diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,124765 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9999278525305724,
+  "global_step": 20790,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.205128205128205e-07,
+      "loss": 1.2097,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.41025641025641e-07,
+      "loss": 3.8474,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.615384615384617e-07,
+      "loss": 2.7615,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.282051282051282e-06,
+      "loss": 4.1008,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6025641025641025e-06,
+      "loss": 2.9705,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9230769230769234e-06,
+      "loss": 3.7029,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.243589743589744e-06,
+      "loss": 3.0706,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.564102564102564e-06,
+      "loss": 3.2559,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.884615384615385e-06,
+      "loss": 2.644,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.205128205128205e-06,
+      "loss": 2.7444,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.525641025641026e-06,
+      "loss": 2.6755,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.846153846153847e-06,
+      "loss": 2.3552,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 2.5999,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.487179487179488e-06,
+      "loss": 2.9094,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.807692307692308e-06,
+      "loss": 2.4919,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.128205128205128e-06,
+      "loss": 2.2581,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.448717948717949e-06,
+      "loss": 2.1614,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.76923076923077e-06,
+      "loss": 2.5745,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.08974358974359e-06,
+      "loss": 2.0354,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.41025641025641e-06,
+      "loss": 2.3157,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.730769230769231e-06,
+      "loss": 2.0914,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.051282051282052e-06,
+      "loss": 2.2717,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.371794871794873e-06,
+      "loss": 1.9167,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.692307692307694e-06,
+      "loss": 2.2988,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.012820512820515e-06,
+      "loss": 2.2263,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.9414,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.653846153846155e-06,
+      "loss": 2.1445,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.974358974358976e-06,
+      "loss": 2.2583,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.294871794871795e-06,
+      "loss": 2.1433,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.615384615384616e-06,
+      "loss": 1.8269,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.935897435897435e-06,
+      "loss": 2.0798,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0256410256410256e-05,
+      "loss": 1.8551,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0576923076923077e-05,
+      "loss": 1.8931,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0897435897435898e-05,
+      "loss": 1.9421,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1217948717948719e-05,
+      "loss": 1.9478,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.153846153846154e-05,
+      "loss": 1.9423,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1858974358974359e-05,
+      "loss": 1.8324,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.217948717948718e-05,
+      "loss": 1.8652,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.25e-05,
+      "loss": 1.9647,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.282051282051282e-05,
+      "loss": 1.7139,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3141025641025642e-05,
+      "loss": 1.724,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3461538461538462e-05,
+      "loss": 1.0638,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3782051282051284e-05,
+      "loss": 1.6985,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4102564102564104e-05,
+      "loss": 1.5717,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4423076923076923e-05,
+      "loss": 1.5884,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4743589743589745e-05,
+      "loss": 1.5651,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5064102564102565e-05,
+      "loss": 1.0618,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5384615384615387e-05,
+      "loss": 1.5717,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5705128205128205e-05,
+      "loss": 1.5131,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.602564102564103e-05,
+      "loss": 1.421,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6346153846153847e-05,
+      "loss": 1.4989,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 1.5144,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.698717948717949e-05,
+      "loss": 1.3035,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.730769230769231e-05,
+      "loss": 1.3749,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.762820512820513e-05,
+      "loss": 1.3035,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.794871794871795e-05,
+      "loss": 1.1986,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.826923076923077e-05,
+      "loss": 1.2202,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.858974358974359e-05,
+      "loss": 1.3077,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.891025641025641e-05,
+      "loss": 1.287,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.923076923076923e-05,
+      "loss": 1.2346,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9551282051282052e-05,
+      "loss": 1.1927,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.987179487179487e-05,
+      "loss": 1.2969,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0192307692307694e-05,
+      "loss": 1.1413,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0512820512820512e-05,
+      "loss": 1.2056,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0833333333333336e-05,
+      "loss": 1.3599,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1153846153846154e-05,
+      "loss": 1.1625,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1474358974358974e-05,
+      "loss": 1.2861,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1794871794871795e-05,
+      "loss": 1.2278,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2115384615384616e-05,
+      "loss": 1.273,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2435897435897437e-05,
+      "loss": 1.2761,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2756410256410258e-05,
+      "loss": 1.1467,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.307692307692308e-05,
+      "loss": 1.2167,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.33974358974359e-05,
+      "loss": 1.2682,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3717948717948718e-05,
+      "loss": 1.2108,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4038461538461542e-05,
+      "loss": 1.2944,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.435897435897436e-05,
+      "loss": 1.2542,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.467948717948718e-05,
+      "loss": 1.3472,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5e-05,
+      "loss": 1.2112,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5320512820512822e-05,
+      "loss": 1.15,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.564102564102564e-05,
+      "loss": 1.2941,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5961538461538464e-05,
+      "loss": 1.2072,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6282051282051285e-05,
+      "loss": 1.2238,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6602564102564102e-05,
+      "loss": 1.1791,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6923076923076923e-05,
+      "loss": 1.1682,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.724358974358974e-05,
+      "loss": 1.2296,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.756410256410257e-05,
+      "loss": 1.2314,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7884615384615386e-05,
+      "loss": 1.0502,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8205128205128207e-05,
+      "loss": 1.1976,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8525641025641025e-05,
+      "loss": 1.3031,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8846153846153845e-05,
+      "loss": 1.2539,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.916666666666667e-05,
+      "loss": 1.233,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.948717948717949e-05,
+      "loss": 1.0952,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9807692307692308e-05,
+      "loss": 1.1255,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.012820512820513e-05,
+      "loss": 1.3066,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0448717948717947e-05,
+      "loss": 1.1212,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0769230769230774e-05,
+      "loss": 1.2925,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.108974358974359e-05,
+      "loss": 1.1298,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.141025641025641e-05,
+      "loss": 1.212,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1730769230769234e-05,
+      "loss": 1.2452,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.205128205128206e-05,
+      "loss": 1.2534,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2371794871794876e-05,
+      "loss": 1.0829,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.269230769230769e-05,
+      "loss": 1.3234,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.301282051282051e-05,
+      "loss": 1.1354,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2908,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.365384615384616e-05,
+      "loss": 1.3258,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.397435897435898e-05,
+      "loss": 1.1371,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4294871794871794e-05,
+      "loss": 1.1572,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.461538461538462e-05,
+      "loss": 1.2714,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4935897435897436e-05,
+      "loss": 1.1824,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.525641025641026e-05,
+      "loss": 1.2281,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.557692307692308e-05,
+      "loss": 1.2307,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.58974358974359e-05,
+      "loss": 1.1691,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.621794871794872e-05,
+      "loss": 0.9863,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.653846153846154e-05,
+      "loss": 1.2526,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.685897435897436e-05,
+      "loss": 1.1141,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.717948717948718e-05,
+      "loss": 1.2239,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 1.2163,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.782051282051282e-05,
+      "loss": 1.1091,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.814102564102564e-05,
+      "loss": 1.5128,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.846153846153846e-05,
+      "loss": 1.2294,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.878205128205129e-05,
+      "loss": 1.3827,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9102564102564105e-05,
+      "loss": 1.0765,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.942307692307692e-05,
+      "loss": 1.2123,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.974358974358974e-05,
+      "loss": 1.2217,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.006410256410257e-05,
+      "loss": 1.2347,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.038461538461539e-05,
+      "loss": 1.1611,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0705128205128206e-05,
+      "loss": 1.1985,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1025641025641023e-05,
+      "loss": 1.2704,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.134615384615385e-05,
+      "loss": 1.2271,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 1.3187,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.198717948717949e-05,
+      "loss": 1.1122,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.230769230769231e-05,
+      "loss": 1.2,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.262820512820513e-05,
+      "loss": 1.3294,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.294871794871795e-05,
+      "loss": 1.2346,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.326923076923077e-05,
+      "loss": 1.131,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.358974358974359e-05,
+      "loss": 1.2383,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3910256410256415e-05,
+      "loss": 1.0711,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.423076923076923e-05,
+      "loss": 1.1963,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.455128205128206e-05,
+      "loss": 1.1652,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4871794871794874e-05,
+      "loss": 1.1675,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.519230769230769e-05,
+      "loss": 1.3281,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5512820512820516e-05,
+      "loss": 1.1197,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5833333333333334e-05,
+      "loss": 1.0944,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.615384615384616e-05,
+      "loss": 1.2092,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6474358974358976e-05,
+      "loss": 1.1639,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.67948717948718e-05,
+      "loss": 1.1274,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.711538461538462e-05,
+      "loss": 1.2219,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7435897435897435e-05,
+      "loss": 1.0591,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.775641025641026e-05,
+      "loss": 1.257,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8076923076923084e-05,
+      "loss": 1.0964,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.83974358974359e-05,
+      "loss": 1.2895,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.871794871794872e-05,
+      "loss": 1.1784,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9038461538461536e-05,
+      "loss": 1.1359,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.935897435897436e-05,
+      "loss": 1.2637,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9679487179487185e-05,
+      "loss": 1.1094,
+      "step": 155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5e-05,
+      "loss": 1.2735,
+      "step": 156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.032051282051282e-05,
+      "loss": 1.1107,
+      "step": 157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0641025641025644e-05,
+      "loss": 1.1062,
+      "step": 158
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.096153846153846e-05,
+      "loss": 1.3615,
+      "step": 159
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.128205128205128e-05,
+      "loss": 1.2217,
+      "step": 160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.160256410256411e-05,
+      "loss": 1.1394,
+      "step": 161
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.192307692307693e-05,
+      "loss": 1.3116,
+      "step": 162
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.224358974358975e-05,
+      "loss": 1.1503,
+      "step": 163
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.256410256410257e-05,
+      "loss": 1.166,
+      "step": 164
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.288461538461539e-05,
+      "loss": 1.2635,
+      "step": 165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3205128205128205e-05,
+      "loss": 1.2399,
+      "step": 166
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.352564102564103e-05,
+      "loss": 1.1425,
+      "step": 167
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.384615384615385e-05,
+      "loss": 1.2505,
+      "step": 168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4166666666666664e-05,
+      "loss": 1.2308,
+      "step": 169
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.448717948717948e-05,
+      "loss": 1.1836,
+      "step": 170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.480769230769231e-05,
+      "loss": 1.2206,
+      "step": 171
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.512820512820514e-05,
+      "loss": 1.1654,
+      "step": 172
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5448717948717955e-05,
+      "loss": 1.2266,
+      "step": 173
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.576923076923077e-05,
+      "loss": 1.2803,
+      "step": 174
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.608974358974359e-05,
+      "loss": 1.3199,
+      "step": 175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6410256410256414e-05,
+      "loss": 1.3012,
+      "step": 176
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.673076923076923e-05,
+      "loss": 1.2215,
+      "step": 177
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.705128205128205e-05,
+      "loss": 1.1546,
+      "step": 178
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.737179487179487e-05,
+      "loss": 1.177,
+      "step": 179
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.769230769230769e-05,
+      "loss": 1.2593,
+      "step": 180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.801282051282052e-05,
+      "loss": 1.2021,
+      "step": 181
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.833333333333334e-05,
+      "loss": 1.2537,
+      "step": 182
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.865384615384616e-05,
+      "loss": 1.2743,
+      "step": 183
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.897435897435898e-05,
+      "loss": 1.162,
+      "step": 184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.92948717948718e-05,
+      "loss": 1.0334,
+      "step": 185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9615384615384616e-05,
+      "loss": 1.202,
+      "step": 186
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9935897435897434e-05,
+      "loss": 1.3008,
+      "step": 187
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.025641025641026e-05,
+      "loss": 1.3156,
+      "step": 188
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0576923076923076e-05,
+      "loss": 1.3701,
+      "step": 189
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.089743589743589e-05,
+      "loss": 1.2094,
+      "step": 190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.121794871794873e-05,
+      "loss": 1.2415,
+      "step": 191
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.153846153846155e-05,
+      "loss": 1.2115,
+      "step": 192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.185897435897437e-05,
+      "loss": 1.3333,
+      "step": 193
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.217948717948718e-05,
+      "loss": 1.2576,
+      "step": 194
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.25e-05,
+      "loss": 1.2723,
+      "step": 195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.282051282051282e-05,
+      "loss": 1.3177,
+      "step": 196
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.314102564102564e-05,
+      "loss": 1.3256,
+      "step": 197
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.346153846153847e-05,
+      "loss": 1.3448,
+      "step": 198
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.378205128205128e-05,
+      "loss": 1.2383,
+      "step": 199
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.410256410256412e-05,
+      "loss": 1.4899,
+      "step": 200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.442307692307693e-05,
+      "loss": 1.2758,
+      "step": 201
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.474358974358975e-05,
+      "loss": 1.1879,
+      "step": 202
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.506410256410257e-05,
+      "loss": 1.313,
+      "step": 203
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.538461538461539e-05,
+      "loss": 1.2532,
+      "step": 204
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.57051282051282e-05,
+      "loss": 1.2519,
+      "step": 205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.602564102564102e-05,
+      "loss": 1.3068,
+      "step": 206
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.634615384615385e-05,
+      "loss": 1.2358,
+      "step": 207
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 1.1265,
+      "step": 208
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.698717948717949e-05,
+      "loss": 1.338,
+      "step": 209
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.730769230769232e-05,
+      "loss": 1.2711,
+      "step": 210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.762820512820514e-05,
+      "loss": 1.2308,
+      "step": 211
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.794871794871795e-05,
+      "loss": 1.2569,
+      "step": 212
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.826923076923077e-05,
+      "loss": 1.2239,
+      "step": 213
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.858974358974359e-05,
+      "loss": 1.265,
+      "step": 214
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.891025641025642e-05,
+      "loss": 1.4725,
+      "step": 215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.923076923076924e-05,
+      "loss": 1.4658,
+      "step": 216
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.955128205128205e-05,
+      "loss": 1.2056,
+      "step": 217
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.987179487179487e-05,
+      "loss": 1.3459,
+      "step": 218
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.019230769230769e-05,
+      "loss": 1.0946,
+      "step": 219
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.051282051282052e-05,
+      "loss": 1.1262,
+      "step": 220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.083333333333334e-05,
+      "loss": 1.1431,
+      "step": 221
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.115384615384616e-05,
+      "loss": 1.2853,
+      "step": 222
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.147435897435897e-05,
+      "loss": 1.1622,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.17948717948718e-05,
+      "loss": 1.4183,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.211538461538462e-05,
+      "loss": 1.3922,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.243589743589744e-05,
+      "loss": 1.3527,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.275641025641026e-05,
+      "loss": 1.2037,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.307692307692307e-05,
+      "loss": 1.2659,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.339743589743589e-05,
+      "loss": 1.2704,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.371794871794872e-05,
+      "loss": 1.2053,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.403846153846154e-05,
+      "loss": 1.2684,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.435897435897436e-05,
+      "loss": 1.2317,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.467948717948719e-05,
+      "loss": 1.197,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.500000000000001e-05,
+      "loss": 1.2866,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.532051282051282e-05,
+      "loss": 1.4103,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.564102564102564e-05,
+      "loss": 1.3437,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.596153846153846e-05,
+      "loss": 1.3799,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.628205128205128e-05,
+      "loss": 1.2673,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.660256410256411e-05,
+      "loss": 1.352,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.692307692307693e-05,
+      "loss": 1.1896,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.724358974358976e-05,
+      "loss": 1.184,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.756410256410257e-05,
+      "loss": 1.3645,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.788461538461539e-05,
+      "loss": 1.1803,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.820512820512821e-05,
+      "loss": 1.3551,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.852564102564103e-05,
+      "loss": 1.3145,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.884615384615384e-05,
+      "loss": 1.2596,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.916666666666666e-05,
+      "loss": 1.1985,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.948717948717948e-05,
+      "loss": 1.101,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.980769230769231e-05,
+      "loss": 1.226,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.012820512820514e-05,
+      "loss": 1.3992,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.044871794871796e-05,
+      "loss": 1.3301,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.076923076923078e-05,
+      "loss": 1.2438,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.10897435897436e-05,
+      "loss": 1.47,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.141025641025641e-05,
+      "loss": 1.1796,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.173076923076923e-05,
+      "loss": 1.4141,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.205128205128205e-05,
+      "loss": 1.4766,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.237179487179488e-05,
+      "loss": 1.4297,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.26923076923077e-05,
+      "loss": 1.3507,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.301282051282053e-05,
+      "loss": 1.2249,
+      "step": 259
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 1.3733,
+      "step": 260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.365384615384616e-05,
+      "loss": 1.3402,
+      "step": 261
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.397435897435898e-05,
+      "loss": 1.2733,
+      "step": 262
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.42948717948718e-05,
+      "loss": 1.2306,
+      "step": 263
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.461538461538461e-05,
+      "loss": 1.1844,
+      "step": 264
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.493589743589743e-05,
+      "loss": 1.1473,
+      "step": 265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.525641025641026e-05,
+      "loss": 1.1098,
+      "step": 266
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.557692307692308e-05,
+      "loss": 1.3178,
+      "step": 267
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.58974358974359e-05,
+      "loss": 1.1595,
+      "step": 268
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.621794871794873e-05,
+      "loss": 1.2321,
+      "step": 269
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.653846153846155e-05,
+      "loss": 1.2149,
+      "step": 270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.685897435897436e-05,
+      "loss": 1.3434,
+      "step": 271
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.717948717948718e-05,
+      "loss": 1.3607,
+      "step": 272
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.75e-05,
+      "loss": 1.424,
+      "step": 273
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.782051282051283e-05,
+      "loss": 1.1359,
+      "step": 274
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.814102564102565e-05,
+      "loss": 1.2603,
+      "step": 275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.846153846153847e-05,
+      "loss": 1.3045,
+      "step": 276
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.878205128205128e-05,
+      "loss": 1.3151,
+      "step": 277
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.910256410256411e-05,
+      "loss": 1.2609,
+      "step": 278
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.942307692307693e-05,
+      "loss": 1.3109,
+      "step": 279
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.974358974358975e-05,
+      "loss": 1.2557,
+      "step": 280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.006410256410257e-05,
+      "loss": 1.3837,
+      "step": 281
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.038461538461538e-05,
+      "loss": 1.4088,
+      "step": 282
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.070512820512822e-05,
+      "loss": 1.4019,
+      "step": 283
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.102564102564103e-05,
+      "loss": 1.3484,
+      "step": 284
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.134615384615385e-05,
+      "loss": 1.3193,
+      "step": 285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.166666666666667e-05,
+      "loss": 1.3019,
+      "step": 286
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.198717948717949e-05,
+      "loss": 1.4363,
+      "step": 287
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.230769230769232e-05,
+      "loss": 1.4053,
+      "step": 288
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.262820512820513e-05,
+      "loss": 1.1002,
+      "step": 289
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.294871794871795e-05,
+      "loss": 1.282,
+      "step": 290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.326923076923077e-05,
+      "loss": 1.3273,
+      "step": 291
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.35897435897436e-05,
+      "loss": 1.1774,
+      "step": 292
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.391025641025642e-05,
+      "loss": 1.3241,
+      "step": 293
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.423076923076924e-05,
+      "loss": 1.2277,
+      "step": 294
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.455128205128205e-05,
+      "loss": 1.4067,
+      "step": 295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.487179487179487e-05,
+      "loss": 1.3493,
+      "step": 296
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.519230769230769e-05,
+      "loss": 1.2564,
+      "step": 297
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.551282051282052e-05,
+      "loss": 1.3539,
+      "step": 298
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.583333333333334e-05,
+      "loss": 1.3604,
+      "step": 299
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.615384615384617e-05,
+      "loss": 1.244,
+      "step": 300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.647435897435898e-05,
+      "loss": 1.2059,
+      "step": 301
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.67948717948718e-05,
+      "loss": 1.3532,
+      "step": 302
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.711538461538462e-05,
+      "loss": 1.3338,
+      "step": 303
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.743589743589744e-05,
+      "loss": 1.5115,
+      "step": 304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.775641025641025e-05,
+      "loss": 1.2999,
+      "step": 305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.807692307692307e-05,
+      "loss": 1.1766,
+      "step": 306
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.839743589743589e-05,
+      "loss": 1.4078,
+      "step": 307
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.871794871794872e-05,
+      "loss": 1.2423,
+      "step": 308
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903846153846155e-05,
+      "loss": 1.2604,
+      "step": 309
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.935897435897437e-05,
+      "loss": 1.4801,
+      "step": 310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.967948717948719e-05,
+      "loss": 1.1519,
+      "step": 311
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 1.2312,
+      "step": 312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010032051282051282,
+      "loss": 1.3387,
+      "step": 313
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010064102564102564,
+      "loss": 1.4274,
+      "step": 314
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010096153846153846,
+      "loss": 1.316,
+      "step": 315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010128205128205129,
+      "loss": 1.2845,
+      "step": 316
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001016025641025641,
+      "loss": 1.2109,
+      "step": 317
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010192307692307692,
+      "loss": 1.2597,
+      "step": 318
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010224358974358974,
+      "loss": 1.3304,
+      "step": 319
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010256410256410256,
+      "loss": 1.3048,
+      "step": 320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010288461538461538,
+      "loss": 1.3345,
+      "step": 321
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010320512820512822,
+      "loss": 1.3448,
+      "step": 322
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010352564102564104,
+      "loss": 1.4158,
+      "step": 323
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010384615384615386,
+      "loss": 1.3158,
+      "step": 324
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010416666666666667,
+      "loss": 1.3131,
+      "step": 325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001044871794871795,
+      "loss": 1.2463,
+      "step": 326
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010480769230769232,
+      "loss": 1.2988,
+      "step": 327
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010512820512820514,
+      "loss": 1.2376,
+      "step": 328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010544871794871796,
+      "loss": 1.5585,
+      "step": 329
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010576923076923077,
+      "loss": 1.2731,
+      "step": 330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010608974358974359,
+      "loss": 1.4762,
+      "step": 331
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010641025641025641,
+      "loss": 1.2585,
+      "step": 332
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010673076923076924,
+      "loss": 1.4725,
+      "step": 333
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010705128205128206,
+      "loss": 1.3646,
+      "step": 334
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010737179487179488,
+      "loss": 1.3685,
+      "step": 335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001076923076923077,
+      "loss": 1.2375,
+      "step": 336
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010801282051282051,
+      "loss": 1.3247,
+      "step": 337
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010833333333333333,
+      "loss": 1.4144,
+      "step": 338
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010865384615384615,
+      "loss": 1.4942,
+      "step": 339
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010897435897435896,
+      "loss": 1.3749,
+      "step": 340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001092948717948718,
+      "loss": 1.195,
+      "step": 341
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010961538461538463,
+      "loss": 1.1417,
+      "step": 342
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010993589743589746,
+      "loss": 1.3799,
+      "step": 343
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011025641025641027,
+      "loss": 1.3052,
+      "step": 344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011057692307692309,
+      "loss": 1.3634,
+      "step": 345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011089743589743591,
+      "loss": 1.4243,
+      "step": 346
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011121794871794873,
+      "loss": 1.2507,
+      "step": 347
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011153846153846154,
+      "loss": 1.3423,
+      "step": 348
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011185897435897436,
+      "loss": 1.2177,
+      "step": 349
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011217948717948718,
+      "loss": 1.3934,
+      "step": 350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011250000000000001,
+      "loss": 1.3373,
+      "step": 351
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011282051282051283,
+      "loss": 1.3648,
+      "step": 352
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011314102564102565,
+      "loss": 1.3414,
+      "step": 353
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011346153846153846,
+      "loss": 1.4576,
+      "step": 354
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011378205128205128,
+      "loss": 1.436,
+      "step": 355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001141025641025641,
+      "loss": 1.2189,
+      "step": 356
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011442307692307692,
+      "loss": 1.1581,
+      "step": 357
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011474358974358975,
+      "loss": 1.3297,
+      "step": 358
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011506410256410256,
+      "loss": 1.3185,
+      "step": 359
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011538461538461538,
+      "loss": 1.1854,
+      "step": 360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011570512820512823,
+      "loss": 1.2455,
+      "step": 361
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011602564102564104,
+      "loss": 1.3552,
+      "step": 362
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011634615384615386,
+      "loss": 1.1776,
+      "step": 363
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 1.136,
+      "step": 364
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001169871794871795,
+      "loss": 1.3972,
+      "step": 365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011730769230769231,
+      "loss": 1.2357,
+      "step": 366
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011762820512820513,
+      "loss": 1.4335,
+      "step": 367
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011794871794871796,
+      "loss": 1.2301,
+      "step": 368
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011826923076923078,
+      "loss": 1.2747,
+      "step": 369
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001185897435897436,
+      "loss": 1.2722,
+      "step": 370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011891025641025642,
+      "loss": 1.288,
+      "step": 371
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011923076923076923,
+      "loss": 1.1549,
+      "step": 372
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011955128205128205,
+      "loss": 1.3526,
+      "step": 373
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011987179487179487,
+      "loss": 1.4153,
+      "step": 374
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001201923076923077,
+      "loss": 1.3292,
+      "step": 375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012051282051282052,
+      "loss": 1.3436,
+      "step": 376
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012083333333333333,
+      "loss": 1.2394,
+      "step": 377
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012115384615384615,
+      "loss": 1.2216,
+      "step": 378
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012147435897435897,
+      "loss": 1.2316,
+      "step": 379
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012179487179487179,
+      "loss": 1.5677,
+      "step": 380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012211538461538463,
+      "loss": 1.3278,
+      "step": 381
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012243589743589746,
+      "loss": 1.3874,
+      "step": 382
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012275641025641027,
+      "loss": 1.2203,
+      "step": 383
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001230769230769231,
+      "loss": 1.4141,
+      "step": 384
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001233974358974359,
+      "loss": 1.4352,
+      "step": 385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012371794871794873,
+      "loss": 1.2845,
+      "step": 386
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012403846153846154,
+      "loss": 1.2909,
+      "step": 387
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012435897435897437,
+      "loss": 1.3852,
+      "step": 388
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001246794871794872,
+      "loss": 1.4934,
+      "step": 389
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000125,
+      "loss": 1.4709,
+      "step": 390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012532051282051283,
+      "loss": 1.3496,
+      "step": 391
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012564102564102564,
+      "loss": 1.3711,
+      "step": 392
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012596153846153847,
+      "loss": 1.2912,
+      "step": 393
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012628205128205127,
+      "loss": 1.3054,
+      "step": 394
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001266025641025641,
+      "loss": 1.321,
+      "step": 395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012692307692307693,
+      "loss": 1.3116,
+      "step": 396
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012724358974358974,
+      "loss": 1.4526,
+      "step": 397
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012756410256410257,
+      "loss": 1.38,
+      "step": 398
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012788461538461537,
+      "loss": 1.2608,
+      "step": 399
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012820512820512823,
+      "loss": 1.5284,
+      "step": 400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012852564102564104,
+      "loss": 1.2776,
+      "step": 401
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012884615384615387,
+      "loss": 1.3851,
+      "step": 402
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012916666666666667,
+      "loss": 1.677,
+      "step": 403
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001294871794871795,
+      "loss": 1.2305,
+      "step": 404
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001298076923076923,
+      "loss": 1.349,
+      "step": 405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013012820512820514,
+      "loss": 1.6062,
+      "step": 406
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013044871794871797,
+      "loss": 1.3967,
+      "step": 407
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013076923076923077,
+      "loss": 1.3998,
+      "step": 408
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001310897435897436,
+      "loss": 1.3334,
+      "step": 409
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001314102564102564,
+      "loss": 1.4702,
+      "step": 410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013173076923076924,
+      "loss": 1.2489,
+      "step": 411
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013205128205128204,
+      "loss": 2.2756,
+      "step": 412
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013237179487179487,
+      "loss": 1.3003,
+      "step": 413
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001326923076923077,
+      "loss": 1.3096,
+      "step": 414
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001330128205128205,
+      "loss": 1.3901,
+      "step": 415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 1.5628,
+      "step": 416
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013365384615384614,
+      "loss": 1.4507,
+      "step": 417
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013397435897435897,
+      "loss": 1.52,
+      "step": 418
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013429487179487178,
+      "loss": 1.3978,
+      "step": 419
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013461538461538464,
+      "loss": 1.3574,
+      "step": 420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013493589743589744,
+      "loss": 1.3671,
+      "step": 421
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013525641025641027,
+      "loss": 1.3524,
+      "step": 422
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001355769230769231,
+      "loss": 1.4384,
+      "step": 423
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001358974358974359,
+      "loss": 1.5883,
+      "step": 424
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013621794871794874,
+      "loss": 1.3484,
+      "step": 425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013653846153846154,
+      "loss": 1.3143,
+      "step": 426
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013685897435897437,
+      "loss": 1.3557,
+      "step": 427
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013717948717948718,
+      "loss": 1.3649,
+      "step": 428
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001375,
+      "loss": 1.5351,
+      "step": 429
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013782051282051284,
+      "loss": 1.4175,
+      "step": 430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013814102564102564,
+      "loss": 1.3977,
+      "step": 431
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013846153846153847,
+      "loss": 1.5039,
+      "step": 432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013878205128205128,
+      "loss": 1.428,
+      "step": 433
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001391025641025641,
+      "loss": 1.6113,
+      "step": 434
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001394230769230769,
+      "loss": 1.4344,
+      "step": 435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013974358974358974,
+      "loss": 2.592,
+      "step": 436
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014006410256410255,
+      "loss": 1.4396,
+      "step": 437
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014038461538461538,
+      "loss": 1.3906,
+      "step": 438
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001407051282051282,
+      "loss": 1.3861,
+      "step": 439
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014102564102564104,
+      "loss": 1.3002,
+      "step": 440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014134615384615387,
+      "loss": 1.4088,
+      "step": 441
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014166666666666668,
+      "loss": 1.311,
+      "step": 442
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001419871794871795,
+      "loss": 1.4863,
+      "step": 443
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001423076923076923,
+      "loss": 1.3951,
+      "step": 444
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014262820512820514,
+      "loss": 1.5728,
+      "step": 445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014294871794871795,
+      "loss": 1.432,
+      "step": 446
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014326923076923078,
+      "loss": 1.4373,
+      "step": 447
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001435897435897436,
+      "loss": 2.6995,
+      "step": 448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001439102564102564,
+      "loss": 1.463,
+      "step": 449
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014423076923076924,
+      "loss": 1.505,
+      "step": 450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014455128205128205,
+      "loss": 1.5612,
+      "step": 451
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014487179487179488,
+      "loss": 1.5046,
+      "step": 452
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014519230769230768,
+      "loss": 1.6698,
+      "step": 453
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014551282051282051,
+      "loss": 1.4561,
+      "step": 454
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014583333333333335,
+      "loss": 1.2006,
+      "step": 455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014615384615384615,
+      "loss": 1.5446,
+      "step": 456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014647435897435898,
+      "loss": 1.6074,
+      "step": 457
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014679487179487178,
+      "loss": 1.4962,
+      "step": 458
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014711538461538464,
+      "loss": 1.5663,
+      "step": 459
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014743589743589745,
+      "loss": 1.3107,
+      "step": 460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014775641025641028,
+      "loss": 1.3486,
+      "step": 461
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014807692307692308,
+      "loss": 1.3926,
+      "step": 462
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001483974358974359,
+      "loss": 1.294,
+      "step": 463
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014871794871794872,
+      "loss": 1.3783,
+      "step": 464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014903846153846155,
+      "loss": 2.7993,
+      "step": 465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014935897435897438,
+      "loss": 1.4825,
+      "step": 466
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014967948717948718,
+      "loss": 1.4956,
+      "step": 467
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 1.38,
+      "step": 468
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015032051282051282,
+      "loss": 1.3677,
+      "step": 469
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015064102564102565,
+      "loss": 1.3489,
+      "step": 470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015096153846153845,
+      "loss": 1.3956,
+      "step": 471
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015128205128205128,
+      "loss": 1.3643,
+      "step": 472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015160256410256412,
+      "loss": 1.3514,
+      "step": 473
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015192307692307692,
+      "loss": 1.4231,
+      "step": 474
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015224358974358975,
+      "loss": 1.5139,
+      "step": 475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015256410256410255,
+      "loss": 1.4025,
+      "step": 476
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015288461538461539,
+      "loss": 1.5206,
+      "step": 477
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015320512820512822,
+      "loss": 1.3357,
+      "step": 478
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015352564102564105,
+      "loss": 1.3908,
+      "step": 479
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015384615384615385,
+      "loss": 1.5217,
+      "step": 480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015416666666666668,
+      "loss": 1.4873,
+      "step": 481
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015448717948717951,
+      "loss": 1.4792,
+      "step": 482
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015480769230769232,
+      "loss": 1.4873,
+      "step": 483
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015512820512820515,
+      "loss": 1.3326,
+      "step": 484
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015544871794871795,
+      "loss": 1.3914,
+      "step": 485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015576923076923078,
+      "loss": 1.4803,
+      "step": 486
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001560897435897436,
+      "loss": 1.4568,
+      "step": 487
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015641025641025642,
+      "loss": 1.356,
+      "step": 488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015673076923076925,
+      "loss": 1.5001,
+      "step": 489
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015705128205128205,
+      "loss": 1.3523,
+      "step": 490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015737179487179488,
+      "loss": 1.3945,
+      "step": 491
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001576923076923077,
+      "loss": 1.6667,
+      "step": 492
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015801282051282052,
+      "loss": 1.5746,
+      "step": 493
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015833333333333332,
+      "loss": 1.411,
+      "step": 494
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015865384615384616,
+      "loss": 1.4044,
+      "step": 495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015897435897435896,
+      "loss": 1.3642,
+      "step": 496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001592948717948718,
+      "loss": 1.4652,
+      "step": 497
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015961538461538462,
+      "loss": 1.4265,
+      "step": 498
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015993589743589745,
+      "loss": 3.2566,
+      "step": 499
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016025641025641028,
+      "loss": 1.5878,
+      "step": 500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001605769230769231,
+      "loss": 1.2536,
+      "step": 501
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016089743589743592,
+      "loss": 1.6327,
+      "step": 502
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016121794871794872,
+      "loss": 1.5992,
+      "step": 503
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016153846153846155,
+      "loss": 1.3093,
+      "step": 504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016185897435897436,
+      "loss": 1.4969,
+      "step": 505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001621794871794872,
+      "loss": 1.644,
+      "step": 506
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 1.4565,
+      "step": 507
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016282051282051282,
+      "loss": 1.537,
+      "step": 508
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016314102564102565,
+      "loss": 1.3722,
+      "step": 509
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016346153846153846,
+      "loss": 1.5176,
+      "step": 510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001637820512820513,
+      "loss": 1.4415,
+      "step": 511
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001641025641025641,
+      "loss": 1.5892,
+      "step": 512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016442307692307692,
+      "loss": 1.5308,
+      "step": 513
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016474358974358976,
+      "loss": 1.5605,
+      "step": 514
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016506410256410256,
+      "loss": 1.4969,
+      "step": 515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001653846153846154,
+      "loss": 1.5289,
+      "step": 516
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016570512820512822,
+      "loss": 1.4208,
+      "step": 517
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016602564102564105,
+      "loss": 1.4248,
+      "step": 518
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016634615384615386,
+      "loss": 1.6902,
+      "step": 519
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 1.5161,
+      "step": 520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001669871794871795,
+      "loss": 1.4665,
+      "step": 521
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016730769230769232,
+      "loss": 1.4636,
+      "step": 522
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016762820512820513,
+      "loss": 1.2588,
+      "step": 523
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016794871794871796,
+      "loss": 1.5936,
+      "step": 524
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001682692307692308,
+      "loss": 1.5758,
+      "step": 525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001685897435897436,
+      "loss": 1.4481,
+      "step": 526
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016891025641025642,
+      "loss": 1.3649,
+      "step": 527
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016923076923076923,
+      "loss": 1.3674,
+      "step": 528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016955128205128206,
+      "loss": 1.5795,
+      "step": 529
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016987179487179486,
+      "loss": 1.3672,
+      "step": 530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001701923076923077,
+      "loss": 1.5421,
+      "step": 531
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017051282051282053,
+      "loss": 1.3828,
+      "step": 532
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017083333333333333,
+      "loss": 1.4669,
+      "step": 533
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017115384615384616,
+      "loss": 2.9575,
+      "step": 534
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017147435897435896,
+      "loss": 1.4958,
+      "step": 535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001717948717948718,
+      "loss": 1.5287,
+      "step": 536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017211538461538463,
+      "loss": 1.4166,
+      "step": 537
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017243589743589746,
+      "loss": 3.1895,
+      "step": 538
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017275641025641026,
+      "loss": 1.3806,
+      "step": 539
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001730769230769231,
+      "loss": 1.3597,
+      "step": 540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017339743589743592,
+      "loss": 1.4064,
+      "step": 541
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017371794871794873,
+      "loss": 1.4582,
+      "step": 542
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017403846153846156,
+      "loss": 1.3268,
+      "step": 543
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017435897435897436,
+      "loss": 1.5826,
+      "step": 544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001746794871794872,
+      "loss": 1.4862,
+      "step": 545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000175,
+      "loss": 1.5404,
+      "step": 546
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017532051282051283,
+      "loss": 2.9912,
+      "step": 547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017564102564102566,
+      "loss": 1.5543,
+      "step": 548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017596153846153846,
+      "loss": 1.5145,
+      "step": 549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001762820512820513,
+      "loss": 1.4047,
+      "step": 550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001766025641025641,
+      "loss": 1.4511,
+      "step": 551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017692307692307693,
+      "loss": 4.3474,
+      "step": 552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017724358974358973,
+      "loss": 1.4698,
+      "step": 553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017756410256410257,
+      "loss": 1.6018,
+      "step": 554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017788461538461537,
+      "loss": 1.511,
+      "step": 555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017820512820512823,
+      "loss": 1.5264,
+      "step": 556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017852564102564103,
+      "loss": 1.5397,
+      "step": 557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017884615384615386,
+      "loss": 1.5033,
+      "step": 558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001791666666666667,
+      "loss": 3.6326,
+      "step": 559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001794871794871795,
+      "loss": 1.4312,
+      "step": 560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017980769230769233,
+      "loss": 1.5906,
+      "step": 561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018012820512820513,
+      "loss": 1.4738,
+      "step": 562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018044871794871796,
+      "loss": 1.4945,
+      "step": 563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018076923076923077,
+      "loss": 1.5715,
+      "step": 564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001810897435897436,
+      "loss": 1.7129,
+      "step": 565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018141025641025643,
+      "loss": 1.6078,
+      "step": 566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018173076923076923,
+      "loss": 1.5311,
+      "step": 567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018205128205128207,
+      "loss": 1.4482,
+      "step": 568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018237179487179487,
+      "loss": 1.5851,
+      "step": 569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001826923076923077,
+      "loss": 1.5714,
+      "step": 570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001830128205128205,
+      "loss": 1.522,
+      "step": 571
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 1.4672,
+      "step": 572
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018365384615384617,
+      "loss": 3.3252,
+      "step": 573
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018397435897435897,
+      "loss": 1.4758,
+      "step": 574
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001842948717948718,
+      "loss": 1.6704,
+      "step": 575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018461538461538463,
+      "loss": 1.3447,
+      "step": 576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018493589743589746,
+      "loss": 2.8279,
+      "step": 577
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018525641025641027,
+      "loss": 1.4277,
+      "step": 578
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001855769230769231,
+      "loss": 1.5401,
+      "step": 579
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001858974358974359,
+      "loss": 1.4904,
+      "step": 580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018621794871794873,
+      "loss": 1.6947,
+      "step": 581
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018653846153846154,
+      "loss": 1.6758,
+      "step": 582
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018685897435897437,
+      "loss": 1.5337,
+      "step": 583
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001871794871794872,
+      "loss": 1.4462,
+      "step": 584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001875,
+      "loss": 1.4485,
+      "step": 585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018782051282051283,
+      "loss": 1.531,
+      "step": 586
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018814102564102564,
+      "loss": 1.6816,
+      "step": 587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018846153846153847,
+      "loss": 1.3767,
+      "step": 588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018878205128205127,
+      "loss": 1.6293,
+      "step": 589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001891025641025641,
+      "loss": 1.5774,
+      "step": 590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018942307692307694,
+      "loss": 1.3569,
+      "step": 591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018974358974358974,
+      "loss": 2.979,
+      "step": 592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019006410256410257,
+      "loss": 1.4167,
+      "step": 593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019038461538461538,
+      "loss": 1.631,
+      "step": 594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019070512820512823,
+      "loss": 1.4644,
+      "step": 595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019102564102564104,
+      "loss": 1.4549,
+      "step": 596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019134615384615387,
+      "loss": 1.3843,
+      "step": 597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019166666666666667,
+      "loss": 1.3257,
+      "step": 598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001919871794871795,
+      "loss": 1.62,
+      "step": 599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019230769230769233,
+      "loss": 1.5458,
+      "step": 600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019262820512820514,
+      "loss": 1.5022,
+      "step": 601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019294871794871797,
+      "loss": 1.421,
+      "step": 602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019326923076923077,
+      "loss": 1.4482,
+      "step": 603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001935897435897436,
+      "loss": 1.5341,
+      "step": 604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001939102564102564,
+      "loss": 1.6016,
+      "step": 605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019423076923076924,
+      "loss": 1.5006,
+      "step": 606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019455128205128207,
+      "loss": 1.6342,
+      "step": 607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019487179487179487,
+      "loss": 1.373,
+      "step": 608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001951923076923077,
+      "loss": 1.5488,
+      "step": 609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001955128205128205,
+      "loss": 1.3313,
+      "step": 610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019583333333333334,
+      "loss": 1.4532,
+      "step": 611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019615384615384615,
+      "loss": 1.6852,
+      "step": 612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019647435897435898,
+      "loss": 1.4606,
+      "step": 613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019679487179487178,
+      "loss": 1.3102,
+      "step": 614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019711538461538464,
+      "loss": 1.5806,
+      "step": 615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019743589743589744,
+      "loss": 1.3732,
+      "step": 616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019775641025641027,
+      "loss": 1.5289,
+      "step": 617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001980769230769231,
+      "loss": 1.3823,
+      "step": 618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001983974358974359,
+      "loss": 1.6394,
+      "step": 619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019871794871794874,
+      "loss": 1.5786,
+      "step": 620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019903846153846154,
+      "loss": 1.4338,
+      "step": 621
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019935897435897437,
+      "loss": 1.6427,
+      "step": 622
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019967948717948718,
+      "loss": 1.626,
+      "step": 623
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 1.4133,
+      "step": 624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999999987865267,
+      "loss": 1.58,
+      "step": 625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999999514610684,
+      "loss": 1.2836,
+      "step": 626
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999998907874053,
+      "loss": 1.3904,
+      "step": 627
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999999805844279,
+      "loss": 1.4369,
+      "step": 628
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999996966316912,
+      "loss": 1.5276,
+      "step": 629
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999999563149645,
+      "loss": 1.3872,
+      "step": 630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999994053981435,
+      "loss": 1.3025,
+      "step": 631
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999992233771905,
+      "loss": 1.4634,
+      "step": 632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999990170867905,
+      "loss": 2.8477,
+      "step": 633
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999987865269484,
+      "loss": 1.3983,
+      "step": 634
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999985316976698,
+      "loss": 1.4039,
+      "step": 635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999998252598961,
+      "loss": 1.4847,
+      "step": 636
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999979492308286,
+      "loss": 1.527,
+      "step": 637
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999976215932802,
+      "loss": 1.4094,
+      "step": 638
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999972696863237,
+      "loss": 1.3877,
+      "step": 639
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999968935099674,
+      "loss": 1.4311,
+      "step": 640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999996493064221,
+      "loss": 1.3539,
+      "step": 641
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999960683490936,
+      "loss": 1.5114,
+      "step": 642
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999956193645954,
+      "loss": 1.6512,
+      "step": 643
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999995146110738,
+      "loss": 1.3157,
+      "step": 644
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999946485875324,
+      "loss": 1.4208,
+      "step": 645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999941267949907,
+      "loss": 1.2658,
+      "step": 646
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999935807331256,
+      "loss": 1.4424,
+      "step": 647
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999930104019504,
+      "loss": 1.4415,
+      "step": 648
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999924158014792,
+      "loss": 1.5613,
+      "step": 649
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999917969317258,
+      "loss": 1.4696,
+      "step": 650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999911537927058,
+      "loss": 1.5459,
+      "step": 651
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999904863844345,
+      "loss": 1.4451,
+      "step": 652
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999897947069287,
+      "loss": 1.6232,
+      "step": 653
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999989078760204,
+      "loss": 1.4712,
+      "step": 654
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999883385442787,
+      "loss": 1.512,
+      "step": 655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999875740591704,
+      "loss": 1.5485,
+      "step": 656
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999986785304898,
+      "loss": 1.5325,
+      "step": 657
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999859722814804,
+      "loss": 1.5182,
+      "step": 658
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999851349889375,
+      "loss": 1.4222,
+      "step": 659
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999842734272892,
+      "loss": 1.6524,
+      "step": 660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999833875965567,
+      "loss": 1.3808,
+      "step": 661
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999824774967614,
+      "loss": 2.9238,
+      "step": 662
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999815431279256,
+      "loss": 1.5963,
+      "step": 663
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999980584490072,
+      "loss": 1.402,
+      "step": 664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999796015832233,
+      "loss": 1.4633,
+      "step": 665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999785944074038,
+      "loss": 1.5857,
+      "step": 666
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999775629626383,
+      "loss": 1.5887,
+      "step": 667
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999976507248951,
+      "loss": 1.4605,
+      "step": 668
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999754272663684,
+      "loss": 1.4944,
+      "step": 669
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999974323014916,
+      "loss": 1.3303,
+      "step": 670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999731944946213,
+      "loss": 1.7747,
+      "step": 671
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999972041705511,
+      "loss": 1.4428,
+      "step": 672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999708646476128,
+      "loss": 1.4781,
+      "step": 673
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019999696633209564,
+      "loss": 1.6444,
+      "step": 674
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199996843772557,
+      "loss": 1.489,
+      "step": 675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999671878614841,
+      "loss": 1.5785,
+      "step": 676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999659137287285,
+      "loss": 1.5798,
+      "step": 677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999646153273342,
+      "loss": 1.4347,
+      "step": 678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999632926573328,
+      "loss": 1.4399,
+      "step": 679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999619457187562,
+      "loss": 1.6311,
+      "step": 680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999605745116375,
+      "loss": 1.5757,
+      "step": 681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999591790360096,
+      "loss": 1.6315,
+      "step": 682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999577592919065,
+      "loss": 1.6035,
+      "step": 683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999563152793627,
+      "loss": 1.4647,
+      "step": 684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999548469984134,
+      "loss": 1.452,
+      "step": 685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999533544490938,
+      "loss": 1.5322,
+      "step": 686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999518376314406,
+      "loss": 1.4581,
+      "step": 687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199995029654549,
+      "loss": 1.4832,
+      "step": 688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199994873119128,
+      "loss": 1.4857,
+      "step": 689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999471415688486,
+      "loss": 1.4626,
+      "step": 690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999455276782337,
+      "loss": 1.3961,
+      "step": 691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999943889519475,
+      "loss": 1.4093,
+      "step": 692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999422270926124,
+      "loss": 1.6358,
+      "step": 693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999405403976861,
+      "loss": 1.5935,
+      "step": 694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999388294347369,
+      "loss": 1.6075,
+      "step": 695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999370942038063,
+      "loss": 1.6202,
+      "step": 696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999353347049364,
+      "loss": 1.4464,
+      "step": 697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199993355093817,
+      "loss": 1.5712,
+      "step": 698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999317429035505,
+      "loss": 1.4066,
+      "step": 699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999299106011214,
+      "loss": 1.4241,
+      "step": 700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999928054030928,
+      "loss": 1.5084,
+      "step": 701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999926173193014,
+      "loss": 1.5668,
+      "step": 702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999242680874266,
+      "loss": 1.4658,
+      "step": 703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999922338714211,
+      "loss": 1.5093,
+      "step": 704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999203850734143,
+      "loss": 1.4899,
+      "step": 705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999184071650838,
+      "loss": 1.4249,
+      "step": 706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999164049892677,
+      "loss": 1.3561,
+      "step": 707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999143785460146,
+      "loss": 1.7449,
+      "step": 708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999123278353735,
+      "loss": 1.41,
+      "step": 709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999102528573943,
+      "loss": 1.5277,
+      "step": 710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999081536121273,
+      "loss": 1.4089,
+      "step": 711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999060300996234,
+      "loss": 1.5157,
+      "step": 712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019999038823199344,
+      "loss": 1.7435,
+      "step": 713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999901710273112,
+      "loss": 2.8894,
+      "step": 714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998995139592092,
+      "loss": 1.5248,
+      "step": 715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998972933782792,
+      "loss": 1.4646,
+      "step": 716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998950485303763,
+      "loss": 1.5674,
+      "step": 717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998927794155543,
+      "loss": 2.5623,
+      "step": 718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998904860338686,
+      "loss": 1.4246,
+      "step": 719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999888168385375,
+      "loss": 1.5161,
+      "step": 720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998858264701292,
+      "loss": 1.5895,
+      "step": 721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999883460288189,
+      "loss": 1.4507,
+      "step": 722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999881069839611,
+      "loss": 1.7253,
+      "step": 723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998786551244532,
+      "loss": 1.3788,
+      "step": 724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999876216142775,
+      "loss": 1.7255,
+      "step": 725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998737528946348,
+      "loss": 3.1646,
+      "step": 726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998712653800923,
+      "loss": 1.4082,
+      "step": 727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998687535992087,
+      "loss": 1.7003,
+      "step": 728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998662175520443,
+      "loss": 1.5866,
+      "step": 729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999863657238661,
+      "loss": 1.3619,
+      "step": 730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998610726591206,
+      "loss": 1.7402,
+      "step": 731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998584638134862,
+      "loss": 1.4684,
+      "step": 732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998558307018208,
+      "loss": 1.4713,
+      "step": 733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998531733241882,
+      "loss": 2.7832,
+      "step": 734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998504916806536,
+      "loss": 1.6801,
+      "step": 735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998477857712814,
+      "loss": 1.6721,
+      "step": 736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998450555961373,
+      "loss": 1.4811,
+      "step": 737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998423011552877,
+      "loss": 1.5057,
+      "step": 738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998395224487997,
+      "loss": 1.6306,
+      "step": 739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998367194767403,
+      "loss": 2.844,
+      "step": 740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998338922391778,
+      "loss": 1.5182,
+      "step": 741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998310407361806,
+      "loss": 1.6722,
+      "step": 742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998281649678183,
+      "loss": 1.5746,
+      "step": 743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998252649341605,
+      "loss": 1.5643,
+      "step": 744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998223406352775,
+      "loss": 1.4564,
+      "step": 745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199981939207124,
+      "loss": 1.5887,
+      "step": 746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199981641924212,
+      "loss": 1.5804,
+      "step": 747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998134221479896,
+      "loss": 1.5279,
+      "step": 748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998104007889215,
+      "loss": 1.4505,
+      "step": 749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998073551649888,
+      "loss": 1.4658,
+      "step": 750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998042852762657,
+      "loss": 1.4562,
+      "step": 751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019998011911228267,
+      "loss": 1.5229,
+      "step": 752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997980727047465,
+      "loss": 1.4885,
+      "step": 753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997949300221013,
+      "loss": 1.4441,
+      "step": 754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999791763074967,
+      "loss": 1.5632,
+      "step": 755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997885718634208,
+      "loss": 1.5361,
+      "step": 756
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199978535638754,
+      "loss": 1.4656,
+      "step": 757
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997821166474022,
+      "loss": 1.596,
+      "step": 758
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999778852643087,
+      "loss": 1.6135,
+      "step": 759
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997755643746726,
+      "loss": 1.3664,
+      "step": 760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997722518422391,
+      "loss": 1.5856,
+      "step": 761
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997689150458674,
+      "loss": 1.5898,
+      "step": 762
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999765553985638,
+      "loss": 1.4598,
+      "step": 763
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997621686616322,
+      "loss": 1.4672,
+      "step": 764
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999758759073933,
+      "loss": 1.3257,
+      "step": 765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997553252226226,
+      "loss": 1.5513,
+      "step": 766
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997518671077845,
+      "loss": 1.4875,
+      "step": 767
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997483847295023,
+      "loss": 1.3716,
+      "step": 768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997448780878613,
+      "loss": 1.4734,
+      "step": 769
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997413471829458,
+      "loss": 1.4241,
+      "step": 770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997377920148414,
+      "loss": 1.4582,
+      "step": 771
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997342125836352,
+      "loss": 1.4074,
+      "step": 772
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997306088894137,
+      "loss": 1.3172,
+      "step": 773
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997269809322638,
+      "loss": 1.618,
+      "step": 774
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997233287122744,
+      "loss": 1.4379,
+      "step": 775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997196522295336,
+      "loss": 1.5123,
+      "step": 776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997159514841307,
+      "loss": 1.4014,
+      "step": 777
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019997122264761558,
+      "loss": 1.568,
+      "step": 778
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001999708477205699,
+      "loss": 1.4359,
+      "step": 779
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019997047036728516,
+      "loss": 1.4312,
+      "step": 780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999700905877705,
+      "loss": 1.5643,
+      "step": 781
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996970838203511,
+      "loss": 3.0566,
+      "step": 782
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996932375008833,
+      "loss": 1.5677,
+      "step": 783
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996893669193942,
+      "loss": 1.4764,
+      "step": 784
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999685472075978,
+      "loss": 1.523,
+      "step": 785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996815529707297,
+      "loss": 1.4725,
+      "step": 786
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999677609603744,
+      "loss": 1.4398,
+      "step": 787
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996736419751162,
+      "loss": 1.5062,
+      "step": 788
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996696500849435,
+      "loss": 1.4377,
+      "step": 789
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996656339333222,
+      "loss": 1.3828,
+      "step": 790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000199966159352035,
+      "loss": 1.3472,
+      "step": 791
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996575288461248,
+      "loss": 1.5846,
+      "step": 792
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999653439910745,
+      "loss": 1.3438,
+      "step": 793
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996493267143107,
+      "loss": 1.4521,
+      "step": 794
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996451892569206,
+      "loss": 2.8743,
+      "step": 795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999641027538676,
+      "loss": 1.5577,
+      "step": 796
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996368415596774,
+      "loss": 1.4708,
+      "step": 797
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996326313200267,
+      "loss": 1.6376,
+      "step": 798
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999628396819826,
+      "loss": 1.5259,
+      "step": 799
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999624138059178,
+      "loss": 1.5378,
+      "step": 800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996198550381857,
+      "loss": 1.4274,
+      "step": 801
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996155477569537,
+      "loss": 1.6514,
+      "step": 802
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996112162155863,
+      "loss": 1.5749,
+      "step": 803
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996068604141885,
+      "loss": 1.6273,
+      "step": 804
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019996024803528662,
+      "loss": 1.4516,
+      "step": 805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995980760317256,
+      "loss": 1.4517,
+      "step": 806
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995936474508734,
+      "loss": 1.3772,
+      "step": 807
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995891946104175,
+      "loss": 1.4654,
+      "step": 808
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995847175104654,
+      "loss": 1.3627,
+      "step": 809
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995802161511264,
+      "loss": 1.2999,
+      "step": 810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995756905325093,
+      "loss": 3.0757,
+      "step": 811
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995711406547238,
+      "loss": 1.4835,
+      "step": 812
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995665665178811,
+      "loss": 1.4625,
+      "step": 813
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995619681220913,
+      "loss": 1.3574,
+      "step": 814
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995573454674668,
+      "loss": 1.582,
+      "step": 815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995526985541192,
+      "loss": 1.3428,
+      "step": 816
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995480273821614,
+      "loss": 1.5225,
+      "step": 817
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995433319517066,
+      "loss": 1.5199,
+      "step": 818
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995386122628696,
+      "loss": 1.4183,
+      "step": 819
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995338683157638,
+      "loss": 1.5167,
+      "step": 820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995291001105052,
+      "loss": 1.307,
+      "step": 821
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995243076472092,
+      "loss": 1.3964,
+      "step": 822
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995194909259921,
+      "loss": 1.5752,
+      "step": 823
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995146499469707,
+      "loss": 1.3373,
+      "step": 824
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995097847102624,
+      "loss": 1.3958,
+      "step": 825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019995048952159859,
+      "loss": 1.5267,
+      "step": 826
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999499981464259,
+      "loss": 1.366,
+      "step": 827
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994950434552015,
+      "loss": 1.5209,
+      "step": 828
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994900811889333,
+      "loss": 3.0723,
+      "step": 829
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994850946655746,
+      "loss": 1.4015,
+      "step": 830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994800838852466,
+      "loss": 1.3834,
+      "step": 831
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994750488480705,
+      "loss": 1.3389,
+      "step": 832
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999469989554169,
+      "loss": 1.5292,
+      "step": 833
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994649060036643,
+      "loss": 1.4444,
+      "step": 834
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994597981966807,
+      "loss": 1.3785,
+      "step": 835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994546661333414,
+      "loss": 1.1985,
+      "step": 836
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999449509813771,
+      "loss": 1.3976,
+      "step": 837
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999444329238095,
+      "loss": 1.5219,
+      "step": 838
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994391244064388,
+      "loss": 1.4634,
+      "step": 839
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999433895318929,
+      "loss": 1.5004,
+      "step": 840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994286419756924,
+      "loss": 1.4417,
+      "step": 841
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994233643768562,
+      "loss": 1.4594,
+      "step": 842
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999418062522549,
+      "loss": 1.6108,
+      "step": 843
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994127364128991,
+      "loss": 3.1621,
+      "step": 844
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019994073860480358,
+      "loss": 1.4096,
+      "step": 845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999402011428089,
+      "loss": 1.4713,
+      "step": 846
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993966125531895,
+      "loss": 1.4561,
+      "step": 847
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993911894234677,
+      "loss": 1.2775,
+      "step": 848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993857420390557,
+      "loss": 1.3473,
+      "step": 849
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993802704000853,
+      "loss": 1.4385,
+      "step": 850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993747745066895,
+      "loss": 1.3546,
+      "step": 851
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993692543590023,
+      "loss": 1.4238,
+      "step": 852
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993637099571563,
+      "loss": 1.35,
+      "step": 853
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993581413012873,
+      "loss": 1.7217,
+      "step": 854
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993525483915297,
+      "loss": 1.4962,
+      "step": 855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993469312280194,
+      "loss": 1.3221,
+      "step": 856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999341289810893,
+      "loss": 1.4601,
+      "step": 857
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993356241402874,
+      "loss": 1.317,
+      "step": 858
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993299342163399,
+      "loss": 1.2933,
+      "step": 859
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993242200391884,
+      "loss": 1.454,
+      "step": 860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999318481608972,
+      "loss": 1.33,
+      "step": 861
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993127189258297,
+      "loss": 1.4575,
+      "step": 862
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019993069319899017,
+      "loss": 1.493,
+      "step": 863
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999301120801328,
+      "loss": 1.2653,
+      "step": 864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992952853602498,
+      "loss": 1.4161,
+      "step": 865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992894256668086,
+      "loss": 1.5934,
+      "step": 866
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992835417211472,
+      "loss": 1.5547,
+      "step": 867
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992776335234078,
+      "loss": 1.517,
+      "step": 868
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999271701073734,
+      "loss": 3.2307,
+      "step": 869
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992657443722696,
+      "loss": 1.442,
+      "step": 870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999259763419159,
+      "loss": 1.4335,
+      "step": 871
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992537582145482,
+      "loss": 1.5765,
+      "step": 872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999247728758582,
+      "loss": 1.5186,
+      "step": 873
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992416750514073,
+      "loss": 1.3657,
+      "step": 874
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992355970931708,
+      "loss": 1.4119,
+      "step": 875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992294948840202,
+      "loss": 1.603,
+      "step": 876
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992233684241033,
+      "loss": 1.5272,
+      "step": 877
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999217217713569,
+      "loss": 1.3218,
+      "step": 878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992110427525665,
+      "loss": 1.3726,
+      "step": 879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019992048435412457,
+      "loss": 1.5192,
+      "step": 880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001999198620079757,
+      "loss": 3.0979,
+      "step": 881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019991923723682514,
+      "loss": 1.506,
+      "step": 882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019991861004068807,
+      "loss": 1.4982,
+      "step": 883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999179804195797,
+      "loss": 1.412,
+      "step": 884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999173483735153,
+      "loss": 1.4269,
+      "step": 885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991671390251023,
+      "loss": 1.4437,
+      "step": 886
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999160770065799,
+      "loss": 1.253,
+      "step": 887
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991543768573973,
+      "loss": 1.5706,
+      "step": 888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999147959400052,
+      "loss": 1.3418,
+      "step": 889
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991415176939201,
+      "loss": 1.4053,
+      "step": 890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999135051739157,
+      "loss": 1.4786,
+      "step": 891
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991285615359198,
+      "loss": 1.4184,
+      "step": 892
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991220470843662,
+      "loss": 1.5692,
+      "step": 893
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991155083846538,
+      "loss": 1.2952,
+      "step": 894
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999108945436942,
+      "loss": 1.4401,
+      "step": 895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019991023582413897,
+      "loss": 1.4539,
+      "step": 896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990957467981566,
+      "loss": 1.5515,
+      "step": 897
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990891111074035,
+      "loss": 1.379,
+      "step": 898
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990824511692908,
+      "loss": 1.3561,
+      "step": 899
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990757669839814,
+      "loss": 1.4042,
+      "step": 900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990690585516363,
+      "loss": 2.7703,
+      "step": 901
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990623258724189,
+      "loss": 1.3943,
+      "step": 902
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990555689464924,
+      "loss": 1.3747,
+      "step": 903
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999048787774021,
+      "loss": 1.3889,
+      "step": 904
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990419823551686,
+      "loss": 1.6121,
+      "step": 905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990351526901013,
+      "loss": 1.5873,
+      "step": 906
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990282987789843,
+      "loss": 1.6811,
+      "step": 907
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990214206219844,
+      "loss": 1.469,
+      "step": 908
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001999014518219268,
+      "loss": 1.3315,
+      "step": 909
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990075915710028,
+      "loss": 1.4561,
+      "step": 910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019990006406773566,
+      "loss": 1.424,
+      "step": 911
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989936655384987,
+      "loss": 1.3544,
+      "step": 912
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989866661545983,
+      "loss": 1.7487,
+      "step": 913
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998979642525825,
+      "loss": 1.4219,
+      "step": 914
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998972594652349,
+      "loss": 2.7748,
+      "step": 915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998965522534342,
+      "loss": 1.3303,
+      "step": 916
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989584261719752,
+      "loss": 1.6017,
+      "step": 917
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998951305565421,
+      "loss": 1.3536,
+      "step": 918
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998944160714852,
+      "loss": 2.4822,
+      "step": 919
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989369916204418,
+      "loss": 1.3567,
+      "step": 920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989297982823647,
+      "loss": 1.5486,
+      "step": 921
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989225807007946,
+      "loss": 1.424,
+      "step": 922
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998915338875907,
+      "loss": 1.5638,
+      "step": 923
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989080728078778,
+      "loss": 1.4938,
+      "step": 924
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019989007824968832,
+      "loss": 1.5725,
+      "step": 925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988934679431,
+      "loss": 1.4449,
+      "step": 926
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998886129146706,
+      "loss": 1.4055,
+      "step": 927
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988787661078788,
+      "loss": 1.4605,
+      "step": 928
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998871378826798,
+      "loss": 1.1881,
+      "step": 929
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988639673036417,
+      "loss": 1.3839,
+      "step": 930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998856531538591,
+      "loss": 1.2905,
+      "step": 931
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988490715318254,
+      "loss": 1.4794,
+      "step": 932
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988415872835265,
+      "loss": 2.646,
+      "step": 933
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988340787938756,
+      "loss": 1.4117,
+      "step": 934
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988265460630554,
+      "loss": 1.3015,
+      "step": 935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988189890912482,
+      "loss": 1.3241,
+      "step": 936
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988114078786374,
+      "loss": 1.4426,
+      "step": 937
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019988038024254073,
+      "loss": 1.3596,
+      "step": 938
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987961727317428,
+      "loss": 1.4706,
+      "step": 939
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987885187978282,
+      "loss": 1.3509,
+      "step": 940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987808406238497,
+      "loss": 1.5063,
+      "step": 941
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998773138209994,
+      "loss": 1.5446,
+      "step": 942
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987654115564475,
+      "loss": 1.3951,
+      "step": 943
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987576606633978,
+      "loss": 2.2068,
+      "step": 944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998749885531033,
+      "loss": 1.2994,
+      "step": 945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987420861595424,
+      "loss": 1.4902,
+      "step": 946
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987342625491143,
+      "loss": 1.476,
+      "step": 947
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987264146999393,
+      "loss": 1.4453,
+      "step": 948
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998718542612208,
+      "loss": 2.4902,
+      "step": 949
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987106462861104,
+      "loss": 1.3621,
+      "step": 950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019987027257218395,
+      "loss": 1.5012,
+      "step": 951
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986947809195863,
+      "loss": 1.4974,
+      "step": 952
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986868118795442,
+      "loss": 1.5447,
+      "step": 953
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998678818601907,
+      "loss": 1.4855,
+      "step": 954
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998670801086868,
+      "loss": 1.3825,
+      "step": 955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998662759334622,
+      "loss": 1.1675,
+      "step": 956
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986546933453646,
+      "loss": 1.5409,
+      "step": 957
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998646603119291,
+      "loss": 1.3831,
+      "step": 958
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986384886565972,
+      "loss": 1.4044,
+      "step": 959
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986303499574813,
+      "loss": 1.4462,
+      "step": 960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986221870221398,
+      "loss": 1.4345,
+      "step": 961
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986139998507714,
+      "loss": 1.5774,
+      "step": 962
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019986057884435746,
+      "loss": 3.1431,
+      "step": 963
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985975528007485,
+      "loss": 1.694,
+      "step": 964
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985892929224935,
+      "loss": 1.42,
+      "step": 965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985810088090093,
+      "loss": 1.486,
+      "step": 966
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985727004604974,
+      "loss": 1.4766,
+      "step": 967
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985643678771596,
+      "loss": 1.3188,
+      "step": 968
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998556011059198,
+      "loss": 1.4187,
+      "step": 969
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998547630006815,
+      "loss": 1.4343,
+      "step": 970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985392247202146,
+      "loss": 1.5824,
+      "step": 971
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985307951996006,
+      "loss": 1.3843,
+      "step": 972
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985223414451775,
+      "loss": 1.5154,
+      "step": 973
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985138634571502,
+      "loss": 1.2908,
+      "step": 974
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019985053612357252,
+      "loss": 1.3906,
+      "step": 975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019984968347811082,
+      "loss": 1.363,
+      "step": 976
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998488284093506,
+      "loss": 1.4627,
+      "step": 977
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019984797091731265,
+      "loss": 1.3708,
+      "step": 978
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019984711100201784,
+      "loss": 1.3052,
+      "step": 979
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998462486634869,
+      "loss": 1.4784,
+      "step": 980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998453839017409,
+      "loss": 1.3481,
+      "step": 981
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998445167168007,
+      "loss": 1.3386,
+      "step": 982
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019984364710868744,
+      "loss": 1.4259,
+      "step": 983
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019984277507742218,
+      "loss": 1.4078,
+      "step": 984
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998419006230261,
+      "loss": 1.2881,
+      "step": 985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019984102374552042,
+      "loss": 1.5122,
+      "step": 986
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998401444449264,
+      "loss": 1.4207,
+      "step": 987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983926272126542,
+      "loss": 1.463,
+      "step": 988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983837857455886,
+      "loss": 2.4246,
+      "step": 989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983749200482815,
+      "loss": 1.4871,
+      "step": 990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983660301209485,
+      "loss": 1.4069,
+      "step": 991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983571159638052,
+      "loss": 1.4352,
+      "step": 992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983481775770682,
+      "loss": 1.5542,
+      "step": 993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983392149609537,
+      "loss": 2.4883,
+      "step": 994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199833022811568,
+      "loss": 1.4938,
+      "step": 995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983212170414646,
+      "loss": 1.4174,
+      "step": 996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019983121817385263,
+      "loss": 1.4463,
+      "step": 997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998303122207085,
+      "loss": 1.4537,
+      "step": 998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199829403844736,
+      "loss": 1.3315,
+      "step": 999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982849304595718,
+      "loss": 1.6179,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982757982439417,
+      "loss": 1.4041,
+      "step": 1001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998266641800691,
+      "loss": 1.3577,
+      "step": 1002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998257461130042,
+      "loss": 1.2948,
+      "step": 1003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982482562322178,
+      "loss": 1.3276,
+      "step": 1004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982390271074417,
+      "loss": 1.5827,
+      "step": 1005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998229773755937,
+      "loss": 1.4539,
+      "step": 1006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982204961779295,
+      "loss": 1.3677,
+      "step": 1007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982111943736433,
+      "loss": 1.4044,
+      "step": 1008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019982018683433047,
+      "loss": 2.4802,
+      "step": 1009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199819251808714,
+      "loss": 1.3583,
+      "step": 1010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019981831436053756,
+      "loss": 1.4619,
+      "step": 1011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199817374489824,
+      "loss": 1.2853,
+      "step": 1012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019981643219659604,
+      "loss": 1.5219,
+      "step": 1013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998154874808766,
+      "loss": 1.3192,
+      "step": 1014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019981454034268857,
+      "loss": 1.2955,
+      "step": 1015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199813590782055,
+      "loss": 1.5754,
+      "step": 1016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019981263879899884,
+      "loss": 1.5167,
+      "step": 1017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019981168439354326,
+      "loss": 1.3116,
+      "step": 1018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019981072756571143,
+      "loss": 1.4313,
+      "step": 1019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980976831552653,
+      "loss": 1.3685,
+      "step": 1020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998088066430119,
+      "loss": 1.5477,
+      "step": 1021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998078425481908,
+      "loss": 2.3882,
+      "step": 1022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980687603108667,
+      "loss": 1.441,
+      "step": 1023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199805907091723,
+      "loss": 1.4786,
+      "step": 1024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980493573012326,
+      "loss": 1.3886,
+      "step": 1025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980396194631102,
+      "loss": 1.3796,
+      "step": 1026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980298574030994,
+      "loss": 1.239,
+      "step": 1027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980200711214367,
+      "loss": 1.3887,
+      "step": 1028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019980102606183603,
+      "loss": 1.5867,
+      "step": 1029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001998000425894108,
+      "loss": 1.5561,
+      "step": 1030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997990566948918,
+      "loss": 1.3492,
+      "step": 1031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019979806837830305,
+      "loss": 1.5181,
+      "step": 1032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019979707763966844,
+      "loss": 1.2076,
+      "step": 1033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019979608447901205,
+      "loss": 2.7576,
+      "step": 1034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000199795088896358,
+      "loss": 1.5134,
+      "step": 1035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997940908917305,
+      "loss": 1.3705,
+      "step": 1036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019979309046515365,
+      "loss": 1.2934,
+      "step": 1037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997920876166518,
+      "loss": 1.5539,
+      "step": 1038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997910823462493,
+      "loss": 1.2078,
+      "step": 1039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019979007465397052,
+      "loss": 1.3914,
+      "step": 1040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978906453983994,
+      "loss": 1.4635,
+      "step": 1041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978805200388205,
+      "loss": 1.4818,
+      "step": 1042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978703704612145,
+      "loss": 1.4357,
+      "step": 1043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978601966658274,
+      "loss": 1.5032,
+      "step": 1044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978499986529063,
+      "loss": 1.339,
+      "step": 1045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978397764226988,
+      "loss": 1.2891,
+      "step": 1046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997829529975453,
+      "loss": 1.4408,
+      "step": 1047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978192593114171,
+      "loss": 1.4813,
+      "step": 1048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019978089644308409,
+      "loss": 1.4106,
+      "step": 1049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997798645333974,
+      "loss": 1.1915,
+      "step": 1050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997788302021067,
+      "loss": 2.8347,
+      "step": 1051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019977779344923708,
+      "loss": 1.4545,
+      "step": 1052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997767542748137,
+      "loss": 1.1927,
+      "step": 1053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997757126788618,
+      "loss": 1.4093,
+      "step": 1054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019977466866140663,
+      "loss": 1.4605,
+      "step": 1055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019977362222247354,
+      "loss": 1.4183,
+      "step": 1056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019977257336208795,
+      "loss": 1.3164,
+      "step": 1057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019977152208027526,
+      "loss": 1.1531,
+      "step": 1058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019977046837706103,
+      "loss": 1.364,
+      "step": 1059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019976941225247083,
+      "loss": 1.5323,
+      "step": 1060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019976835370653028,
+      "loss": 2.6775,
+      "step": 1061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997672927392651,
+      "loss": 1.2211,
+      "step": 1062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019976622935070096,
+      "loss": 1.4146,
+      "step": 1063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997651635408637,
+      "loss": 1.3404,
+      "step": 1064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019976409530977928,
+      "loss": 1.3492,
+      "step": 1065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997630246574735,
+      "loss": 1.4123,
+      "step": 1066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019976195158397241,
+      "loss": 1.2955,
+      "step": 1067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019976087608930207,
+      "loss": 1.5095,
+      "step": 1068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975979817348852,
+      "loss": 1.541,
+      "step": 1069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975871783655796,
+      "loss": 1.4614,
+      "step": 1070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975763507853659,
+      "loss": 1.4205,
+      "step": 1071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997565498994507,
+      "loss": 1.4288,
+      "step": 1072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997554622993266,
+      "loss": 1.4366,
+      "step": 1073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975437227819074,
+      "loss": 1.4062,
+      "step": 1074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975327983606954,
+      "loss": 1.5307,
+      "step": 1075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975218497298952,
+      "loss": 1.5583,
+      "step": 1076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975108768897725,
+      "loss": 1.2849,
+      "step": 1077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974998798405933,
+      "loss": 1.4448,
+      "step": 1078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997488858582625,
+      "loss": 2.8604,
+      "step": 1079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974778131161348,
+      "loss": 2.7998,
+      "step": 1080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997466743441391,
+      "loss": 1.3814,
+      "step": 1081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997455649558662,
+      "loss": 1.2513,
+      "step": 1082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997444531468217,
+      "loss": 1.3699,
+      "step": 1083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974333891703262,
+      "loss": 1.6053,
+      "step": 1084
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974222226652597,
+      "loss": 1.4005,
+      "step": 1085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974110319532885,
+      "loss": 1.375,
+      "step": 1086
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973998170346843,
+      "loss": 1.6002,
+      "step": 1087
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997388577909719,
+      "loss": 1.483,
+      "step": 1088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997377314578666,
+      "loss": 1.3686,
+      "step": 1089
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973660270417983,
+      "loss": 1.3499,
+      "step": 1090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973547152993897,
+      "loss": 1.4411,
+      "step": 1091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997343379351715,
+      "loss": 1.3098,
+      "step": 1092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997332019199049,
+      "loss": 2.6201,
+      "step": 1093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019973206348416675,
+      "loss": 1.5662,
+      "step": 1094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997309226279847,
+      "loss": 2.5325,
+      "step": 1095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997297793513864,
+      "loss": 1.485,
+      "step": 1096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019972863365439966,
+      "loss": 1.3949,
+      "step": 1097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019972748553705226,
+      "loss": 1.2967,
+      "step": 1098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019972633499937202,
+      "loss": 1.3758,
+      "step": 1099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997251820413869,
+      "loss": 1.4957,
+      "step": 1100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019972402666312483,
+      "loss": 1.4612,
+      "step": 1101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019972286886461396,
+      "loss": 1.3873,
+      "step": 1102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997217086458823,
+      "loss": 1.2997,
+      "step": 1103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000199720546006958,
+      "loss": 1.3331,
+      "step": 1104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971938094786936,
+      "loss": 1.4343,
+      "step": 1105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971821346864458,
+      "loss": 1.4286,
+      "step": 1106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971704356931206,
+      "loss": 1.1808,
+      "step": 1107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997158712499001,
+      "loss": 1.3896,
+      "step": 1108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971469651043722,
+      "loss": 1.4058,
+      "step": 1109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971351935095194,
+      "loss": 1.4969,
+      "step": 1110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971233977147276,
+      "loss": 1.3506,
+      "step": 1111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019971115777202836,
+      "loss": 1.3152,
+      "step": 1112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019970997335264743,
+      "loss": 1.3448,
+      "step": 1113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997087865133587,
+      "loss": 1.427,
+      "step": 1114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019970759725419096,
+      "loss": 1.4181,
+      "step": 1115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019970640557517312,
+      "loss": 1.5149,
+      "step": 1116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019970521147633406,
+      "loss": 1.4673,
+      "step": 1117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019970401495770276,
+      "loss": 1.2904,
+      "step": 1118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019970281601930826,
+      "loss": 1.3669,
+      "step": 1119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997016146611797,
+      "loss": 1.5199,
+      "step": 1120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001997004108833462,
+      "loss": 2.5884,
+      "step": 1121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969920468583695,
+      "loss": 1.0798,
+      "step": 1122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969799606868124,
+      "loss": 1.4226,
+      "step": 1123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969678503190845,
+      "loss": 1.349,
+      "step": 1124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969557157554794,
+      "loss": 1.296,
+      "step": 1125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969435569962914,
+      "loss": 1.4727,
+      "step": 1126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969313740418157,
+      "loss": 1.2501,
+      "step": 1127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996919166892348,
+      "loss": 1.2891,
+      "step": 1128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969069355481846,
+      "loss": 1.486,
+      "step": 1129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968946800096225,
+      "loss": 2.4839,
+      "step": 1130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968824002769586,
+      "loss": 1.4199,
+      "step": 1131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968700963504915,
+      "loss": 1.3383,
+      "step": 1132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968577682305197,
+      "loss": 1.2454,
+      "step": 1133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968454159173424,
+      "loss": 1.4378,
+      "step": 1134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968330394112588,
+      "loss": 2.3938,
+      "step": 1135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968206387125701,
+      "loss": 1.4212,
+      "step": 1136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996808213821577,
+      "loss": 1.3906,
+      "step": 1137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967957647385807,
+      "loss": 1.5772,
+      "step": 1138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967832914638835,
+      "loss": 1.4005,
+      "step": 1139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967707939977887,
+      "loss": 1.2999,
+      "step": 1140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967582723405986,
+      "loss": 1.5354,
+      "step": 1141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996745726492618,
+      "loss": 1.3397,
+      "step": 1142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967331564541504,
+      "loss": 1.4835,
+      "step": 1143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996720562225502,
+      "loss": 1.3047,
+      "step": 1144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967079438069777,
+      "loss": 1.3495,
+      "step": 1145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996695301198884,
+      "loss": 1.3755,
+      "step": 1146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966826344015276,
+      "loss": 1.2992,
+      "step": 1147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966699434152162,
+      "loss": 1.3494,
+      "step": 1148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966572282402572,
+      "loss": 1.4196,
+      "step": 1149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000199664448887696,
+      "loss": 1.4446,
+      "step": 1150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996631725325633,
+      "loss": 1.4056,
+      "step": 1151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966189375865867,
+      "loss": 2.5405,
+      "step": 1152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996606125660131,
+      "loss": 1.4131,
+      "step": 1153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965932895465768,
+      "loss": 1.3335,
+      "step": 1154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996580429246236,
+      "loss": 1.3549,
+      "step": 1155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965675447594202,
+      "loss": 1.2302,
+      "step": 1156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965546360864423,
+      "loss": 1.3395,
+      "step": 1157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965417032276157,
+      "loss": 1.6714,
+      "step": 1158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965287461832541,
+      "loss": 1.3227,
+      "step": 1159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965157649536725,
+      "loss": 2.7903,
+      "step": 1160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965027595391852,
+      "loss": 1.3638,
+      "step": 1161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964897299401085,
+      "loss": 1.4622,
+      "step": 1162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964766761567578,
+      "loss": 1.2408,
+      "step": 1163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964635981894506,
+      "loss": 1.4168,
+      "step": 1164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996450496038504,
+      "loss": 2.2061,
+      "step": 1165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964373697042365,
+      "loss": 1.1731,
+      "step": 1166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964242191869658,
+      "loss": 1.4346,
+      "step": 1167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964110444870117,
+      "loss": 1.468,
+      "step": 1168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963978456046934,
+      "loss": 1.4205,
+      "step": 1169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996384622540332,
+      "loss": 1.3127,
+      "step": 1170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963713752942476,
+      "loss": 1.3829,
+      "step": 1171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963581038667626,
+      "loss": 1.4243,
+      "step": 1172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963448082581983,
+      "loss": 1.2914,
+      "step": 1173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963314884688775,
+      "loss": 2.2295,
+      "step": 1174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963181444991238,
+      "loss": 1.2989,
+      "step": 1175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996304776349261,
+      "loss": 1.2226,
+      "step": 1176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996291384019613,
+      "loss": 1.366,
+      "step": 1177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019962779675105054,
+      "loss": 1.2612,
+      "step": 1178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019962645268222637,
+      "loss": 1.4337,
+      "step": 1179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996251061955214,
+      "loss": 1.2694,
+      "step": 1180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996237572909683,
+      "loss": 1.4537,
+      "step": 1181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019962240596859988,
+      "loss": 1.5082,
+      "step": 1182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996210522284488,
+      "loss": 1.3876,
+      "step": 1183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019961969607054804,
+      "loss": 1.3921,
+      "step": 1184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019961833749493045,
+      "loss": 1.2957,
+      "step": 1185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019961697650162902,
+      "loss": 1.3102,
+      "step": 1186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019961561309067677,
+      "loss": 1.316,
+      "step": 1187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996142472621068,
+      "loss": 1.4897,
+      "step": 1188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019961287901595226,
+      "loss": 1.5396,
+      "step": 1189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019961150835224632,
+      "loss": 2.2117,
+      "step": 1190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996101352710223,
+      "loss": 1.3918,
+      "step": 1191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019960875977231347,
+      "loss": 1.3723,
+      "step": 1192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019960738185615328,
+      "loss": 2.479,
+      "step": 1193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019960600152257513,
+      "loss": 1.4702,
+      "step": 1194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019960461877161252,
+      "loss": 1.377,
+      "step": 1195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000199603233603299,
+      "loss": 1.4587,
+      "step": 1196
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001996018460176682,
+      "loss": 1.259,
+      "step": 1197
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001996004560147538,
+      "loss": 1.3672,
+      "step": 1198
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995990635945895,
+      "loss": 1.1883,
+      "step": 1199
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959766875720915,
+      "loss": 1.2522,
+      "step": 1200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959627150264658,
+      "loss": 1.2027,
+      "step": 1201
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995948718309357,
+      "loss": 1.4946,
+      "step": 1202
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959346974211047,
+      "loss": 1.5393,
+      "step": 1203
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959206523620493,
+      "loss": 1.3676,
+      "step": 1204
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959065831325314,
+      "loss": 1.3526,
+      "step": 1205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958924897328927,
+      "loss": 2.269,
+      "step": 1206
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995878372163475,
+      "loss": 1.4207,
+      "step": 1207
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958642304246214,
+      "loss": 1.3535,
+      "step": 1208
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958500645166747,
+      "loss": 1.3733,
+      "step": 1209
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958358744399788,
+      "loss": 1.3479,
+      "step": 1210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958216601948783,
+      "loss": 1.1961,
+      "step": 1211
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958074217817177,
+      "loss": 1.3563,
+      "step": 1212
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957931592008433,
+      "loss": 1.4818,
+      "step": 1213
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957788724526,
+      "loss": 1.3216,
+      "step": 1214
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995764561537336,
+      "loss": 1.3558,
+      "step": 1215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957502264553978,
+      "loss": 1.3586,
+      "step": 1216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995735867207133,
+      "loss": 1.3787,
+      "step": 1217
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957214837928913,
+      "loss": 1.4003,
+      "step": 1218
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957070762130204,
+      "loss": 1.4015,
+      "step": 1219
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956926444678709,
+      "loss": 1.2514,
+      "step": 1220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956781885577924,
+      "loss": 1.269,
+      "step": 1221
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956637084831363,
+      "loss": 1.4005,
+      "step": 1222
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956492042442537,
+      "loss": 1.1825,
+      "step": 1223
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995634675841497,
+      "loss": 2.5085,
+      "step": 1224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995620123275218,
+      "loss": 1.504,
+      "step": 1225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956055465457706,
+      "loss": 1.3427,
+      "step": 1226
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019955909456535083,
+      "loss": 1.1729,
+      "step": 1227
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019955763205987856,
+      "loss": 1.5201,
+      "step": 1228
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995561671381957,
+      "loss": 1.2286,
+      "step": 1229
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019955469980033788,
+      "loss": 2.429,
+      "step": 1230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019955323004634067,
+      "loss": 1.4011,
+      "step": 1231
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019955175787623973,
+      "loss": 1.2764,
+      "step": 1232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019955028329007076,
+      "loss": 1.2153,
+      "step": 1233
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995488062878696,
+      "loss": 2.2827,
+      "step": 1234
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995473268696721,
+      "loss": 1.2842,
+      "step": 1235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019954584503551414,
+      "loss": 2.2311,
+      "step": 1236
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995443607854317,
+      "loss": 1.3298,
+      "step": 1237
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019954287411946077,
+      "loss": 1.545,
+      "step": 1238
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019954138503763748,
+      "loss": 1.2855,
+      "step": 1239
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019953989353999792,
+      "loss": 1.5215,
+      "step": 1240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995383996265783,
+      "loss": 1.3301,
+      "step": 1241
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995369032974149,
+      "loss": 1.3337,
+      "step": 1242
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000199535404552544,
+      "loss": 1.5585,
+      "step": 1243
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019953390339200202,
+      "loss": 1.2982,
+      "step": 1244
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019953239981582534,
+      "loss": 1.1445,
+      "step": 1245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995308938240505,
+      "loss": 1.5102,
+      "step": 1246
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952938541671403,
+      "loss": 1.2635,
+      "step": 1247
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952787459385252,
+      "loss": 1.4388,
+      "step": 1248
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952636135550265,
+      "loss": 2.6399,
+      "step": 1249
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952484570170116,
+      "loss": 1.2636,
+      "step": 1250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952332763248482,
+      "loss": 1.206,
+      "step": 1251
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952180714789047,
+      "loss": 1.1779,
+      "step": 1252
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019952028424795502,
+      "loss": 1.3255,
+      "step": 1253
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995187589327154,
+      "loss": 1.4745,
+      "step": 1254
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019951723120220866,
+      "loss": 1.5084,
+      "step": 1255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995157010564719,
+      "loss": 1.3644,
+      "step": 1256
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995141684955422,
+      "loss": 1.4667,
+      "step": 1257
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995126335194568,
+      "loss": 1.3004,
+      "step": 1258
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019951109612825294,
+      "loss": 1.3111,
+      "step": 1259
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995095563219679,
+      "loss": 1.2372,
+      "step": 1260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019950801410063907,
+      "loss": 1.5061,
+      "step": 1261
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019950646946430392,
+      "loss": 1.4622,
+      "step": 1262
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019950492241299988,
+      "loss": 1.358,
+      "step": 1263
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019950337294676455,
+      "loss": 1.3166,
+      "step": 1264
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995018210656355,
+      "loss": 1.4039,
+      "step": 1265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019950026676965037,
+      "loss": 1.2487,
+      "step": 1266
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019949871005884692,
+      "loss": 1.4191,
+      "step": 1267
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994971509332629,
+      "loss": 1.4567,
+      "step": 1268
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994955893929362,
+      "loss": 1.2487,
+      "step": 1269
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019949402543790467,
+      "loss": 1.374,
+      "step": 1270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994924590682063,
+      "loss": 1.2972,
+      "step": 1271
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019949089028387906,
+      "loss": 1.4188,
+      "step": 1272
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019948931908496108,
+      "loss": 1.3813,
+      "step": 1273
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019948774547149043,
+      "loss": 1.3561,
+      "step": 1274
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994861694435054,
+      "loss": 1.2957,
+      "step": 1275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994845910010441,
+      "loss": 1.335,
+      "step": 1276
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019948301014414494,
+      "loss": 1.3713,
+      "step": 1277
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019948142687284626,
+      "loss": 2.3376,
+      "step": 1278
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019947984118718647,
+      "loss": 1.2426,
+      "step": 1279
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019947825308720408,
+      "loss": 1.4272,
+      "step": 1280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019947666257293762,
+      "loss": 1.2981,
+      "step": 1281
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019947506964442568,
+      "loss": 1.2781,
+      "step": 1282
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019947347430170694,
+      "loss": 1.5093,
+      "step": 1283
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994718765448201,
+      "loss": 1.4892,
+      "step": 1284
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019947027637380393,
+      "loss": 1.3362,
+      "step": 1285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019946867378869732,
+      "loss": 1.2612,
+      "step": 1286
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994670687895391,
+      "loss": 1.2914,
+      "step": 1287
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019946546137636822,
+      "loss": 1.289,
+      "step": 1288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019946385154922378,
+      "loss": 1.2849,
+      "step": 1289
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019946223930814475,
+      "loss": 1.3795,
+      "step": 1290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001994606246531703,
+      "loss": 1.155,
+      "step": 1291
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019945900758433963,
+      "loss": 1.2755,
+      "step": 1292
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019945738810169192,
+      "loss": 1.3625,
+      "step": 1293
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019945576620526657,
+      "loss": 1.6033,
+      "step": 1294
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019945414189510285,
+      "loss": 1.3835,
+      "step": 1295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019945251517124025,
+      "loss": 1.3647,
+      "step": 1296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019945088603371823,
+      "loss": 2.5881,
+      "step": 1297
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019944925448257632,
+      "loss": 1.4953,
+      "step": 1298
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019944762051785412,
+      "loss": 1.3403,
+      "step": 1299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019944598413959127,
+      "loss": 1.2826,
+      "step": 1300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019944434534782752,
+      "loss": 1.3683,
+      "step": 1301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019944270414260263,
+      "loss": 1.4086,
+      "step": 1302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994410605239564,
+      "loss": 1.4802,
+      "step": 1303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019943941449192875,
+      "loss": 1.3358,
+      "step": 1304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994377660465596,
+      "loss": 1.1549,
+      "step": 1305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000199436115187889,
+      "loss": 1.4201,
+      "step": 1306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000199434461915957,
+      "loss": 1.244,
+      "step": 1307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019943280623080373,
+      "loss": 1.3093,
+      "step": 1308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994311481324693,
+      "loss": 1.4625,
+      "step": 1309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019942948762099408,
+      "loss": 2.4441,
+      "step": 1310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019942782469641827,
+      "loss": 1.2189,
+      "step": 1311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019942615935878227,
+      "loss": 1.2536,
+      "step": 1312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994244916081265,
+      "loss": 1.3077,
+      "step": 1313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019942282144449138,
+      "loss": 1.3793,
+      "step": 1314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994211488679175,
+      "loss": 1.4312,
+      "step": 1315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019941947387844543,
+      "loss": 1.4084,
+      "step": 1316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019941779647611587,
+      "loss": 1.5374,
+      "step": 1317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019941611666096946,
+      "loss": 1.3583,
+      "step": 1318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000199414434433047,
+      "loss": 1.3862,
+      "step": 1319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994127497923893,
+      "loss": 1.4857,
+      "step": 1320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019941106273903728,
+      "loss": 1.2931,
+      "step": 1321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019940937327303187,
+      "loss": 1.38,
+      "step": 1322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019940768139441403,
+      "loss": 1.3583,
+      "step": 1323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019940598710322492,
+      "loss": 2.6174,
+      "step": 1324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019940429039950553,
+      "loss": 1.4135,
+      "step": 1325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019940259128329712,
+      "loss": 1.6381,
+      "step": 1326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019940088975464092,
+      "loss": 1.3175,
+      "step": 1327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019939918581357823,
+      "loss": 1.2014,
+      "step": 1328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993974794601504,
+      "loss": 1.4037,
+      "step": 1329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993957706943988,
+      "loss": 1.3086,
+      "step": 1330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019939405951636494,
+      "loss": 1.3069,
+      "step": 1331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019939234592609032,
+      "loss": 1.4897,
+      "step": 1332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019939062992361658,
+      "loss": 1.5116,
+      "step": 1333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019938891150898537,
+      "loss": 1.3778,
+      "step": 1334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993871906822383,
+      "loss": 1.4597,
+      "step": 1335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019938546744341723,
+      "loss": 1.4529,
+      "step": 1336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019938374179256392,
+      "loss": 1.2809,
+      "step": 1337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993820137297203,
+      "loss": 1.3126,
+      "step": 1338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019938028325492828,
+      "loss": 1.6152,
+      "step": 1339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019937855036822986,
+      "loss": 1.4332,
+      "step": 1340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993768150696671,
+      "loss": 2.2944,
+      "step": 1341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019937507735928213,
+      "loss": 1.4097,
+      "step": 1342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019937333723711711,
+      "loss": 1.4976,
+      "step": 1343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019937159470321424,
+      "loss": 1.2839,
+      "step": 1344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019936984975761588,
+      "loss": 1.3192,
+      "step": 1345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019936810240036432,
+      "loss": 1.3065,
+      "step": 1346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000199366352631502,
+      "loss": 1.2587,
+      "step": 1347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019936460045107138,
+      "loss": 1.3171,
+      "step": 1348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019936284585911495,
+      "loss": 1.3168,
+      "step": 1349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019936108885567534,
+      "loss": 1.4879,
+      "step": 1350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019935932944079517,
+      "loss": 1.3398,
+      "step": 1351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019935756761451712,
+      "loss": 1.2465,
+      "step": 1352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000199355803376884,
+      "loss": 1.5333,
+      "step": 1353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993540367279386,
+      "loss": 1.3116,
+      "step": 1354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019935226766772379,
+      "loss": 1.2364,
+      "step": 1355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993504961962825,
+      "loss": 1.4521,
+      "step": 1356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019934872231365774,
+      "loss": 2.571,
+      "step": 1357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019934694601989253,
+      "loss": 1.4968,
+      "step": 1358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019934516731503005,
+      "loss": 1.4961,
+      "step": 1359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019934338619911337,
+      "loss": 1.3668,
+      "step": 1360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019934160267218577,
+      "loss": 1.3557,
+      "step": 1361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019933981673429053,
+      "loss": 1.4724,
+      "step": 1362
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019933802838547102,
+      "loss": 1.3223,
+      "step": 1363
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019933623762577064,
+      "loss": 1.1697,
+      "step": 1364
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001993344444552328,
+      "loss": 1.3833,
+      "step": 1365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019933264887390104,
+      "loss": 1.17,
+      "step": 1366
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019933085088181892,
+      "loss": 1.242,
+      "step": 1367
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019932905047903018,
+      "loss": 1.3112,
+      "step": 1368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019932724766557835,
+      "loss": 1.4902,
+      "step": 1369
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019932544244150732,
+      "loss": 1.5319,
+      "step": 1370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019932363480686086,
+      "loss": 1.3426,
+      "step": 1371
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019932182476168283,
+      "loss": 1.4307,
+      "step": 1372
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019932001230601714,
+      "loss": 1.2283,
+      "step": 1373
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019931819743990777,
+      "loss": 1.312,
+      "step": 1374
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019931638016339886,
+      "loss": 1.1206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019931456047653443,
+      "loss": 1.532,
+      "step": 1376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019931273837935865,
+      "loss": 1.3641,
+      "step": 1377
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019931091387191574,
+      "loss": 1.5071,
+      "step": 1378
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019930908695425,
+      "loss": 1.261,
+      "step": 1379
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019930725762640577,
+      "loss": 1.3938,
+      "step": 1380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019930542588842743,
+      "loss": 1.2971,
+      "step": 1381
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019930359174035947,
+      "loss": 1.2678,
+      "step": 1382
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019930175518224634,
+      "loss": 1.3978,
+      "step": 1383
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001992999162141327,
+      "loss": 1.4238,
+      "step": 1384
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019929807483606306,
+      "loss": 1.3574,
+      "step": 1385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019929623104808223,
+      "loss": 1.5353,
+      "step": 1386
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019929438485023488,
+      "loss": 1.3238,
+      "step": 1387
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019929253624256584,
+      "loss": 1.442,
+      "step": 1388
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019929068522512,
+      "loss": 2.2673,
+      "step": 1389
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019928883179794226,
+      "loss": 1.4218,
+      "step": 1390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001992869759610776,
+      "loss": 1.3657,
+      "step": 1391
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019928511771457106,
+      "loss": 1.4358,
+      "step": 1392
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019928325705846774,
+      "loss": 1.235,
+      "step": 1393
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001992813939928128,
+      "loss": 1.3609,
+      "step": 1394
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019927952851765145,
+      "loss": 2.4048,
+      "step": 1395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019927766063302898,
+      "loss": 1.2972,
+      "step": 1396
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001992757903389907,
+      "loss": 1.3609,
+      "step": 1397
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019927391763558202,
+      "loss": 1.4427,
+      "step": 1398
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001992720425228484,
+      "loss": 1.3965,
+      "step": 1399
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001992701650008353,
+      "loss": 1.3317,
+      "step": 1400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019926828506958833,
+      "loss": 1.2821,
+      "step": 1401
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019926640272915309,
+      "loss": 1.3101,
+      "step": 1402
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019926451797957527,
+      "loss": 1.462,
+      "step": 1403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019926263082090063,
+      "loss": 2.4929,
+      "step": 1404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019926074125317496,
+      "loss": 1.3949,
+      "step": 1405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992588492764441,
+      "loss": 1.419,
+      "step": 1406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000199256954890754,
+      "loss": 1.4662,
+      "step": 1407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992550580961506,
+      "loss": 1.5015,
+      "step": 1408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019925315889267995,
+      "loss": 1.2159,
+      "step": 1409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019925125728038816,
+      "loss": 1.3923,
+      "step": 1410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019924935325932137,
+      "loss": 1.2089,
+      "step": 1411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992474468295258,
+      "loss": 1.4399,
+      "step": 1412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019924553799104765,
+      "loss": 1.2386,
+      "step": 1413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019924362674393335,
+      "loss": 2.4705,
+      "step": 1414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992417130882292,
+      "loss": 1.479,
+      "step": 1415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992397970239817,
+      "loss": 2.4287,
+      "step": 1416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019923787855123735,
+      "loss": 1.3961,
+      "step": 1417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019923595767004266,
+      "loss": 1.3965,
+      "step": 1418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019923403438044428,
+      "loss": 1.4807,
+      "step": 1419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992321086824889,
+      "loss": 1.459,
+      "step": 1420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019923018057622325,
+      "loss": 1.3331,
+      "step": 1421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019922825006169409,
+      "loss": 1.343,
+      "step": 1422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019922631713894831,
+      "loss": 1.3718,
+      "step": 1423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019922438180803282,
+      "loss": 1.3567,
+      "step": 1424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992224440689946,
+      "loss": 1.377,
+      "step": 1425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992205039218806,
+      "loss": 1.2949,
+      "step": 1426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000199218561366738,
+      "loss": 1.2052,
+      "step": 1427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992166164036139,
+      "loss": 1.3522,
+      "step": 1428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019921466903255553,
+      "loss": 1.2747,
+      "step": 1429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992127192536101,
+      "loss": 1.4711,
+      "step": 1430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000199210767066825,
+      "loss": 1.2849,
+      "step": 1431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019920881247224753,
+      "loss": 1.322,
+      "step": 1432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992068554699252,
+      "loss": 1.2972,
+      "step": 1433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992048960599055,
+      "loss": 1.5016,
+      "step": 1434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001992029342422359,
+      "loss": 1.3845,
+      "step": 1435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019920097001696412,
+      "loss": 1.6105,
+      "step": 1436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019919900338413778,
+      "loss": 2.4165,
+      "step": 1437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019919703434380457,
+      "loss": 1.3759,
+      "step": 1438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019919506289601234,
+      "loss": 1.5316,
+      "step": 1439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019919308904080892,
+      "loss": 1.2128,
+      "step": 1440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019919111277824221,
+      "loss": 1.4467,
+      "step": 1441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019918913410836014,
+      "loss": 1.4872,
+      "step": 1442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019918715303121078,
+      "loss": 1.4001,
+      "step": 1443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019918516954684222,
+      "loss": 1.2869,
+      "step": 1444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019918318365530252,
+      "loss": 1.4656,
+      "step": 1445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019918119535663998,
+      "loss": 1.2494,
+      "step": 1446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019917920465090276,
+      "loss": 1.4601,
+      "step": 1447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019917721153813923,
+      "loss": 1.4127,
+      "step": 1448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019917521601839778,
+      "loss": 1.2218,
+      "step": 1449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019917321809172675,
+      "loss": 1.2006,
+      "step": 1450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991712177581747,
+      "loss": 1.3047,
+      "step": 1451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019916921501779018,
+      "loss": 1.1677,
+      "step": 1452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991672098706218,
+      "loss": 2.481,
+      "step": 1453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019916520231671816,
+      "loss": 1.3733,
+      "step": 1454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019916319235612805,
+      "loss": 1.3075,
+      "step": 1455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991611799889002,
+      "loss": 1.3938,
+      "step": 1456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991591652150835,
+      "loss": 1.2246,
+      "step": 1457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019915714803472677,
+      "loss": 1.3749,
+      "step": 1458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019915512844787908,
+      "loss": 1.2305,
+      "step": 1459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019915310645458935,
+      "loss": 1.4626,
+      "step": 1460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019915108205490668,
+      "loss": 1.3023,
+      "step": 1461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019914905524888026,
+      "loss": 1.0969,
+      "step": 1462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019914702603655916,
+      "loss": 1.3482,
+      "step": 1463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019914499441799272,
+      "loss": 1.1758,
+      "step": 1464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019914296039323022,
+      "loss": 1.3914,
+      "step": 1465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019914092396232104,
+      "loss": 1.2523,
+      "step": 1466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019913888512531456,
+      "loss": 2.541,
+      "step": 1467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019913684388226028,
+      "loss": 1.2592,
+      "step": 1468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019913480023320775,
+      "loss": 1.2163,
+      "step": 1469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991327541782066,
+      "loss": 1.3896,
+      "step": 1470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019913070571730642,
+      "loss": 1.3684,
+      "step": 1471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000199128654850557,
+      "loss": 1.2441,
+      "step": 1472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019912660157800805,
+      "loss": 1.2774,
+      "step": 1473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991245458997094,
+      "loss": 1.3297,
+      "step": 1474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019912248781571096,
+      "loss": 1.6476,
+      "step": 1475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019912042732606272,
+      "loss": 1.4457,
+      "step": 1476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019911836443081465,
+      "loss": 1.1158,
+      "step": 1477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019911629913001679,
+      "loss": 1.3357,
+      "step": 1478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001991142314237193,
+      "loss": 1.2267,
+      "step": 1479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019911216131197234,
+      "loss": 1.4116,
+      "step": 1480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019911008879482618,
+      "loss": 1.2283,
+      "step": 1481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019910801387233108,
+      "loss": 1.436,
+      "step": 1482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019910593654453744,
+      "loss": 1.1631,
+      "step": 1483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019910385681149562,
+      "loss": 1.5494,
+      "step": 1484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019910177467325616,
+      "loss": 1.3674,
+      "step": 1485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019909969012986954,
+      "loss": 1.3983,
+      "step": 1486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001990976031813864,
+      "loss": 1.4557,
+      "step": 1487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019909551382785731,
+      "loss": 1.4828,
+      "step": 1488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019909342206933306,
+      "loss": 1.3266,
+      "step": 1489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001990913279058644,
+      "loss": 1.5596,
+      "step": 1490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019908923133750208,
+      "loss": 1.4944,
+      "step": 1491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019908713236429707,
+      "loss": 1.4775,
+      "step": 1492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001990850309863003,
+      "loss": 1.2372,
+      "step": 1493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001990829272035627,
+      "loss": 1.3052,
+      "step": 1494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019908082101613542,
+      "loss": 1.2774,
+      "step": 1495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019907871242406952,
+      "loss": 1.4888,
+      "step": 1496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019907660142741619,
+      "loss": 2.4226,
+      "step": 1497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019907448802622665,
+      "loss": 2.2137,
+      "step": 1498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019907237222055224,
+      "loss": 1.0767,
+      "step": 1499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019907025401044425,
+      "loss": 1.3332,
+      "step": 1500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019906813339595409,
+      "loss": 1.3849,
+      "step": 1501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019906601037713326,
+      "loss": 1.4187,
+      "step": 1502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019906388495403326,
+      "loss": 1.3795,
+      "step": 1503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001990617571267057,
+      "loss": 1.3226,
+      "step": 1504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001990596268952022,
+      "loss": 1.4041,
+      "step": 1505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019905749425957445,
+      "loss": 1.4164,
+      "step": 1506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019905535921987423,
+      "loss": 1.27,
+      "step": 1507
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019905322177615333,
+      "loss": 1.2452,
+      "step": 1508
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019905108192846366,
+      "loss": 1.4108,
+      "step": 1509
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019904893967685717,
+      "loss": 1.3245,
+      "step": 1510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019904679502138578,
+      "loss": 1.4313,
+      "step": 1511
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019904464796210156,
+      "loss": 1.4353,
+      "step": 1512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019904249849905665,
+      "loss": 1.6332,
+      "step": 1513
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990403466323032,
+      "loss": 1.1067,
+      "step": 1514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019903819236189344,
+      "loss": 1.2529,
+      "step": 1515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019903603568787964,
+      "loss": 1.5041,
+      "step": 1516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019903387661031414,
+      "loss": 1.3623,
+      "step": 1517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990317151292494,
+      "loss": 1.4736,
+      "step": 1518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019902955124473777,
+      "loss": 1.4678,
+      "step": 1519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019902738495683185,
+      "loss": 1.5181,
+      "step": 1520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019902521626558416,
+      "loss": 1.3259,
+      "step": 1521
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990230451710474,
+      "loss": 1.4799,
+      "step": 1522
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990208716732742,
+      "loss": 1.3517,
+      "step": 1523
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019901869577231734,
+      "loss": 1.1886,
+      "step": 1524
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019901651746822964,
+      "loss": 1.299,
+      "step": 1525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990143367610639,
+      "loss": 1.3921,
+      "step": 1526
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019901215365087312,
+      "loss": 1.3165,
+      "step": 1527
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019900996813771024,
+      "loss": 1.2281,
+      "step": 1528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990077802216283,
+      "loss": 1.396,
+      "step": 1529
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019900558990268045,
+      "loss": 1.3363,
+      "step": 1530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990033971809198,
+      "loss": 2.6904,
+      "step": 1531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001990012020563996,
+      "loss": 1.4464,
+      "step": 1532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019899900452917307,
+      "loss": 1.3493,
+      "step": 1533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989968045992936,
+      "loss": 1.4495,
+      "step": 1534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019899460226681454,
+      "loss": 2.5449,
+      "step": 1535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019899239753178936,
+      "loss": 1.2609,
+      "step": 1536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019899019039427155,
+      "loss": 1.3409,
+      "step": 1537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989879808543147,
+      "loss": 1.4409,
+      "step": 1538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019898576891197244,
+      "loss": 1.5015,
+      "step": 1539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019898355456729846,
+      "loss": 2.4626,
+      "step": 1540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019898133782034642,
+      "loss": 1.6218,
+      "step": 1541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989791186711702,
+      "loss": 1.3379,
+      "step": 1542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019897689711982365,
+      "loss": 1.2855,
+      "step": 1543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019897467316636064,
+      "loss": 1.4173,
+      "step": 1544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989724468108352,
+      "loss": 1.4475,
+      "step": 1545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019897021805330133,
+      "loss": 2.3887,
+      "step": 1546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019896798689381315,
+      "loss": 1.4346,
+      "step": 1547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019896575333242476,
+      "loss": 1.3881,
+      "step": 1548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019896351736919038,
+      "loss": 1.2063,
+      "step": 1549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019896127900416436,
+      "loss": 1.2668,
+      "step": 1550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019895903823740087,
+      "loss": 1.2391,
+      "step": 1551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989567950689544,
+      "loss": 1.4105,
+      "step": 1552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019895454949887942,
+      "loss": 1.6268,
+      "step": 1553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989523015272303,
+      "loss": 1.3533,
+      "step": 1554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989500511540617,
+      "loss": 1.4204,
+      "step": 1555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019894779837942826,
+      "loss": 1.34,
+      "step": 1556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019894554320338455,
+      "loss": 1.2388,
+      "step": 1557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019894328562598536,
+      "loss": 2.144,
+      "step": 1558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019894102564728546,
+      "loss": 1.6028,
+      "step": 1559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019893876326733972,
+      "loss": 1.3203,
+      "step": 1560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019893649848620306,
+      "loss": 1.3303,
+      "step": 1561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989342313039304,
+      "loss": 1.3973,
+      "step": 1562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019893196172057683,
+      "loss": 1.5212,
+      "step": 1563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019892968973619733,
+      "loss": 1.379,
+      "step": 1564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019892741535084714,
+      "loss": 1.5262,
+      "step": 1565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019892513856458138,
+      "loss": 1.2191,
+      "step": 1566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019892285937745538,
+      "loss": 1.4086,
+      "step": 1567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019892057778952442,
+      "loss": 1.3751,
+      "step": 1568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019891829380084386,
+      "loss": 1.1333,
+      "step": 1569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019891600741146913,
+      "loss": 1.3796,
+      "step": 1570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019891371862145577,
+      "loss": 1.2065,
+      "step": 1571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019891142743085923,
+      "loss": 1.3066,
+      "step": 1572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001989091338397352,
+      "loss": 1.3521,
+      "step": 1573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019890683784813936,
+      "loss": 1.353,
+      "step": 1574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019890453945612733,
+      "loss": 1.3843,
+      "step": 1575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019890223866375498,
+      "loss": 1.3361,
+      "step": 1576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019889993547107812,
+      "loss": 1.3272,
+      "step": 1577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019889762987815264,
+      "loss": 1.3306,
+      "step": 1578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019889532188503454,
+      "loss": 1.4775,
+      "step": 1579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019889301149177976,
+      "loss": 1.3392,
+      "step": 1580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988906986984444,
+      "loss": 1.0821,
+      "step": 1581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019888838350508463,
+      "loss": 1.2904,
+      "step": 1582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019888606591175664,
+      "loss": 2.3408,
+      "step": 1583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019888374591851658,
+      "loss": 1.4431,
+      "step": 1584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019888142352542088,
+      "loss": 1.3941,
+      "step": 1585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988790987325258,
+      "loss": 1.1987,
+      "step": 1586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019887677153988782,
+      "loss": 1.3639,
+      "step": 1587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988744419475634,
+      "loss": 1.3235,
+      "step": 1588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988721099556091,
+      "loss": 1.281,
+      "step": 1589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019886977556408147,
+      "loss": 1.3613,
+      "step": 1590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988674387730372,
+      "loss": 1.4803,
+      "step": 1591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000198865099582533,
+      "loss": 1.366,
+      "step": 1592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019886275799262565,
+      "loss": 1.2172,
+      "step": 1593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019886041400337198,
+      "loss": 1.4335,
+      "step": 1594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019885806761482883,
+      "loss": 1.4559,
+      "step": 1595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988557188270532,
+      "loss": 1.4895,
+      "step": 1596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019885336764010207,
+      "loss": 1.3853,
+      "step": 1597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988510140540325,
+      "loss": 1.4191,
+      "step": 1598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019884865806890162,
+      "loss": 1.2427,
+      "step": 1599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988462996847666,
+      "loss": 1.3309,
+      "step": 1600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019884393890168468,
+      "loss": 2.272,
+      "step": 1601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019884157571971317,
+      "loss": 1.3972,
+      "step": 1602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988392101389094,
+      "loss": 1.3771,
+      "step": 1603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019883684215933078,
+      "loss": 1.3771,
+      "step": 1604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988344717810348,
+      "loss": 1.5574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019883209900407897,
+      "loss": 1.4735,
+      "step": 1606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019882972382852091,
+      "loss": 1.2806,
+      "step": 1607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988273462544182,
+      "loss": 1.3207,
+      "step": 1608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019882496628182862,
+      "loss": 1.3989,
+      "step": 1609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019882258391080987,
+      "loss": 1.3307,
+      "step": 1610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001988201991414198,
+      "loss": 1.3369,
+      "step": 1611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019881781197371624,
+      "loss": 1.3325,
+      "step": 1612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001988154224077572,
+      "loss": 1.2339,
+      "step": 1613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019881303044360065,
+      "loss": 1.4139,
+      "step": 1614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001988106360813046,
+      "loss": 1.2574,
+      "step": 1615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019880823932092717,
+      "loss": 1.1522,
+      "step": 1616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019880584016252658,
+      "loss": 1.42,
+      "step": 1617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019880343860616102,
+      "loss": 1.3337,
+      "step": 1618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019880103465188875,
+      "loss": 1.2777,
+      "step": 1619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019879862829976817,
+      "loss": 1.3268,
+      "step": 1620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019879621954985762,
+      "loss": 1.2748,
+      "step": 1621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019879380840221562,
+      "loss": 1.1884,
+      "step": 1622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019879139485690065,
+      "loss": 1.3275,
+      "step": 1623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019878897891397127,
+      "loss": 1.2805,
+      "step": 1624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019878656057348615,
+      "loss": 1.4293,
+      "step": 1625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019878413983550396,
+      "loss": 1.46,
+      "step": 1626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019878171670008346,
+      "loss": 1.213,
+      "step": 1627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019877929116728346,
+      "loss": 2.3972,
+      "step": 1628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001987768632371628,
+      "loss": 1.3728,
+      "step": 1629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019877443290978046,
+      "loss": 1.3592,
+      "step": 1630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019877200018519537,
+      "loss": 1.3641,
+      "step": 1631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001987695650634666,
+      "loss": 1.5263,
+      "step": 1632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019876712754465323,
+      "loss": 1.326,
+      "step": 1633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019876468762881444,
+      "loss": 1.3033,
+      "step": 1634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019876224531600943,
+      "loss": 1.3423,
+      "step": 1635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019875980060629744,
+      "loss": 1.5399,
+      "step": 1636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019875735349973787,
+      "loss": 1.1664,
+      "step": 1637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019875490399639008,
+      "loss": 1.5121,
+      "step": 1638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001987524520963135,
+      "loss": 1.2173,
+      "step": 1639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019874999779956768,
+      "loss": 2.0286,
+      "step": 1640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019874754110621212,
+      "loss": 1.2538,
+      "step": 1641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019874508201630652,
+      "loss": 1.5557,
+      "step": 1642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019874262052991046,
+      "loss": 2.2576,
+      "step": 1643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001987401566470838,
+      "loss": 1.1409,
+      "step": 1644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019873769036788622,
+      "loss": 1.3545,
+      "step": 1645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019873522169237769,
+      "loss": 1.4097,
+      "step": 1646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019873275062061804,
+      "loss": 1.3505,
+      "step": 1647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019873027715266726,
+      "loss": 1.1072,
+      "step": 1648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019872780128858537,
+      "loss": 1.2184,
+      "step": 1649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001987253230284325,
+      "loss": 2.1921,
+      "step": 1650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019872284237226875,
+      "loss": 1.2704,
+      "step": 1651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019872035932015435,
+      "loss": 1.363,
+      "step": 1652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019871787387214958,
+      "loss": 1.267,
+      "step": 1653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001987153860283147,
+      "loss": 1.2054,
+      "step": 1654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019871289578871014,
+      "loss": 1.5161,
+      "step": 1655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019871040315339633,
+      "loss": 1.1838,
+      "step": 1656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019870790812243376,
+      "loss": 1.4406,
+      "step": 1657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019870541069588296,
+      "loss": 1.3598,
+      "step": 1658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019870291087380456,
+      "loss": 1.5849,
+      "step": 1659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019870040865625925,
+      "loss": 1.4438,
+      "step": 1660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019869790404330773,
+      "loss": 1.3679,
+      "step": 1661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986953970350108,
+      "loss": 1.2437,
+      "step": 1662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986928876314293,
+      "loss": 1.3528,
+      "step": 1663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019869037583262412,
+      "loss": 1.2974,
+      "step": 1664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019868786163865622,
+      "loss": 1.1664,
+      "step": 1665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019868534504958665,
+      "loss": 1.4449,
+      "step": 1666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019868282606547646,
+      "loss": 1.2401,
+      "step": 1667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019868030468638676,
+      "loss": 1.3789,
+      "step": 1668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986777809123788,
+      "loss": 1.3149,
+      "step": 1669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019867525474351382,
+      "loss": 1.3469,
+      "step": 1670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019867272617985306,
+      "loss": 1.2805,
+      "step": 1671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019867019522145797,
+      "loss": 1.1955,
+      "step": 1672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019866766186838993,
+      "loss": 1.3085,
+      "step": 1673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019866512612071047,
+      "loss": 1.399,
+      "step": 1674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019866258797848107,
+      "loss": 1.3853,
+      "step": 1675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019866004744176335,
+      "loss": 1.2859,
+      "step": 1676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019865750451061894,
+      "loss": 1.4097,
+      "step": 1677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019865495918510963,
+      "loss": 1.4717,
+      "step": 1678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019865241146529717,
+      "loss": 1.4397,
+      "step": 1679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019864986135124336,
+      "loss": 1.3109,
+      "step": 1680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019864730884301007,
+      "loss": 1.3412,
+      "step": 1681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019864475394065932,
+      "loss": 1.4032,
+      "step": 1682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019864219664425306,
+      "loss": 1.214,
+      "step": 1683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019863963695385337,
+      "loss": 1.2105,
+      "step": 1684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986370748695224,
+      "loss": 1.2984,
+      "step": 1685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019863451039132227,
+      "loss": 1.2432,
+      "step": 1686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019863194351931528,
+      "loss": 1.4362,
+      "step": 1687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019862937425356369,
+      "loss": 1.3925,
+      "step": 1688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986268025941299,
+      "loss": 1.3543,
+      "step": 1689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019862422854107626,
+      "loss": 1.4346,
+      "step": 1690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019862165209446527,
+      "loss": 1.2203,
+      "step": 1691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019861907325435945,
+      "loss": 1.3715,
+      "step": 1692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019861649202082137,
+      "loss": 1.4133,
+      "step": 1693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019861390839391374,
+      "loss": 1.2842,
+      "step": 1694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019861132237369923,
+      "loss": 1.4491,
+      "step": 1695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019860873396024059,
+      "loss": 1.3616,
+      "step": 1696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986061431536006,
+      "loss": 1.2916,
+      "step": 1697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001986035499538422,
+      "loss": 1.3693,
+      "step": 1698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019860095436102833,
+      "loss": 1.3739,
+      "step": 1699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019859835637522197,
+      "loss": 1.459,
+      "step": 1700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019859575599648612,
+      "loss": 1.3842,
+      "step": 1701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019859315322488395,
+      "loss": 1.5579,
+      "step": 1702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019859054806047865,
+      "loss": 1.3315,
+      "step": 1703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019858794050333338,
+      "loss": 1.2994,
+      "step": 1704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019858533055351144,
+      "loss": 1.4409,
+      "step": 1705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019858271821107616,
+      "loss": 1.5114,
+      "step": 1706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019858010347609102,
+      "loss": 2.1318,
+      "step": 1707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019857748634861936,
+      "loss": 1.3005,
+      "step": 1708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001985748668287248,
+      "loss": 1.2803,
+      "step": 1709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019857224491647086,
+      "loss": 1.3036,
+      "step": 1710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001985696206119212,
+      "loss": 1.3179,
+      "step": 1711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019856699391513946,
+      "loss": 1.1817,
+      "step": 1712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019856436482618942,
+      "loss": 1.4294,
+      "step": 1713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019856173334513493,
+      "loss": 1.354,
+      "step": 1714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001985590994720398,
+      "loss": 1.4125,
+      "step": 1715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019855646320696794,
+      "loss": 1.1793,
+      "step": 1716
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019855382454998338,
+      "loss": 1.3584,
+      "step": 1717
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001985511835011501,
+      "loss": 1.4469,
+      "step": 1718
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001985485400605323,
+      "loss": 1.3284,
+      "step": 1719
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019854589422819402,
+      "loss": 1.2076,
+      "step": 1720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019854324600419957,
+      "loss": 1.2736,
+      "step": 1721
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019854059538861312,
+      "loss": 1.1725,
+      "step": 1722
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019853794238149905,
+      "loss": 1.3113,
+      "step": 1723
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019853528698292176,
+      "loss": 1.5062,
+      "step": 1724
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019853262919294567,
+      "loss": 1.3578,
+      "step": 1725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019852996901163533,
+      "loss": 1.3783,
+      "step": 1726
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019852730643905525,
+      "loss": 1.382,
+      "step": 1727
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001985246414752701,
+      "loss": 1.3376,
+      "step": 1728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019852197412034448,
+      "loss": 1.3921,
+      "step": 1729
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001985193043743432,
+      "loss": 1.4752,
+      "step": 1730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000198516632237331,
+      "loss": 1.3753,
+      "step": 1731
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019851395770937273,
+      "loss": 1.1561,
+      "step": 1732
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019851128079053333,
+      "loss": 1.3932,
+      "step": 1733
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001985086014808778,
+      "loss": 1.4188,
+      "step": 1734
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001985059197804711,
+      "loss": 1.426,
+      "step": 1735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019850323568937835,
+      "loss": 1.4018,
+      "step": 1736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019850054920766467,
+      "loss": 1.3332,
+      "step": 1737
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984978603353953,
+      "loss": 1.2999,
+      "step": 1738
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019849516907263542,
+      "loss": 1.245,
+      "step": 1739
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019849247541945043,
+      "loss": 1.2296,
+      "step": 1740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019848977937590565,
+      "loss": 2.4915,
+      "step": 1741
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019848708094206654,
+      "loss": 1.2448,
+      "step": 1742
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019848438011799857,
+      "loss": 1.571,
+      "step": 1743
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984816769037673,
+      "loss": 1.2971,
+      "step": 1744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019847897129943832,
+      "loss": 1.3862,
+      "step": 1745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019847626330507731,
+      "loss": 1.3227,
+      "step": 1746
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019847355292075,
+      "loss": 1.4326,
+      "step": 1747
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019847084014652216,
+      "loss": 1.2124,
+      "step": 1748
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984681249824596,
+      "loss": 1.4976,
+      "step": 1749
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019846540742862823,
+      "loss": 1.3144,
+      "step": 1750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019846268748509405,
+      "loss": 1.3876,
+      "step": 1751
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000198459965151923,
+      "loss": 1.4425,
+      "step": 1752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984572404291812,
+      "loss": 1.3728,
+      "step": 1753
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019845451331693476,
+      "loss": 1.2633,
+      "step": 1754
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019845178381524986,
+      "loss": 1.3228,
+      "step": 1755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019844905192419277,
+      "loss": 1.2717,
+      "step": 1756
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019844631764382974,
+      "loss": 1.321,
+      "step": 1757
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984435809742272,
+      "loss": 1.445,
+      "step": 1758
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984408419154515,
+      "loss": 1.5338,
+      "step": 1759
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019843810046756913,
+      "loss": 1.3071,
+      "step": 1760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019843535663064668,
+      "loss": 1.4542,
+      "step": 1761
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019843261040475068,
+      "loss": 1.1802,
+      "step": 1762
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984298617899478,
+      "loss": 1.189,
+      "step": 1763
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019842711078630476,
+      "loss": 1.4727,
+      "step": 1764
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984243573938883,
+      "loss": 1.3444,
+      "step": 1765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019842160161276522,
+      "loss": 1.4611,
+      "step": 1766
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984188434430025,
+      "loss": 1.4852,
+      "step": 1767
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019841608288466696,
+      "loss": 1.2761,
+      "step": 1768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019841331993782568,
+      "loss": 1.3477,
+      "step": 1769
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019841055460254568,
+      "loss": 1.3934,
+      "step": 1770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001984077868788941,
+      "loss": 1.2669,
+      "step": 1771
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019840501676693807,
+      "loss": 1.4415,
+      "step": 1772
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019840224426674485,
+      "loss": 1.4117,
+      "step": 1773
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983994693783817,
+      "loss": 1.2072,
+      "step": 1774
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019839669210191602,
+      "loss": 1.4197,
+      "step": 1775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019839391243741514,
+      "loss": 1.448,
+      "step": 1776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983911303849466,
+      "loss": 1.4377,
+      "step": 1777
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019838834594457786,
+      "loss": 1.441,
+      "step": 1778
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983855591163765,
+      "loss": 1.2754,
+      "step": 1779
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019838276990041015,
+      "loss": 1.4517,
+      "step": 1780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019837997829674658,
+      "loss": 1.4423,
+      "step": 1781
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019837718430545343,
+      "loss": 1.161,
+      "step": 1782
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019837438792659858,
+      "loss": 1.3768,
+      "step": 1783
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019837158916024986,
+      "loss": 1.2401,
+      "step": 1784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019836878800647524,
+      "loss": 2.4258,
+      "step": 1785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019836598446534268,
+      "loss": 2.6602,
+      "step": 1786
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983631785369202,
+      "loss": 1.3002,
+      "step": 1787
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983603702212759,
+      "loss": 2.3352,
+      "step": 1788
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019835755951847793,
+      "loss": 1.3635,
+      "step": 1789
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019835474642859458,
+      "loss": 1.1349,
+      "step": 1790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019835193095169406,
+      "loss": 1.2309,
+      "step": 1791
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019834911308784468,
+      "loss": 1.295,
+      "step": 1792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019834629283711484,
+      "loss": 1.5465,
+      "step": 1793
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000198343470199573,
+      "loss": 1.3639,
+      "step": 1794
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983406451752877,
+      "loss": 1.3755,
+      "step": 1795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019833781776432745,
+      "loss": 1.3419,
+      "step": 1796
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019833498796676088,
+      "loss": 1.3041,
+      "step": 1797
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983321557826567,
+      "loss": 1.4042,
+      "step": 1798
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983293212120836,
+      "loss": 1.4758,
+      "step": 1799
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019832648425511037,
+      "loss": 1.2054,
+      "step": 1800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983236449118059,
+      "loss": 1.3898,
+      "step": 1801
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983208031822391,
+      "loss": 1.2371,
+      "step": 1802
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983179590664789,
+      "loss": 1.3584,
+      "step": 1803
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019831511256459435,
+      "loss": 1.2725,
+      "step": 1804
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983122636766545,
+      "loss": 1.443,
+      "step": 1805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019830941240272859,
+      "loss": 1.3124,
+      "step": 1806
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001983065587428857,
+      "loss": 1.2633,
+      "step": 1807
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019830370269719513,
+      "loss": 1.2477,
+      "step": 1808
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019830084426572623,
+      "loss": 1.2917,
+      "step": 1809
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019829798344854832,
+      "loss": 1.4185,
+      "step": 1810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019829512024573088,
+      "loss": 1.3132,
+      "step": 1811
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019829225465734334,
+      "loss": 1.4532,
+      "step": 1812
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001982893866834553,
+      "loss": 1.2065,
+      "step": 1813
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001982865163241363,
+      "loss": 3.1946,
+      "step": 1814
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019828364357945608,
+      "loss": 1.3581,
+      "step": 1815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019828076844948434,
+      "loss": 2.5452,
+      "step": 1816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019827789093429082,
+      "loss": 1.3444,
+      "step": 1817
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019827501103394536,
+      "loss": 1.5533,
+      "step": 1818
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001982721287485179,
+      "loss": 1.4025,
+      "step": 1819
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019826924407807835,
+      "loss": 1.296,
+      "step": 1820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019826635702269673,
+      "loss": 1.2987,
+      "step": 1821
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019826346758244312,
+      "loss": 1.452,
+      "step": 1822
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019826057575738764,
+      "loss": 1.2521,
+      "step": 1823
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019825768154760044,
+      "loss": 1.2429,
+      "step": 1824
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019825478495315182,
+      "loss": 1.4266,
+      "step": 1825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019825188597411204,
+      "loss": 1.3057,
+      "step": 1826
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019824898461055145,
+      "loss": 1.4429,
+      "step": 1827
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019824608086254047,
+      "loss": 1.3923,
+      "step": 1828
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982431747301496,
+      "loss": 1.5514,
+      "step": 1829
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019824026621344934,
+      "loss": 2.6946,
+      "step": 1830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982373553125103,
+      "loss": 1.3488,
+      "step": 1831
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019823444202740314,
+      "loss": 1.2719,
+      "step": 1832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982315263581985,
+      "loss": 1.3162,
+      "step": 1833
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982286083049672,
+      "loss": 1.3442,
+      "step": 1834
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019822568786778005,
+      "loss": 1.2705,
+      "step": 1835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019822276504670788,
+      "loss": 1.3514,
+      "step": 1836
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982198398418217,
+      "loss": 1.2843,
+      "step": 1837
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019821691225319245,
+      "loss": 1.2325,
+      "step": 1838
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982139822808912,
+      "loss": 1.3709,
+      "step": 1839
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019821104992498906,
+      "loss": 1.449,
+      "step": 1840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982081151855572,
+      "loss": 1.3032,
+      "step": 1841
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019820517806266682,
+      "loss": 1.569,
+      "step": 1842
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001982022385563892,
+      "loss": 1.3336,
+      "step": 1843
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981992966667958,
+      "loss": 1.2676,
+      "step": 1844
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981963523939578,
+      "loss": 1.3756,
+      "step": 1845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981934057379468,
+      "loss": 1.1925,
+      "step": 1846
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019819045669883435,
+      "loss": 1.1528,
+      "step": 1847
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981875052766919,
+      "loss": 1.4922,
+      "step": 1848
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019818455147159117,
+      "loss": 1.3015,
+      "step": 1849
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019818159528360382,
+      "loss": 1.2695,
+      "step": 1850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019817863671280157,
+      "loss": 1.288,
+      "step": 1851
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019817567575925626,
+      "loss": 1.2379,
+      "step": 1852
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019817271242303974,
+      "loss": 1.3836,
+      "step": 1853
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981697467042239,
+      "loss": 1.4023,
+      "step": 1854
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019816677860288076,
+      "loss": 1.4417,
+      "step": 1855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019816380811908232,
+      "loss": 1.2042,
+      "step": 1856
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981608352529007,
+      "loss": 1.3878,
+      "step": 1857
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000198157860004408,
+      "loss": 1.432,
+      "step": 1858
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981548823736765,
+      "loss": 1.2414,
+      "step": 1859
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019815190236077845,
+      "loss": 1.4075,
+      "step": 1860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019814891996578613,
+      "loss": 1.3357,
+      "step": 1861
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019814593518877196,
+      "loss": 1.3896,
+      "step": 1862
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019814294802980833,
+      "loss": 1.3719,
+      "step": 1863
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019813995848896778,
+      "loss": 1.2733,
+      "step": 1864
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019813696656632284,
+      "loss": 1.5105,
+      "step": 1865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981339722619462,
+      "loss": 1.304,
+      "step": 1866
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981309755759104,
+      "loss": 1.1765,
+      "step": 1867
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019812797650828828,
+      "loss": 1.0733,
+      "step": 1868
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019812497505915255,
+      "loss": 1.4728,
+      "step": 1869
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981219712285761,
+      "loss": 1.2898,
+      "step": 1870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981189650166318,
+      "loss": 1.0563,
+      "step": 1871
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019811595642339262,
+      "loss": 1.4003,
+      "step": 1872
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019811294544893162,
+      "loss": 1.2739,
+      "step": 1873
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981099320933218,
+      "loss": 1.3672,
+      "step": 1874
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019810691635663635,
+      "loss": 1.3521,
+      "step": 1875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019810389823894843,
+      "loss": 1.3331,
+      "step": 1876
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001981008777403313,
+      "loss": 1.2578,
+      "step": 1877
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019809785486085826,
+      "loss": 1.3132,
+      "step": 1878
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019809482960060266,
+      "loss": 1.375,
+      "step": 1879
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019809180195963798,
+      "loss": 1.3902,
+      "step": 1880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019808877193803763,
+      "loss": 1.4234,
+      "step": 1881
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980857395358752,
+      "loss": 1.277,
+      "step": 1882
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980827047532242,
+      "loss": 1.5059,
+      "step": 1883
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019807966759015838,
+      "loss": 1.3479,
+      "step": 1884
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019807662804675142,
+      "loss": 1.328,
+      "step": 1885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980735861230771,
+      "loss": 1.3768,
+      "step": 1886
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980705418192092,
+      "loss": 1.3099,
+      "step": 1887
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019806749513522163,
+      "loss": 1.395,
+      "step": 1888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019806444607118833,
+      "loss": 1.3864,
+      "step": 1889
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980613946271833,
+      "loss": 1.3842,
+      "step": 1890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980583408032806,
+      "loss": 1.3937,
+      "step": 1891
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019805528459955433,
+      "loss": 1.3686,
+      "step": 1892
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980522260160787,
+      "loss": 1.4852,
+      "step": 1893
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980491650529279,
+      "loss": 1.451,
+      "step": 1894
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019804610171017623,
+      "loss": 2.9058,
+      "step": 1895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019804303598789803,
+      "loss": 1.1128,
+      "step": 1896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019803996788616773,
+      "loss": 1.304,
+      "step": 1897
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019803689740505978,
+      "loss": 1.3882,
+      "step": 1898
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019803382454464867,
+      "loss": 1.1032,
+      "step": 1899
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000198030749305009,
+      "loss": 1.4478,
+      "step": 1900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019802767168621542,
+      "loss": 1.3394,
+      "step": 1901
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980245916883426,
+      "loss": 1.2205,
+      "step": 1902
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019802150931146528,
+      "loss": 1.4224,
+      "step": 1903
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019801842455565832,
+      "loss": 1.3174,
+      "step": 1904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001980153374209965,
+      "loss": 1.4989,
+      "step": 1905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019801224790755482,
+      "loss": 1.3573,
+      "step": 1906
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019800915601540821,
+      "loss": 1.109,
+      "step": 1907
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019800606174463176,
+      "loss": 1.4109,
+      "step": 1908
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019800296509530054,
+      "loss": 1.4599,
+      "step": 1909
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019799986606748967,
+      "loss": 1.2398,
+      "step": 1910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001979967646612744,
+      "loss": 1.4049,
+      "step": 1911
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019799366087673,
+      "loss": 1.3743,
+      "step": 1912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001979905547139318,
+      "loss": 1.259,
+      "step": 1913
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019798744617295516,
+      "loss": 1.3524,
+      "step": 1914
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019798433525387553,
+      "loss": 1.266,
+      "step": 1915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019798122195676844,
+      "loss": 1.5295,
+      "step": 1916
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019797810628170942,
+      "loss": 1.306,
+      "step": 1917
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019797498822877408,
+      "loss": 1.3245,
+      "step": 1918
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001979718677980381,
+      "loss": 1.5157,
+      "step": 1919
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019796874498957722,
+      "loss": 1.1952,
+      "step": 1920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001979656198034672,
+      "loss": 1.5042,
+      "step": 1921
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000197962492239784,
+      "loss": 1.526,
+      "step": 1922
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019795936229860336,
+      "loss": 2.9814,
+      "step": 1923
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019795622998000132,
+      "loss": 1.5652,
+      "step": 1924
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001979530952840539,
+      "loss": 1.3555,
+      "step": 1925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001979499582108372,
+      "loss": 1.3326,
+      "step": 1926
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001979468187604273,
+      "loss": 1.3594,
+      "step": 1927
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019794367693290043,
+      "loss": 1.2976,
+      "step": 1928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019794053272833285,
+      "loss": 1.2692,
+      "step": 1929
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019793738614680084,
+      "loss": 1.2911,
+      "step": 1930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019793423718838079,
+      "loss": 1.4584,
+      "step": 1931
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019793108585314907,
+      "loss": 1.3617,
+      "step": 1932
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019792793214118225,
+      "loss": 1.2573,
+      "step": 1933
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019792477605255678,
+      "loss": 1.4785,
+      "step": 1934
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019792161758734933,
+      "loss": 1.2559,
+      "step": 1935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001979184567456365,
+      "loss": 1.3983,
+      "step": 1936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019791529352749504,
+      "loss": 1.4247,
+      "step": 1937
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019791212793300172,
+      "loss": 1.2472,
+      "step": 1938
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001979089599622333,
+      "loss": 1.2368,
+      "step": 1939
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019790578961526676,
+      "loss": 1.3304,
+      "step": 1940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019790261689217896,
+      "loss": 1.4554,
+      "step": 1941
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019789944179304697,
+      "loss": 3.3938,
+      "step": 1942
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019789626431794782,
+      "loss": 1.4618,
+      "step": 1943
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001978930844669586,
+      "loss": 1.4383,
+      "step": 1944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019788990224015648,
+      "loss": 1.3967,
+      "step": 1945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019788671763761877,
+      "loss": 1.1583,
+      "step": 1946
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001978835306594227,
+      "loss": 1.158,
+      "step": 1947
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019788034130564558,
+      "loss": 1.0843,
+      "step": 1948
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001978771495763649,
+      "loss": 1.4836,
+      "step": 1949
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019787395547165803,
+      "loss": 1.4208,
+      "step": 1950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019787075899160256,
+      "loss": 1.4829,
+      "step": 1951
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019786756013627605,
+      "loss": 1.1933,
+      "step": 1952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019786435890575612,
+      "loss": 1.4327,
+      "step": 1953
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019786115530012047,
+      "loss": 1.3044,
+      "step": 1954
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019785794931944682,
+      "loss": 1.6093,
+      "step": 1955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019785474096381306,
+      "loss": 2.7444,
+      "step": 1956
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019785153023329697,
+      "loss": 1.2032,
+      "step": 1957
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001978483171279765,
+      "loss": 1.3243,
+      "step": 1958
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019784510164792967,
+      "loss": 1.3154,
+      "step": 1959
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019784188379323446,
+      "loss": 1.2316,
+      "step": 1960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019783866356396898,
+      "loss": 1.3145,
+      "step": 1961
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001978354409602114,
+      "loss": 1.2709,
+      "step": 1962
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019783221598203992,
+      "loss": 1.3539,
+      "step": 1963
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001978289886295328,
+      "loss": 1.3864,
+      "step": 1964
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019782575890276837,
+      "loss": 1.3108,
+      "step": 1965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019782252680182504,
+      "loss": 1.2332,
+      "step": 1966
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019781929232678126,
+      "loss": 1.3755,
+      "step": 1967
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019781605547771546,
+      "loss": 1.4109,
+      "step": 1968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019781281625470625,
+      "loss": 1.2419,
+      "step": 1969
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019780957465783221,
+      "loss": 1.3603,
+      "step": 1970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019780633068717206,
+      "loss": 1.4098,
+      "step": 1971
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019780308434280447,
+      "loss": 1.2673,
+      "step": 1972
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001977998356248083,
+      "loss": 1.3811,
+      "step": 1973
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019779658453326234,
+      "loss": 1.1541,
+      "step": 1974
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019779333106824553,
+      "loss": 1.3242,
+      "step": 1975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019779007522983678,
+      "loss": 1.0958,
+      "step": 1976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019778681701811514,
+      "loss": 1.2828,
+      "step": 1977
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019778355643315965,
+      "loss": 1.4416,
+      "step": 1978
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001977802934750495,
+      "loss": 1.3939,
+      "step": 1979
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019777702814386387,
+      "loss": 1.4077,
+      "step": 1980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019777376043968197,
+      "loss": 1.3303,
+      "step": 1981
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019777049036258314,
+      "loss": 1.2699,
+      "step": 1982
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019776721791264673,
+      "loss": 1.3431,
+      "step": 1983
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019776394308995213,
+      "loss": 1.3779,
+      "step": 1984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019776066589457885,
+      "loss": 1.2869,
+      "step": 1985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019775738632660646,
+      "loss": 1.326,
+      "step": 1986
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019775410438611446,
+      "loss": 1.2613,
+      "step": 1987
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019775082007318263,
+      "loss": 1.272,
+      "step": 1988
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019774753338789054,
+      "loss": 1.4389,
+      "step": 1989
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019774424433031805,
+      "loss": 1.4249,
+      "step": 1990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019774095290054494,
+      "loss": 1.1832,
+      "step": 1991
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019773765909865114,
+      "loss": 1.3031,
+      "step": 1992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019773436292471654,
+      "loss": 1.2761,
+      "step": 1993
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019773106437882113,
+      "loss": 1.2389,
+      "step": 1994
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000197727763461045,
+      "loss": 1.5646,
+      "step": 1995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019772446017146827,
+      "loss": 1.2639,
+      "step": 1996
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019772115451017106,
+      "loss": 1.4357,
+      "step": 1997
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019771784647723363,
+      "loss": 1.4747,
+      "step": 1998
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019771453607273626,
+      "loss": 1.3185,
+      "step": 1999
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019771122329675928,
+      "loss": 1.4497,
+      "step": 2000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019770790814938312,
+      "loss": 1.5856,
+      "step": 2001
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001977045906306882,
+      "loss": 1.2251,
+      "step": 2002
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019770127074075506,
+      "loss": 1.2194,
+      "step": 2003
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019769794847966426,
+      "loss": 2.7451,
+      "step": 2004
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019769462384749642,
+      "loss": 1.3361,
+      "step": 2005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019769129684433227,
+      "loss": 1.4595,
+      "step": 2006
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019768796747025248,
+      "loss": 1.4272,
+      "step": 2007
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001976846357253379,
+      "loss": 1.4551,
+      "step": 2008
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019768130160966943,
+      "loss": 1.5209,
+      "step": 2009
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001976779651233279,
+      "loss": 1.4364,
+      "step": 2010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019767462626639433,
+      "loss": 1.4327,
+      "step": 2011
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019767128503894976,
+      "loss": 1.4859,
+      "step": 2012
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019766794144107525,
+      "loss": 1.3107,
+      "step": 2013
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000197664595472852,
+      "loss": 1.1105,
+      "step": 2014
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019766124713436117,
+      "loss": 2.7473,
+      "step": 2015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000197657896425684,
+      "loss": 1.5076,
+      "step": 2016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019765454334690188,
+      "loss": 1.2489,
+      "step": 2017
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019765118789809614,
+      "loss": 1.4167,
+      "step": 2018
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001976478300793482,
+      "loss": 1.3504,
+      "step": 2019
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019764446989073962,
+      "loss": 1.3774,
+      "step": 2020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019764110733235188,
+      "loss": 2.313,
+      "step": 2021
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019763774240426664,
+      "loss": 1.2846,
+      "step": 2022
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019763437510656552,
+      "loss": 1.218,
+      "step": 2023
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019763100543933025,
+      "loss": 1.1792,
+      "step": 2024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019762763340264264,
+      "loss": 1.4626,
+      "step": 2025
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019762425899658452,
+      "loss": 1.4993,
+      "step": 2026
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001976208822212378,
+      "loss": 2.5222,
+      "step": 2027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019761750307668437,
+      "loss": 1.4407,
+      "step": 2028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019761412156300628,
+      "loss": 1.282,
+      "step": 2029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001976107376802856,
+      "loss": 1.3998,
+      "step": 2030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019760735142860446,
+      "loss": 1.3007,
+      "step": 2031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000197603962808045,
+      "loss": 1.2855,
+      "step": 2032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019760057181868955,
+      "loss": 1.4737,
+      "step": 2033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019759717846062032,
+      "loss": 1.2529,
+      "step": 2034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019759378273391968,
+      "loss": 1.4651,
+      "step": 2035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019759038463867007,
+      "loss": 1.3662,
+      "step": 2036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019758698417495396,
+      "loss": 1.3912,
+      "step": 2037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019758358134285383,
+      "loss": 1.4421,
+      "step": 2038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019758017614245236,
+      "loss": 1.2966,
+      "step": 2039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001975767685738321,
+      "loss": 1.2401,
+      "step": 2040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019757335863707578,
+      "loss": 1.1548,
+      "step": 2041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019756994633226616,
+      "loss": 1.4118,
+      "step": 2042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019756653165948606,
+      "loss": 1.2413,
+      "step": 2043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019756311461881836,
+      "loss": 1.3873,
+      "step": 2044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019755969521034595,
+      "loss": 1.3897,
+      "step": 2045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019755627343415188,
+      "loss": 1.2664,
+      "step": 2046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019755284929031915,
+      "loss": 1.3298,
+      "step": 2047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019754942277893083,
+      "loss": 1.4023,
+      "step": 2048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019754599390007016,
+      "loss": 1.2127,
+      "step": 2049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001975425626538203,
+      "loss": 1.4061,
+      "step": 2050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019753912904026455,
+      "loss": 1.2292,
+      "step": 2051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019753569305948622,
+      "loss": 1.4474,
+      "step": 2052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019753225471156872,
+      "loss": 2.7346,
+      "step": 2053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019752881399659553,
+      "loss": 1.3234,
+      "step": 2054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019752537091465006,
+      "loss": 1.4885,
+      "step": 2055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019752192546581597,
+      "loss": 1.5128,
+      "step": 2056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001975184776501768,
+      "loss": 1.4395,
+      "step": 2057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001975150274678163,
+      "loss": 1.3332,
+      "step": 2058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019751157491881813,
+      "loss": 1.368,
+      "step": 2059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019750812000326616,
+      "loss": 1.3197,
+      "step": 2060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019750466272124416,
+      "loss": 1.4608,
+      "step": 2061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019750120307283607,
+      "loss": 1.2921,
+      "step": 2062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974977410581259,
+      "loss": 1.3649,
+      "step": 2063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019749427667719758,
+      "loss": 1.3279,
+      "step": 2064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019749080993013526,
+      "loss": 1.4811,
+      "step": 2065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019748734081702303,
+      "loss": 1.4547,
+      "step": 2066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019748386933794514,
+      "loss": 1.2989,
+      "step": 2067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974803954929858,
+      "loss": 1.3719,
+      "step": 2068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974769192822293,
+      "loss": 1.3686,
+      "step": 2069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019747344070576006,
+      "loss": 1.4167,
+      "step": 2070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019746995976366248,
+      "loss": 1.3477,
+      "step": 2071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019746647645602102,
+      "loss": 1.2979,
+      "step": 2072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019746299078292023,
+      "loss": 1.315,
+      "step": 2073
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019745950274444472,
+      "loss": 1.2932,
+      "step": 2074
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019745601234067914,
+      "loss": 1.3178,
+      "step": 2075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019745251957170817,
+      "loss": 1.4282,
+      "step": 2076
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974490244376166,
+      "loss": 1.2303,
+      "step": 2077
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019744552693848927,
+      "loss": 2.8621,
+      "step": 2078
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019744202707441103,
+      "loss": 1.3413,
+      "step": 2079
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019743852484546687,
+      "loss": 1.3202,
+      "step": 2080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019743502025174175,
+      "loss": 1.3649,
+      "step": 2081
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974315132933207,
+      "loss": 2.5044,
+      "step": 2082
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019742800397028887,
+      "loss": 1.238,
+      "step": 2083
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019742449228273142,
+      "loss": 1.229,
+      "step": 2084
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019742097823073358,
+      "loss": 1.5018,
+      "step": 2085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019741746181438063,
+      "loss": 1.2675,
+      "step": 2086
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974139430337579,
+      "loss": 1.211,
+      "step": 2087
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019741042188895085,
+      "loss": 1.3597,
+      "step": 2088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019740689838004482,
+      "loss": 1.2558,
+      "step": 2089
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019740337250712543,
+      "loss": 1.2396,
+      "step": 2090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973998442702782,
+      "loss": 2.634,
+      "step": 2091
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973963136695888,
+      "loss": 1.4913,
+      "step": 2092
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019739278070514288,
+      "loss": 2.542,
+      "step": 2093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019738924537702617,
+      "loss": 1.2723,
+      "step": 2094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973857076853245,
+      "loss": 1.4185,
+      "step": 2095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019738216763012372,
+      "loss": 1.3668,
+      "step": 2096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019737862521150974,
+      "loss": 1.3959,
+      "step": 2097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019737508042956854,
+      "loss": 1.249,
+      "step": 2098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019737153328438614,
+      "loss": 1.4551,
+      "step": 2099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019736798377604863,
+      "loss": 1.2888,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019736443190464218,
+      "loss": 1.3553,
+      "step": 2101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019736087767025296,
+      "loss": 1.4398,
+      "step": 2102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019735732107296723,
+      "loss": 1.1763,
+      "step": 2103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019735376211287134,
+      "loss": 1.1105,
+      "step": 2104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019735020079005163,
+      "loss": 1.348,
+      "step": 2105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019734663710459452,
+      "loss": 2.4585,
+      "step": 2106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019734307105658653,
+      "loss": 1.3737,
+      "step": 2107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973395026461142,
+      "loss": 1.3297,
+      "step": 2108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019733593187326416,
+      "loss": 1.3055,
+      "step": 2109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019733235873812302,
+      "loss": 1.2891,
+      "step": 2110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019732878324077747,
+      "loss": 1.3242,
+      "step": 2111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973252053813144,
+      "loss": 1.3907,
+      "step": 2112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019732162515982056,
+      "loss": 1.4627,
+      "step": 2113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019731804257638283,
+      "loss": 1.1361,
+      "step": 2114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019731445763108822,
+      "loss": 1.2447,
+      "step": 2115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973108703240237,
+      "loss": 1.2704,
+      "step": 2116
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973072806552763,
+      "loss": 1.3111,
+      "step": 2117
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001973036886249332,
+      "loss": 1.3615,
+      "step": 2118
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019730009423308154,
+      "loss": 1.4149,
+      "step": 2119
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019729649747980856,
+      "loss": 1.4006,
+      "step": 2120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019729289836520155,
+      "loss": 1.1508,
+      "step": 2121
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019728929688934786,
+      "loss": 1.1058,
+      "step": 2122
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019728569305233487,
+      "loss": 1.3872,
+      "step": 2123
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019728208685425012,
+      "loss": 1.3768,
+      "step": 2124
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019727847829518108,
+      "loss": 1.0551,
+      "step": 2125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001972748673752153,
+      "loss": 1.3373,
+      "step": 2126
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019727125409444045,
+      "loss": 1.3317,
+      "step": 2127
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019726763845294423,
+      "loss": 1.3545,
+      "step": 2128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019726402045081436,
+      "loss": 1.2512,
+      "step": 2129
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019726040008813866,
+      "loss": 1.4057,
+      "step": 2130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000197256777365005,
+      "loss": 1.5385,
+      "step": 2131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001972531522815013,
+      "loss": 1.2531,
+      "step": 2132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019724952483771556,
+      "loss": 1.2755,
+      "step": 2133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019724589503373576,
+      "loss": 1.4545,
+      "step": 2134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019724226286965007,
+      "loss": 1.4424,
+      "step": 2135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019723862834554657,
+      "loss": 1.432,
+      "step": 2136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019723499146151346,
+      "loss": 1.4091,
+      "step": 2137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019723135221763908,
+      "loss": 1.2578,
+      "step": 2138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019722771061401174,
+      "loss": 1.5706,
+      "step": 2139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019722406665071974,
+      "loss": 1.3045,
+      "step": 2140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019722042032785158,
+      "loss": 1.2688,
+      "step": 2141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019721677164549577,
+      "loss": 1.4009,
+      "step": 2142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019721312060374082,
+      "loss": 1.363,
+      "step": 2143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019720946720267537,
+      "loss": 1.2183,
+      "step": 2144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019720581144238802,
+      "loss": 2.917,
+      "step": 2145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001972021533229676,
+      "loss": 1.3416,
+      "step": 2146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019719849284450283,
+      "loss": 1.3727,
+      "step": 2147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019719483000708255,
+      "loss": 1.1623,
+      "step": 2148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019719116481079562,
+      "loss": 1.4692,
+      "step": 2149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001971874972557311,
+      "loss": 1.4171,
+      "step": 2150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019718382734197786,
+      "loss": 1.1078,
+      "step": 2151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019718015506962506,
+      "loss": 1.3833,
+      "step": 2152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019717648043876182,
+      "loss": 1.3481,
+      "step": 2153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001971728034494773,
+      "loss": 2.7368,
+      "step": 2154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019716912410186073,
+      "loss": 1.4903,
+      "step": 2155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001971654423960014,
+      "loss": 1.3887,
+      "step": 2156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019716175833198868,
+      "loss": 1.3754,
+      "step": 2157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000197158071909912,
+      "loss": 1.1697,
+      "step": 2158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019715438312986082,
+      "loss": 2.4514,
+      "step": 2159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019715069199192464,
+      "loss": 1.4488,
+      "step": 2160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019714699849619305,
+      "loss": 1.2856,
+      "step": 2161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019714330264275568,
+      "loss": 1.383,
+      "step": 2162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019713960443170226,
+      "loss": 1.3564,
+      "step": 2163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001971359038631225,
+      "loss": 1.2305,
+      "step": 2164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019713220093710624,
+      "loss": 1.3511,
+      "step": 2165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019712849565374336,
+      "loss": 1.2071,
+      "step": 2166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019712478801312377,
+      "loss": 1.1759,
+      "step": 2167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019712107801533743,
+      "loss": 1.416,
+      "step": 2168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019711736566047444,
+      "loss": 1.4075,
+      "step": 2169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001971136509486248,
+      "loss": 2.5115,
+      "step": 2170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019710993387987876,
+      "loss": 1.3806,
+      "step": 2171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019710621445432646,
+      "loss": 1.1963,
+      "step": 2172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001971024926720582,
+      "loss": 1.2403,
+      "step": 2173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019709876853316434,
+      "loss": 1.2852,
+      "step": 2174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970950420377352,
+      "loss": 2.5923,
+      "step": 2175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019709131318586124,
+      "loss": 1.2559,
+      "step": 2176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000197087581977633,
+      "loss": 1.2337,
+      "step": 2177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019708384841314098,
+      "loss": 1.2938,
+      "step": 2178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019708011249247581,
+      "loss": 1.3201,
+      "step": 2179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019707637421572817,
+      "loss": 2.5012,
+      "step": 2180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970726335829888,
+      "loss": 1.3138,
+      "step": 2181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019706889059434844,
+      "loss": 1.3154,
+      "step": 2182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019706514524989792,
+      "loss": 1.1316,
+      "step": 2183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970613975497282,
+      "loss": 1.3567,
+      "step": 2184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019705764749393018,
+      "loss": 1.2775,
+      "step": 2185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019705389508259491,
+      "loss": 1.2883,
+      "step": 2186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019705014031581346,
+      "loss": 1.1706,
+      "step": 2187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970463831936769,
+      "loss": 1.2827,
+      "step": 2188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970426237162765,
+      "loss": 1.3704,
+      "step": 2189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970388618837034,
+      "loss": 1.4574,
+      "step": 2190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019703509769604898,
+      "loss": 2.7034,
+      "step": 2191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019703133115340458,
+      "loss": 1.4839,
+      "step": 2192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970275622558616,
+      "loss": 1.2372,
+      "step": 2193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019702379100351144,
+      "loss": 1.2556,
+      "step": 2194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019702001739644578,
+      "loss": 1.35,
+      "step": 2195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019701624143475606,
+      "loss": 1.1733,
+      "step": 2196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000197012463118534,
+      "loss": 1.2009,
+      "step": 2197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019700868244787128,
+      "loss": 1.3602,
+      "step": 2198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019700489942285962,
+      "loss": 1.2043,
+      "step": 2199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001970011140435909,
+      "loss": 1.439,
+      "step": 2200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019699732631015693,
+      "loss": 1.2064,
+      "step": 2201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019699353622264963,
+      "loss": 2.7043,
+      "step": 2202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019698974378116107,
+      "loss": 1.2093,
+      "step": 2203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001969859489857832,
+      "loss": 2.2493,
+      "step": 2204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019698215183660815,
+      "loss": 1.4147,
+      "step": 2205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019697835233372808,
+      "loss": 1.1911,
+      "step": 2206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019697455047723518,
+      "loss": 1.1877,
+      "step": 2207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019697074626722177,
+      "loss": 1.3406,
+      "step": 2208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019696693970378012,
+      "loss": 1.3661,
+      "step": 2209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019696313078700262,
+      "loss": 1.1859,
+      "step": 2210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001969593195169817,
+      "loss": 1.4014,
+      "step": 2211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001969555058938099,
+      "loss": 1.2604,
+      "step": 2212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001969516899175798,
+      "loss": 1.2843,
+      "step": 2213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019694787158838396,
+      "loss": 1.3423,
+      "step": 2214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000196944050906315,
+      "loss": 1.3683,
+      "step": 2215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019694022787146575,
+      "loss": 1.2037,
+      "step": 2216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019693640248392896,
+      "loss": 1.3478,
+      "step": 2217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001969325747437974,
+      "loss": 1.2906,
+      "step": 2218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019692874465116408,
+      "loss": 1.3181,
+      "step": 2219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019692491220612185,
+      "loss": 1.4133,
+      "step": 2220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001969210774087638,
+      "loss": 1.4,
+      "step": 2221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019691724025918295,
+      "loss": 1.2221,
+      "step": 2222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019691340075747243,
+      "loss": 1.2125,
+      "step": 2223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019690955890372546,
+      "loss": 1.3185,
+      "step": 2224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019690571469803528,
+      "loss": 1.3803,
+      "step": 2225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019690186814049512,
+      "loss": 1.2431,
+      "step": 2226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001968980192311984,
+      "loss": 1.4634,
+      "step": 2227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019689416797023847,
+      "loss": 2.6909,
+      "step": 2228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019689031435770886,
+      "loss": 1.2365,
+      "step": 2229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019688645839370308,
+      "loss": 1.2915,
+      "step": 2230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001968826000783147,
+      "loss": 1.3486,
+      "step": 2231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019687873941163735,
+      "loss": 1.2176,
+      "step": 2232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019687487639376473,
+      "loss": 1.3292,
+      "step": 2233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001968710110247906,
+      "loss": 1.3531,
+      "step": 2234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019686714330480879,
+      "loss": 1.4159,
+      "step": 2235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019686327323391317,
+      "loss": 1.3212,
+      "step": 2236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001968594008121976,
+      "loss": 1.2244,
+      "step": 2237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019685552603975612,
+      "loss": 1.2987,
+      "step": 2238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019685164891668276,
+      "loss": 1.3322,
+      "step": 2239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019684776944307162,
+      "loss": 1.2517,
+      "step": 2240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019684388761901682,
+      "loss": 1.2357,
+      "step": 2241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001968400034446126,
+      "loss": 1.4099,
+      "step": 2242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019683611691995322,
+      "loss": 1.3002,
+      "step": 2243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000196832228045133,
+      "loss": 1.2461,
+      "step": 2244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019682833682024633,
+      "loss": 1.3063,
+      "step": 2245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019682444324538764,
+      "loss": 1.3605,
+      "step": 2246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019682054732065142,
+      "loss": 1.288,
+      "step": 2247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019681664904613224,
+      "loss": 1.3779,
+      "step": 2248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001968127484219247,
+      "loss": 1.5381,
+      "step": 2249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019680884544812346,
+      "loss": 1.0481,
+      "step": 2250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019680494012482323,
+      "loss": 1.3881,
+      "step": 2251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019680103245211882,
+      "loss": 1.4547,
+      "step": 2252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019679712243010506,
+      "loss": 2.5371,
+      "step": 2253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019679321005887683,
+      "loss": 1.3026,
+      "step": 2254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967892953385291,
+      "loss": 1.189,
+      "step": 2255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019678537826915685,
+      "loss": 1.2737,
+      "step": 2256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967814588508552,
+      "loss": 1.2211,
+      "step": 2257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967775370837192,
+      "loss": 1.293,
+      "step": 2258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019677361296784407,
+      "loss": 2.4988,
+      "step": 2259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019676968650332504,
+      "loss": 1.4847,
+      "step": 2260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019676575769025744,
+      "loss": 1.357,
+      "step": 2261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967618265287365,
+      "loss": 1.2624,
+      "step": 2262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967578930188578,
+      "loss": 1.2729,
+      "step": 2263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019675395716071666,
+      "loss": 1.1308,
+      "step": 2264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019675001895440867,
+      "loss": 1.3102,
+      "step": 2265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019674607840002942,
+      "loss": 1.2593,
+      "step": 2266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967421354976745,
+      "loss": 1.4514,
+      "step": 2267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019673819024743961,
+      "loss": 1.25,
+      "step": 2268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001967342426494205,
+      "loss": 2.3582,
+      "step": 2269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019673029270371301,
+      "loss": 1.1005,
+      "step": 2270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019672634041041298,
+      "loss": 1.335,
+      "step": 2271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019672238576961632,
+      "loss": 1.4227,
+      "step": 2272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019671842878141903,
+      "loss": 1.4158,
+      "step": 2273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019671446944591713,
+      "loss": 1.2587,
+      "step": 2274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019671050776320673,
+      "loss": 1.2673,
+      "step": 2275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019670654373338392,
+      "loss": 2.3806,
+      "step": 2276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019670257735654497,
+      "loss": 1.3707,
+      "step": 2277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019669860863278613,
+      "loss": 1.431,
+      "step": 2278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019669463756220369,
+      "loss": 1.377,
+      "step": 2279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019669066414489402,
+      "loss": 2.4492,
+      "step": 2280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001966866883809536,
+      "loss": 1.1472,
+      "step": 2281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019668271027047888,
+      "loss": 1.301,
+      "step": 2282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019667872981356642,
+      "loss": 1.4533,
+      "step": 2283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019667474701031283,
+      "loss": 1.3395,
+      "step": 2284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019667076186081478,
+      "loss": 1.4999,
+      "step": 2285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019666677436516893,
+      "loss": 1.2092,
+      "step": 2286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019666278452347213,
+      "loss": 1.4149,
+      "step": 2287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001966587923358212,
+      "loss": 1.3425,
+      "step": 2288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019665479780231295,
+      "loss": 1.2896,
+      "step": 2289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019665080092304442,
+      "loss": 1.2243,
+      "step": 2290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019664680169811255,
+      "loss": 1.2209,
+      "step": 2291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019664280012761444,
+      "loss": 1.4788,
+      "step": 2292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001966387962116472,
+      "loss": 1.367,
+      "step": 2293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019663478995030794,
+      "loss": 1.2875,
+      "step": 2294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000196630781343694,
+      "loss": 1.3807,
+      "step": 2295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019662677039190257,
+      "loss": 1.1984,
+      "step": 2296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019662275709503107,
+      "loss": 1.4072,
+      "step": 2297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019661874145317682,
+      "loss": 1.2731,
+      "step": 2298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019661472346643732,
+      "loss": 1.321,
+      "step": 2299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001966107031349101,
+      "loss": 1.4107,
+      "step": 2300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001966066804586927,
+      "loss": 1.3668,
+      "step": 2301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001966026554378828,
+      "loss": 2.5986,
+      "step": 2302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000196598628072578,
+      "loss": 1.5117,
+      "step": 2303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019659459836287612,
+      "loss": 1.259,
+      "step": 2304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019659056630887497,
+      "loss": 1.4188,
+      "step": 2305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019658653191067228,
+      "loss": 1.4921,
+      "step": 2306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019658249516836608,
+      "loss": 1.1736,
+      "step": 2307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019657845608205435,
+      "loss": 1.2917,
+      "step": 2308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019657441465183504,
+      "loss": 1.261,
+      "step": 2309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019657037087780625,
+      "loss": 2.3442,
+      "step": 2310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019656632476006616,
+      "loss": 1.2753,
+      "step": 2311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019656227629871295,
+      "loss": 1.2296,
+      "step": 2312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019655822549384484,
+      "loss": 1.2623,
+      "step": 2313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019655417234556018,
+      "loss": 1.3798,
+      "step": 2314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019655011685395734,
+      "loss": 1.2796,
+      "step": 2315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001965460590191347,
+      "loss": 1.1545,
+      "step": 2316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019654199884119083,
+      "loss": 1.3947,
+      "step": 2317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019653793632022417,
+      "loss": 1.2505,
+      "step": 2318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019653387145633338,
+      "loss": 1.3147,
+      "step": 2319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019652980424961707,
+      "loss": 1.3852,
+      "step": 2320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019652573470017396,
+      "loss": 1.4519,
+      "step": 2321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019652166280810285,
+      "loss": 1.4507,
+      "step": 2322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019651758857350252,
+      "loss": 1.3444,
+      "step": 2323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019651351199647188,
+      "loss": 2.4641,
+      "step": 2324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019650943307710983,
+      "loss": 1.2661,
+      "step": 2325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019650535181551542,
+      "loss": 1.3615,
+      "step": 2326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019650126821178763,
+      "loss": 1.0881,
+      "step": 2327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019649718226602564,
+      "loss": 1.2239,
+      "step": 2328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019649309397832854,
+      "loss": 1.2476,
+      "step": 2329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019648900334879557,
+      "loss": 1.1989,
+      "step": 2330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019648491037752605,
+      "loss": 1.3827,
+      "step": 2331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001964808150646193,
+      "loss": 1.3419,
+      "step": 2332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001964767174101747,
+      "loss": 1.1886,
+      "step": 2333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019647261741429167,
+      "loss": 1.3413,
+      "step": 2334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019646851507706977,
+      "loss": 1.2222,
+      "step": 2335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019646441039860852,
+      "loss": 1.2959,
+      "step": 2336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019646030337900757,
+      "loss": 1.4576,
+      "step": 2337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019645619401836655,
+      "loss": 1.2897,
+      "step": 2338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019645208231678523,
+      "loss": 1.2498,
+      "step": 2339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019644796827436337,
+      "loss": 1.3636,
+      "step": 2340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019644385189120084,
+      "loss": 1.1542,
+      "step": 2341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019643973316739756,
+      "loss": 1.2609,
+      "step": 2342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019643561210305345,
+      "loss": 1.4397,
+      "step": 2343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019643148869826854,
+      "loss": 1.2404,
+      "step": 2344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019642736295314293,
+      "loss": 1.4906,
+      "step": 2345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019642323486777666,
+      "loss": 1.1165,
+      "step": 2346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019641910444227002,
+      "loss": 1.3617,
+      "step": 2347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019641497167672322,
+      "loss": 1.4578,
+      "step": 2348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019641083657123654,
+      "loss": 1.1461,
+      "step": 2349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019640669912591037,
+      "loss": 1.288,
+      "step": 2350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019640255934084508,
+      "loss": 1.1922,
+      "step": 2351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001963984172161412,
+      "loss": 1.3652,
+      "step": 2352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019639427275189915,
+      "loss": 1.222,
+      "step": 2353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019639012594821966,
+      "loss": 1.3838,
+      "step": 2354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019638597680520325,
+      "loss": 1.3499,
+      "step": 2355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019638182532295066,
+      "loss": 1.4081,
+      "step": 2356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019637767150156268,
+      "loss": 1.2984,
+      "step": 2357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019637351534114008,
+      "loss": 1.3162,
+      "step": 2358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019636935684178372,
+      "loss": 1.1925,
+      "step": 2359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019636519600359456,
+      "loss": 1.3989,
+      "step": 2360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019636103282667352,
+      "loss": 1.3453,
+      "step": 2361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019635686731112173,
+      "loss": 1.4818,
+      "step": 2362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019635269945704023,
+      "loss": 1.4628,
+      "step": 2363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019634852926453015,
+      "loss": 1.3005,
+      "step": 2364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019634435673369273,
+      "loss": 1.1504,
+      "step": 2365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019634018186462924,
+      "loss": 1.2281,
+      "step": 2366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000196336004657441,
+      "loss": 1.4536,
+      "step": 2367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019633182511222938,
+      "loss": 2.5742,
+      "step": 2368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001963276432290958,
+      "loss": 1.2174,
+      "step": 2369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019632345900814175,
+      "loss": 1.2622,
+      "step": 2370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019631927244946883,
+      "loss": 1.1067,
+      "step": 2371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019631508355317864,
+      "loss": 1.3904,
+      "step": 2372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019631089231937274,
+      "loss": 1.2532,
+      "step": 2373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019630669874815297,
+      "loss": 1.55,
+      "step": 2374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019630250283962105,
+      "loss": 1.3932,
+      "step": 2375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019629830459387884,
+      "loss": 1.384,
+      "step": 2376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019629410401102816,
+      "loss": 1.3899,
+      "step": 2377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019628990109117104,
+      "loss": 1.3194,
+      "step": 2378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019628569583440947,
+      "loss": 1.313,
+      "step": 2379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019628148824084546,
+      "loss": 1.2745,
+      "step": 2380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019627727831058117,
+      "loss": 1.3898,
+      "step": 2381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019627306604371873,
+      "loss": 1.371,
+      "step": 2382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019626885144036042,
+      "loss": 1.2703,
+      "step": 2383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001962646345006085,
+      "loss": 1.1922,
+      "step": 2384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019626041522456534,
+      "loss": 1.3167,
+      "step": 2385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019625619361233326,
+      "loss": 1.3141,
+      "step": 2386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019625196966401482,
+      "loss": 1.2903,
+      "step": 2387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019624774337971248,
+      "loss": 1.5103,
+      "step": 2388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019624351475952878,
+      "loss": 1.3611,
+      "step": 2389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001962392838035664,
+      "loss": 1.4178,
+      "step": 2390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000196235050511928,
+      "loss": 1.5929,
+      "step": 2391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019623081488471636,
+      "loss": 1.2751,
+      "step": 2392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001962265769220342,
+      "loss": 1.5563,
+      "step": 2393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001962223366239844,
+      "loss": 1.177,
+      "step": 2394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019621809399066994,
+      "loss": 1.2558,
+      "step": 2395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019621384902219368,
+      "loss": 1.2021,
+      "step": 2396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001962096017186587,
+      "loss": 1.3649,
+      "step": 2397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001962053520801681,
+      "loss": 2.5618,
+      "step": 2398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019620110010682497,
+      "loss": 1.2228,
+      "step": 2399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019619684579873253,
+      "loss": 1.1902,
+      "step": 2400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019619258915599402,
+      "loss": 1.3514,
+      "step": 2401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961883301787127,
+      "loss": 1.1459,
+      "step": 2402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019618406886699204,
+      "loss": 1.2191,
+      "step": 2403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019617980522093541,
+      "loss": 1.2272,
+      "step": 2404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019617553924064623,
+      "loss": 2.6758,
+      "step": 2405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019617127092622812,
+      "loss": 1.3144,
+      "step": 2406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961670002777846,
+      "loss": 1.3997,
+      "step": 2407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019616272729541936,
+      "loss": 1.3604,
+      "step": 2408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961584519792361,
+      "loss": 1.152,
+      "step": 2409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019615417432933857,
+      "loss": 1.4062,
+      "step": 2410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019614989434583056,
+      "loss": 1.1179,
+      "step": 2411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019614561202881597,
+      "loss": 1.1464,
+      "step": 2412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019614132737839874,
+      "loss": 1.4693,
+      "step": 2413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019613704039468286,
+      "loss": 1.2826,
+      "step": 2414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961327510777723,
+      "loss": 2.5913,
+      "step": 2415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961284594277713,
+      "loss": 1.3943,
+      "step": 2416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019612416544478387,
+      "loss": 1.3856,
+      "step": 2417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961198691289143,
+      "loss": 1.2531,
+      "step": 2418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019611557048026686,
+      "loss": 1.2811,
+      "step": 2419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019611126949894584,
+      "loss": 1.238,
+      "step": 2420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019610696618505564,
+      "loss": 1.4783,
+      "step": 2421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019610266053870072,
+      "loss": 1.2946,
+      "step": 2422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019609835255998554,
+      "loss": 1.4962,
+      "step": 2423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019609404224901468,
+      "loss": 1.3088,
+      "step": 2424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019608972960589272,
+      "loss": 1.36,
+      "step": 2425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001960854146307244,
+      "loss": 1.2606,
+      "step": 2426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019608109732361436,
+      "loss": 1.4673,
+      "step": 2427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001960767776846674,
+      "loss": 1.2551,
+      "step": 2428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019607245571398834,
+      "loss": 1.2865,
+      "step": 2429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001960681314116821,
+      "loss": 1.4146,
+      "step": 2430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019606380477785363,
+      "loss": 2.3831,
+      "step": 2431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019605947581260793,
+      "loss": 1.2862,
+      "step": 2432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019605514451605005,
+      "loss": 1.4891,
+      "step": 2433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019605081088828512,
+      "loss": 1.2944,
+      "step": 2434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019604647492941832,
+      "loss": 1.3093,
+      "step": 2435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019604213663955485,
+      "loss": 1.356,
+      "step": 2436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019603779601880004,
+      "loss": 1.1916,
+      "step": 2437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019603345306725922,
+      "loss": 1.2209,
+      "step": 2438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019602910778503775,
+      "loss": 1.267,
+      "step": 2439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019602476017224117,
+      "loss": 1.2302,
+      "step": 2440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019602041022897493,
+      "loss": 1.2657,
+      "step": 2441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001960160579553446,
+      "loss": 1.4055,
+      "step": 2442
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019601170335145583,
+      "loss": 1.4785,
+      "step": 2443
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001960073464174143,
+      "loss": 1.2263,
+      "step": 2444
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019600298715332577,
+      "loss": 1.2875,
+      "step": 2445
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000195998625559296,
+      "loss": 1.4443,
+      "step": 2446
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019599426163543088,
+      "loss": 1.4869,
+      "step": 2447
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019598989538183628,
+      "loss": 1.3224,
+      "step": 2448
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001959855267986182,
+      "loss": 1.2638,
+      "step": 2449
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019598115588588267,
+      "loss": 1.328,
+      "step": 2450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001959767826437357,
+      "loss": 1.3866,
+      "step": 2451
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019597240707228353,
+      "loss": 1.2362,
+      "step": 2452
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001959680291716323,
+      "loss": 1.2834,
+      "step": 2453
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019596364894188822,
+      "loss": 1.4352,
+      "step": 2454
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019595926638315764,
+      "loss": 1.4731,
+      "step": 2455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019595488149554694,
+      "loss": 2.4785,
+      "step": 2456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001959504942791625,
+      "loss": 1.2446,
+      "step": 2457
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001959461047341108,
+      "loss": 1.3796,
+      "step": 2458
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019594171286049837,
+      "loss": 1.3147,
+      "step": 2459
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019593731865843187,
+      "loss": 1.2551,
+      "step": 2460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019593292212801786,
+      "loss": 1.1558,
+      "step": 2461
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019592852326936306,
+      "loss": 2.3259,
+      "step": 2462
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019592412208257424,
+      "loss": 1.2639,
+      "step": 2463
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019591971856775818,
+      "loss": 1.3988,
+      "step": 2464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019591531272502183,
+      "loss": 1.2568,
+      "step": 2465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019591090455447208,
+      "loss": 1.3573,
+      "step": 2466
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019590649405621586,
+      "loss": 1.4579,
+      "step": 2467
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019590208123036024,
+      "loss": 1.2929,
+      "step": 2468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958976660770124,
+      "loss": 1.195,
+      "step": 2469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019589324859627935,
+      "loss": 1.1965,
+      "step": 2470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958888287882684,
+      "loss": 1.2986,
+      "step": 2471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958844066530868,
+      "loss": 1.4559,
+      "step": 2472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019587998219084188,
+      "loss": 1.4476,
+      "step": 2473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019587555540164097,
+      "loss": 1.3589,
+      "step": 2474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019587112628559154,
+      "loss": 1.3504,
+      "step": 2475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019586669484280108,
+      "loss": 1.3112,
+      "step": 2476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019586226107337718,
+      "loss": 1.2952,
+      "step": 2477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019585782497742735,
+      "loss": 1.3832,
+      "step": 2478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019585338655505932,
+      "loss": 1.1869,
+      "step": 2479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958489458063808,
+      "loss": 1.2027,
+      "step": 2480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019584450273149958,
+      "loss": 2.5376,
+      "step": 2481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019584005733052343,
+      "loss": 1.4191,
+      "step": 2482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958356096035603,
+      "loss": 1.2364,
+      "step": 2483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958311595507181,
+      "loss": 1.064,
+      "step": 2484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019582670717210486,
+      "loss": 1.1153,
+      "step": 2485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958222524678286,
+      "loss": 1.3454,
+      "step": 2486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019581779543799746,
+      "loss": 2.502,
+      "step": 2487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958133360827196,
+      "loss": 1.3617,
+      "step": 2488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019580887440210325,
+      "loss": 1.2297,
+      "step": 2489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001958044103962567,
+      "loss": 1.3584,
+      "step": 2490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019579994406528827,
+      "loss": 1.2819,
+      "step": 2491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019579547540930637,
+      "loss": 1.376,
+      "step": 2492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019579100442841942,
+      "loss": 1.2794,
+      "step": 2493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000195786531122736,
+      "loss": 1.4568,
+      "step": 2494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019578205549236459,
+      "loss": 1.1444,
+      "step": 2495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019577757753741387,
+      "loss": 1.3274,
+      "step": 2496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019577309725799248,
+      "loss": 1.419,
+      "step": 2497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001957686146542092,
+      "loss": 1.3378,
+      "step": 2498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019576412972617277,
+      "loss": 1.3282,
+      "step": 2499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019575964247399208,
+      "loss": 1.3168,
+      "step": 2500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000195755152897776,
+      "loss": 1.2529,
+      "step": 2501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001957506609976335,
+      "loss": 1.3562,
+      "step": 2502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001957461667736736,
+      "loss": 1.3641,
+      "step": 2503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019574167022600535,
+      "loss": 0.9711,
+      "step": 2504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019573717135473793,
+      "loss": 1.3865,
+      "step": 2505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019573267015998052,
+      "loss": 1.2855,
+      "step": 2506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019572816664184227,
+      "loss": 1.2939,
+      "step": 2507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001957236608004326,
+      "loss": 1.1519,
+      "step": 2508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019571915263586078,
+      "loss": 1.2841,
+      "step": 2509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019571464214823627,
+      "loss": 1.3775,
+      "step": 2510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001957101293376685,
+      "loss": 1.2879,
+      "step": 2511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019570561420426702,
+      "loss": 1.3201,
+      "step": 2512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019570109674814137,
+      "loss": 1.3324,
+      "step": 2513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019569657696940124,
+      "loss": 1.2178,
+      "step": 2514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019569205486815632,
+      "loss": 2.749,
+      "step": 2515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956875304445163,
+      "loss": 1.3635,
+      "step": 2516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019568300369859104,
+      "loss": 1.411,
+      "step": 2517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956784746304904,
+      "loss": 1.3534,
+      "step": 2518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019567394324032425,
+      "loss": 1.2838,
+      "step": 2519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956694095282026,
+      "loss": 1.3221,
+      "step": 2520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956648734942355,
+      "loss": 1.3458,
+      "step": 2521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000195660335138533,
+      "loss": 1.3421,
+      "step": 2522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019565579446120527,
+      "loss": 1.316,
+      "step": 2523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019565125146236252,
+      "loss": 1.3685,
+      "step": 2524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019564670614211495,
+      "loss": 1.0992,
+      "step": 2525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019564215850057293,
+      "loss": 1.3187,
+      "step": 2526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956376085378468,
+      "loss": 1.4515,
+      "step": 2527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019563305625404698,
+      "loss": 1.3759,
+      "step": 2528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019562850164928398,
+      "loss": 1.4791,
+      "step": 2529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956239447236683,
+      "loss": 1.2376,
+      "step": 2530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019561938547731057,
+      "loss": 1.3502,
+      "step": 2531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001956148239103214,
+      "loss": 1.4001,
+      "step": 2532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019561026002281156,
+      "loss": 1.3926,
+      "step": 2533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019560569381489175,
+      "loss": 1.2526,
+      "step": 2534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019560112528667283,
+      "loss": 1.304,
+      "step": 2535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019559655443826565,
+      "loss": 1.1953,
+      "step": 2536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019559198126978115,
+      "loss": 1.1926,
+      "step": 2537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019558740578133033,
+      "loss": 1.4426,
+      "step": 2538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001955828279730242,
+      "loss": 1.079,
+      "step": 2539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019557824784497392,
+      "loss": 2.4805,
+      "step": 2540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019557366539729062,
+      "loss": 1.4317,
+      "step": 2541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019556908063008546,
+      "loss": 1.3015,
+      "step": 2542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019556449354346978,
+      "loss": 1.4265,
+      "step": 2543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001955599041375549,
+      "loss": 2.4934,
+      "step": 2544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019555531241245214,
+      "loss": 1.3593,
+      "step": 2545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019555071836827303,
+      "loss": 1.324,
+      "step": 2546
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019554612200512904,
+      "loss": 1.2054,
+      "step": 2547
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019554152332313166,
+      "loss": 1.2859,
+      "step": 2548
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019553692232239254,
+      "loss": 1.2476,
+      "step": 2549
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019553231900302332,
+      "loss": 1.2895,
+      "step": 2550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001955277133651358,
+      "loss": 1.4266,
+      "step": 2551
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019552310540884164,
+      "loss": 1.3036,
+      "step": 2552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019551849513425277,
+      "loss": 1.3139,
+      "step": 2553
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019551388254148103,
+      "loss": 2.1743,
+      "step": 2554
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019550926763063837,
+      "loss": 1.4536,
+      "step": 2555
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001955046504018368,
+      "loss": 1.4003,
+      "step": 2556
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019550003085518837,
+      "loss": 1.3746,
+      "step": 2557
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001954954089908052,
+      "loss": 1.4432,
+      "step": 2558
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019549078480879945,
+      "loss": 1.1984,
+      "step": 2559
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019548615830928338,
+      "loss": 1.1257,
+      "step": 2560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019548152949236923,
+      "loss": 1.2216,
+      "step": 2561
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019547689835816936,
+      "loss": 1.3927,
+      "step": 2562
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019547226490679615,
+      "loss": 1.3472,
+      "step": 2563
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001954676291383621,
+      "loss": 1.3816,
+      "step": 2564
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019546299105297962,
+      "loss": 1.4952,
+      "step": 2565
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019545835065076138,
+      "loss": 1.1491,
+      "step": 2566
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019545370793181994,
+      "loss": 1.1562,
+      "step": 2567
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019544906289626802,
+      "loss": 1.4077,
+      "step": 2568
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019544441554421828,
+      "loss": 1.4573,
+      "step": 2569
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019543976587578355,
+      "loss": 1.2211,
+      "step": 2570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001954351138910767,
+      "loss": 1.3684,
+      "step": 2571
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001954304595902106,
+      "loss": 1.3932,
+      "step": 2572
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019542580297329822,
+      "loss": 1.2124,
+      "step": 2573
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019542114404045256,
+      "loss": 1.2897,
+      "step": 2574
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019541648279178673,
+      "loss": 1.1056,
+      "step": 2575
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001954118192274138,
+      "loss": 2.7654,
+      "step": 2576
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000195407153347447,
+      "loss": 1.3382,
+      "step": 2577
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001954024851519995,
+      "loss": 1.2449,
+      "step": 2578
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019539781464118468,
+      "loss": 1.2273,
+      "step": 2579
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019539314181511583,
+      "loss": 1.2648,
+      "step": 2580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019538846667390637,
+      "loss": 1.2612,
+      "step": 2581
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001953837892176698,
+      "loss": 1.1232,
+      "step": 2582
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001953791094465196,
+      "loss": 1.2046,
+      "step": 2583
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019537442736056935,
+      "loss": 1.2088,
+      "step": 2584
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019536974295993268,
+      "loss": 1.2047,
+      "step": 2585
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019536505624472329,
+      "loss": 1.2339,
+      "step": 2586
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019536036721505494,
+      "loss": 1.2913,
+      "step": 2587
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001953556758710414,
+      "loss": 1.3636,
+      "step": 2588
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019535098221279653,
+      "loss": 1.2102,
+      "step": 2589
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019534628624043424,
+      "loss": 1.4659,
+      "step": 2590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019534158795406854,
+      "loss": 1.235,
+      "step": 2591
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001953368873538134,
+      "loss": 1.1607,
+      "step": 2592
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001953321844397829,
+      "loss": 1.371,
+      "step": 2593
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019532747921209126,
+      "loss": 1.3939,
+      "step": 2594
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019532277167085257,
+      "loss": 1.1295,
+      "step": 2595
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019531806181618112,
+      "loss": 1.2766,
+      "step": 2596
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019531334964819122,
+      "loss": 1.3213,
+      "step": 2597
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019530863516699728,
+      "loss": 1.3734,
+      "step": 2598
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019530391837271362,
+      "loss": 1.5527,
+      "step": 2599
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952991992654548,
+      "loss": 1.3939,
+      "step": 2600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952944778453353,
+      "loss": 1.3442,
+      "step": 2601
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952897541124697,
+      "loss": 1.0481,
+      "step": 2602
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952850280669727,
+      "loss": 1.0209,
+      "step": 2603
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019528029970895892,
+      "loss": 1.4948,
+      "step": 2604
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019527556903854316,
+      "loss": 1.3765,
+      "step": 2605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019527083605584022,
+      "loss": 1.3379,
+      "step": 2606
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000195266100760965,
+      "loss": 1.2873,
+      "step": 2607
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952613631540324,
+      "loss": 1.2311,
+      "step": 2608
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019525662323515737,
+      "loss": 2.6736,
+      "step": 2609
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000195251881004455,
+      "loss": 1.2122,
+      "step": 2610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019524713646204033,
+      "loss": 2.4084,
+      "step": 2611
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019524238960802855,
+      "loss": 1.4009,
+      "step": 2612
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019523764044253482,
+      "loss": 1.1674,
+      "step": 2613
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019523288896567445,
+      "loss": 1.4023,
+      "step": 2614
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952281351775627,
+      "loss": 1.1525,
+      "step": 2615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019522337907831497,
+      "loss": 1.2376,
+      "step": 2616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019521862066804673,
+      "loss": 1.2684,
+      "step": 2617
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952138599468734,
+      "loss": 1.4842,
+      "step": 2618
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001952090969149105,
+      "loss": 1.4081,
+      "step": 2619
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019520433157227374,
+      "loss": 1.3339,
+      "step": 2620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019519956391907868,
+      "loss": 1.4656,
+      "step": 2621
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000195194793955441,
+      "loss": 1.3721,
+      "step": 2622
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001951900216814766,
+      "loss": 1.3292,
+      "step": 2623
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019518524709730115,
+      "loss": 1.3918,
+      "step": 2624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001951804702030306,
+      "loss": 1.1356,
+      "step": 2625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019517569099878091,
+      "loss": 1.2736,
+      "step": 2626
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019517090948466801,
+      "loss": 1.3969,
+      "step": 2627
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019516612566080797,
+      "loss": 1.2944,
+      "step": 2628
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001951613395273169,
+      "loss": 1.3107,
+      "step": 2629
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019515655108431093,
+      "loss": 1.325,
+      "step": 2630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019515176033190628,
+      "loss": 1.1174,
+      "step": 2631
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019514696727021923,
+      "loss": 1.3906,
+      "step": 2632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019514217189936612,
+      "loss": 1.2212,
+      "step": 2633
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019513737421946332,
+      "loss": 1.2182,
+      "step": 2634
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019513257423062723,
+      "loss": 0.9931,
+      "step": 2635
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001951277719329744,
+      "loss": 1.4308,
+      "step": 2636
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019512296732662133,
+      "loss": 1.3609,
+      "step": 2637
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019511816041168468,
+      "loss": 1.0994,
+      "step": 2638
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019511335118828108,
+      "loss": 1.2437,
+      "step": 2639
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019510853965652722,
+      "loss": 1.2343,
+      "step": 2640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019510372581653992,
+      "loss": 1.4746,
+      "step": 2641
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019509890966843595,
+      "loss": 1.4254,
+      "step": 2642
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001950940912123323,
+      "loss": 1.2486,
+      "step": 2643
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001950892704483458,
+      "loss": 1.45,
+      "step": 2644
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001950844473765935,
+      "loss": 1.3382,
+      "step": 2645
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019507962199719246,
+      "loss": 1.1989,
+      "step": 2646
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019507479431025976,
+      "loss": 1.4179,
+      "step": 2647
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001950699643159126,
+      "loss": 1.4018,
+      "step": 2648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019506513201426816,
+      "loss": 1.1719,
+      "step": 2649
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019506029740544377,
+      "loss": 1.4113,
+      "step": 2650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019505546048955672,
+      "loss": 1.3224,
+      "step": 2651
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019505062126672442,
+      "loss": 1.3499,
+      "step": 2652
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001950457797370643,
+      "loss": 1.3784,
+      "step": 2653
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019504093590069386,
+      "loss": 1.1205,
+      "step": 2654
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001950360897577307,
+      "loss": 1.1143,
+      "step": 2655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019503124130829238,
+      "loss": 1.2419,
+      "step": 2656
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001950263905524966,
+      "loss": 1.2632,
+      "step": 2657
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019502153749046105,
+      "loss": 1.3422,
+      "step": 2658
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019501668212230355,
+      "loss": 2.9685,
+      "step": 2659
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019501182444814192,
+      "loss": 1.4419,
+      "step": 2660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019500696446809408,
+      "loss": 1.4022,
+      "step": 2661
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001950021021822779,
+      "loss": 1.3105,
+      "step": 2662
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019499723759081153,
+      "loss": 1.3068,
+      "step": 2663
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019499237069381286,
+      "loss": 1.2486,
+      "step": 2664
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019498750149140012,
+      "loss": 1.4446,
+      "step": 2665
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019498262998369147,
+      "loss": 1.3203,
+      "step": 2666
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949777561708051,
+      "loss": 1.547,
+      "step": 2667
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019497288005285932,
+      "loss": 1.2501,
+      "step": 2668
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019496800162997245,
+      "loss": 1.4062,
+      "step": 2669
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019496312090226293,
+      "loss": 1.1915,
+      "step": 2670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019495823786984918,
+      "loss": 1.3735,
+      "step": 2671
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019495335253284972,
+      "loss": 1.2225,
+      "step": 2672
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019494846489138308,
+      "loss": 1.2279,
+      "step": 2673
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019494357494556795,
+      "loss": 1.1998,
+      "step": 2674
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019493868269552294,
+      "loss": 1.4985,
+      "step": 2675
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949337881413668,
+      "loss": 1.4169,
+      "step": 2676
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019492889128321837,
+      "loss": 1.0483,
+      "step": 2677
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949239921211964,
+      "loss": 1.2722,
+      "step": 2678
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019491909065541988,
+      "loss": 1.3651,
+      "step": 2679
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949141868860077,
+      "loss": 1.219,
+      "step": 2680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949092808130789,
+      "loss": 1.3967,
+      "step": 2681
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019490437243675256,
+      "loss": 1.3638,
+      "step": 2682
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948994617571478,
+      "loss": 1.2637,
+      "step": 2683
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019489454877438374,
+      "loss": 1.1901,
+      "step": 2684
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948896334885797,
+      "loss": 1.2479,
+      "step": 2685
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019488471589985493,
+      "loss": 1.4154,
+      "step": 2686
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948797960083288,
+      "loss": 1.4274,
+      "step": 2687
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019487487381412066,
+      "loss": 1.2819,
+      "step": 2688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019486994931735004,
+      "loss": 1.2719,
+      "step": 2689
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948650225181364,
+      "loss": 1.4221,
+      "step": 2690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019486009341659934,
+      "loss": 1.3903,
+      "step": 2691
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948551620128585,
+      "loss": 1.2909,
+      "step": 2692
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948502283070335,
+      "loss": 1.462,
+      "step": 2693
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019484529229924416,
+      "loss": 1.4175,
+      "step": 2694
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948403539896102,
+      "loss": 1.2404,
+      "step": 2695
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948354133782515,
+      "loss": 1.3066,
+      "step": 2696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000194830470465288,
+      "loss": 1.0938,
+      "step": 2697
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948255252508396,
+      "loss": 1.3808,
+      "step": 2698
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019482057773502639,
+      "loss": 1.1803,
+      "step": 2699
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019481562791796834,
+      "loss": 1.4851,
+      "step": 2700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019481067579978572,
+      "loss": 1.3469,
+      "step": 2701
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019480572138059858,
+      "loss": 1.4301,
+      "step": 2702
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019480076466052722,
+      "loss": 1.2864,
+      "step": 2703
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019479580563969194,
+      "loss": 1.1864,
+      "step": 2704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001947908443182131,
+      "loss": 1.2567,
+      "step": 2705
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001947858806962111,
+      "loss": 1.347,
+      "step": 2706
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001947809147738064,
+      "loss": 1.3914,
+      "step": 2707
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001947759465511195,
+      "loss": 1.3757,
+      "step": 2708
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019477097602827104,
+      "loss": 1.3896,
+      "step": 2709
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019476600320538157,
+      "loss": 1.2631,
+      "step": 2710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019476102808257183,
+      "loss": 1.4652,
+      "step": 2711
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019475605065996256,
+      "loss": 1.329,
+      "step": 2712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019475107093767452,
+      "loss": 1.5995,
+      "step": 2713
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019474608891582862,
+      "loss": 1.2951,
+      "step": 2714
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019474110459454576,
+      "loss": 1.1456,
+      "step": 2715
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019473611797394687,
+      "loss": 1.2767,
+      "step": 2716
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019473112905415301,
+      "loss": 1.2433,
+      "step": 2717
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019472613783528522,
+      "loss": 1.4239,
+      "step": 2718
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019472114431746466,
+      "loss": 1.2344,
+      "step": 2719
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019471614850081256,
+      "loss": 1.3995,
+      "step": 2720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001947111503854501,
+      "loss": 1.4028,
+      "step": 2721
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019470614997149857,
+      "loss": 1.4169,
+      "step": 2722
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001947011472590794,
+      "loss": 1.3215,
+      "step": 2723
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019469614224831397,
+      "loss": 1.3196,
+      "step": 2724
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019469113493932375,
+      "loss": 1.3535,
+      "step": 2725
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019468612533223023,
+      "loss": 1.3232,
+      "step": 2726
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019468111342715503,
+      "loss": 1.4887,
+      "step": 2727
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001946760992242198,
+      "loss": 1.4518,
+      "step": 2728
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001946710827235462,
+      "loss": 1.3019,
+      "step": 2729
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019466606392525595,
+      "loss": 1.2728,
+      "step": 2730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019466104282947093,
+      "loss": 1.3184,
+      "step": 2731
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019465601943631297,
+      "loss": 1.4347,
+      "step": 2732
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001946509937459039,
+      "loss": 1.3411,
+      "step": 2733
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019464596575836587,
+      "loss": 1.3999,
+      "step": 2734
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019464093547382074,
+      "loss": 1.2648,
+      "step": 2735
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019463590289239068,
+      "loss": 1.1983,
+      "step": 2736
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019463086801419776,
+      "loss": 1.3077,
+      "step": 2737
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019462583083936424,
+      "loss": 1.3061,
+      "step": 2738
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019462079136801236,
+      "loss": 1.3415,
+      "step": 2739
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019461574960026438,
+      "loss": 1.3226,
+      "step": 2740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019461070553624272,
+      "loss": 1.4459,
+      "step": 2741
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019460565917606978,
+      "loss": 1.3817,
+      "step": 2742
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019460061051986796,
+      "loss": 1.2891,
+      "step": 2743
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019459555956775993,
+      "loss": 1.3752,
+      "step": 2744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019459050631986813,
+      "loss": 3.0249,
+      "step": 2745
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001945854507763153,
+      "loss": 1.3077,
+      "step": 2746
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019458039293722407,
+      "loss": 1.3861,
+      "step": 2747
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019457533280271724,
+      "loss": 1.3828,
+      "step": 2748
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001945702703729176,
+      "loss": 1.326,
+      "step": 2749
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019456520564794798,
+      "loss": 1.3647,
+      "step": 2750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019456013862793135,
+      "loss": 1.3723,
+      "step": 2751
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001945550693129907,
+      "loss": 1.3617,
+      "step": 2752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019454999770324895,
+      "loss": 1.3716,
+      "step": 2753
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019454492379882926,
+      "loss": 1.2213,
+      "step": 2754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001945398475998548,
+      "loss": 2.8047,
+      "step": 2755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019453476910644873,
+      "loss": 2.6938,
+      "step": 2756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001945296883187343,
+      "loss": 1.1132,
+      "step": 2757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001945246052368348,
+      "loss": 1.3061,
+      "step": 2758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019451951986087363,
+      "loss": 1.316,
+      "step": 2759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001945144321909742,
+      "loss": 1.3086,
+      "step": 2760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019450934222726,
+      "loss": 1.2506,
+      "step": 2761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019450424996985451,
+      "loss": 1.2447,
+      "step": 2762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019449915541888137,
+      "loss": 1.2835,
+      "step": 2763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001944940585744642,
+      "loss": 1.5424,
+      "step": 2764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001944889594367267,
+      "loss": 2.717,
+      "step": 2765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001944838580057926,
+      "loss": 1.3024,
+      "step": 2766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019447875428178576,
+      "loss": 1.4038,
+      "step": 2767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019447364826483,
+      "loss": 1.3238,
+      "step": 2768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019446853995504923,
+      "loss": 1.5327,
+      "step": 2769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019446342935256747,
+      "loss": 2.4456,
+      "step": 2770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019445831645750877,
+      "loss": 1.2037,
+      "step": 2771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019445320126999712,
+      "loss": 1.2792,
+      "step": 2772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019444808379015678,
+      "loss": 1.3405,
+      "step": 2773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019444296401811183,
+      "loss": 1.5231,
+      "step": 2774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019443784195398664,
+      "loss": 1.2548,
+      "step": 2775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019443271759790544,
+      "loss": 1.2123,
+      "step": 2776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019442759094999264,
+      "loss": 1.292,
+      "step": 2777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019442246201037262,
+      "loss": 1.3259,
+      "step": 2778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019441733077916988,
+      "loss": 1.0843,
+      "step": 2779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019441219725650893,
+      "loss": 1.2998,
+      "step": 2780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019440706144251442,
+      "loss": 1.5153,
+      "step": 2781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019440192333731093,
+      "loss": 2.627,
+      "step": 2782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019439678294102319,
+      "loss": 1.3379,
+      "step": 2783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019439164025377597,
+      "loss": 1.1515,
+      "step": 2784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019438649527569402,
+      "loss": 1.2799,
+      "step": 2785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019438134800690225,
+      "loss": 1.231,
+      "step": 2786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001943761984475256,
+      "loss": 1.4067,
+      "step": 2787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000194371046597689,
+      "loss": 1.2659,
+      "step": 2788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019436589245751747,
+      "loss": 1.374,
+      "step": 2789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019436073602713616,
+      "loss": 1.3206,
+      "step": 2790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019435557730667017,
+      "loss": 1.2702,
+      "step": 2791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019435041629624474,
+      "loss": 1.434,
+      "step": 2792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001943452529959851,
+      "loss": 1.3132,
+      "step": 2793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019434008740601655,
+      "loss": 1.298,
+      "step": 2794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019433491952646447,
+      "loss": 1.3118,
+      "step": 2795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019432974935745424,
+      "loss": 2.5823,
+      "step": 2796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019432457689911144,
+      "loss": 1.2078,
+      "step": 2797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001943194021515615,
+      "loss": 1.2593,
+      "step": 2798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019431422511493004,
+      "loss": 1.3957,
+      "step": 2799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019430904578934273,
+      "loss": 1.3396,
+      "step": 2800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019430386417492525,
+      "loss": 1.3137,
+      "step": 2801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019429868027180333,
+      "loss": 1.2792,
+      "step": 2802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019429349408010284,
+      "loss": 1.3079,
+      "step": 2803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001942883055999496,
+      "loss": 2.8792,
+      "step": 2804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019428311483146954,
+      "loss": 1.2794,
+      "step": 2805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019427792177478863,
+      "loss": 1.36,
+      "step": 2806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019427272643003288,
+      "loss": 1.1813,
+      "step": 2807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001942675287973285,
+      "loss": 1.3068,
+      "step": 2808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019426232887680147,
+      "loss": 1.37,
+      "step": 2809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019425712666857809,
+      "loss": 1.2812,
+      "step": 2810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019425192217278458,
+      "loss": 1.1065,
+      "step": 2811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019424671538954727,
+      "loss": 1.369,
+      "step": 2812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001942415063189925,
+      "loss": 1.1882,
+      "step": 2813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019423629496124673,
+      "loss": 1.4106,
+      "step": 2814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019423108131643642,
+      "loss": 1.3103,
+      "step": 2815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019422586538468805,
+      "loss": 1.2426,
+      "step": 2816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001942206471661283,
+      "loss": 1.4198,
+      "step": 2817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019421542666088376,
+      "loss": 1.1226,
+      "step": 2818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019421020386908112,
+      "loss": 1.502,
+      "step": 2819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019420497879084713,
+      "loss": 1.3792,
+      "step": 2820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019419975142630867,
+      "loss": 1.3985,
+      "step": 2821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019419452177559253,
+      "loss": 1.2443,
+      "step": 2822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019418928983882565,
+      "loss": 1.4071,
+      "step": 2823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019418405561613498,
+      "loss": 1.3599,
+      "step": 2824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019417881910764766,
+      "loss": 1.3322,
+      "step": 2825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019417358031349066,
+      "loss": 1.1931,
+      "step": 2826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019416833923379117,
+      "loss": 1.3652,
+      "step": 2827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019416309586867638,
+      "loss": 2.7463,
+      "step": 2828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019415785021827353,
+      "loss": 1.3873,
+      "step": 2829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019415260228270999,
+      "loss": 1.4028,
+      "step": 2830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019414735206211306,
+      "loss": 1.1996,
+      "step": 2831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019414209955661015,
+      "loss": 1.3122,
+      "step": 2832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001941368447663288,
+      "loss": 1.3469,
+      "step": 2833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001941315876913965,
+      "loss": 2.5562,
+      "step": 2834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019412632833194082,
+      "loss": 1.2379,
+      "step": 2835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019412106668808944,
+      "loss": 1.1613,
+      "step": 2836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019411580275997006,
+      "loss": 1.2786,
+      "step": 2837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001941105365477104,
+      "loss": 1.2858,
+      "step": 2838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019410526805143828,
+      "loss": 1.2721,
+      "step": 2839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019409999727128155,
+      "loss": 1.4202,
+      "step": 2840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019409472420736816,
+      "loss": 1.1422,
+      "step": 2841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019408944885982606,
+      "loss": 1.4341,
+      "step": 2842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019408417122878332,
+      "loss": 1.2724,
+      "step": 2843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000194078891314368,
+      "loss": 1.2595,
+      "step": 2844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001940736091167082,
+      "loss": 1.393,
+      "step": 2845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019406832463593216,
+      "loss": 1.2343,
+      "step": 2846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019406303787216812,
+      "loss": 1.3124,
+      "step": 2847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001940577488255444,
+      "loss": 1.3317,
+      "step": 2848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019405245749618937,
+      "loss": 1.4493,
+      "step": 2849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019404716388423142,
+      "loss": 1.2329,
+      "step": 2850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000194041867989799,
+      "loss": 1.172,
+      "step": 2851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019403656981302068,
+      "loss": 2.5815,
+      "step": 2852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019403126935402506,
+      "loss": 1.5428,
+      "step": 2853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001940259666129408,
+      "loss": 1.321,
+      "step": 2854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019402066158989646,
+      "loss": 1.279,
+      "step": 2855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019401535428502094,
+      "loss": 1.2709,
+      "step": 2856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019401004469844297,
+      "loss": 1.3996,
+      "step": 2857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001940047328302914,
+      "loss": 1.438,
+      "step": 2858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001939994186806952,
+      "loss": 1.2256,
+      "step": 2859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001939941022497833,
+      "loss": 1.2535,
+      "step": 2860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019398878353768474,
+      "loss": 1.2882,
+      "step": 2861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019398346254452864,
+      "loss": 1.2783,
+      "step": 2862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019397813927044406,
+      "loss": 1.2568,
+      "step": 2863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019397281371556026,
+      "loss": 2.626,
+      "step": 2864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019396748588000647,
+      "loss": 2.6433,
+      "step": 2865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019396215576391194,
+      "loss": 1.1419,
+      "step": 2866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001939568233674061,
+      "loss": 1.3892,
+      "step": 2867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019395148869061834,
+      "loss": 1.3394,
+      "step": 2868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019394615173367813,
+      "loss": 1.5326,
+      "step": 2869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000193940812496715,
+      "loss": 1.3674,
+      "step": 2870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019393547097985853,
+      "loss": 1.2303,
+      "step": 2871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019393012718323834,
+      "loss": 1.3493,
+      "step": 2872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001939247811069841,
+      "loss": 1.3127,
+      "step": 2873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019391943275122567,
+      "loss": 1.2943,
+      "step": 2874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001939140821160927,
+      "loss": 1.3349,
+      "step": 2875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001939087292017151,
+      "loss": 1.2329,
+      "step": 2876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019390337400822287,
+      "loss": 2.718,
+      "step": 2877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019389801653574588,
+      "loss": 1.1615,
+      "step": 2878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019389265678441416,
+      "loss": 1.462,
+      "step": 2879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019388729475435784,
+      "loss": 1.5823,
+      "step": 2880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000193881930445707,
+      "loss": 1.2173,
+      "step": 2881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019387656385859184,
+      "loss": 1.1921,
+      "step": 2882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019387119499314265,
+      "loss": 1.371,
+      "step": 2883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019386582384948965,
+      "loss": 1.3013,
+      "step": 2884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019386045042776327,
+      "loss": 1.1969,
+      "step": 2885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019385507472809388,
+      "loss": 1.3335,
+      "step": 2886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019384969675061197,
+      "loss": 1.1823,
+      "step": 2887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019384431649544802,
+      "loss": 1.3303,
+      "step": 2888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019383893396273267,
+      "loss": 1.2969,
+      "step": 2889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001938335491525965,
+      "loss": 1.3253,
+      "step": 2890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019382816206517015,
+      "loss": 1.3633,
+      "step": 2891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001938227727005845,
+      "loss": 1.3267,
+      "step": 2892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019381738105897023,
+      "loss": 1.1577,
+      "step": 2893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019381198714045822,
+      "loss": 2.7327,
+      "step": 2894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001938065909451794,
+      "loss": 1.5077,
+      "step": 2895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019380119247326474,
+      "loss": 1.2432,
+      "step": 2896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019379579172484524,
+      "loss": 2.8447,
+      "step": 2897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019379038870005194,
+      "loss": 1.3674,
+      "step": 2898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019378498339901601,
+      "loss": 1.1952,
+      "step": 2899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001937795758218686,
+      "loss": 1.205,
+      "step": 2900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000193774165968741,
+      "loss": 1.1997,
+      "step": 2901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001937687538397645,
+      "loss": 1.427,
+      "step": 2902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019376333943507042,
+      "loss": 1.2893,
+      "step": 2903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019375792275479013,
+      "loss": 1.3407,
+      "step": 2904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019375250379905515,
+      "loss": 1.1943,
+      "step": 2905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000193747082567997,
+      "loss": 1.3254,
+      "step": 2906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019374165906174722,
+      "loss": 1.1428,
+      "step": 2907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001937362332804374,
+      "loss": 1.2136,
+      "step": 2908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001937308052241993,
+      "loss": 1.1443,
+      "step": 2909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019372537489316461,
+      "loss": 1.4359,
+      "step": 2910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019371994228746514,
+      "loss": 1.2753,
+      "step": 2911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001937145074072327,
+      "loss": 1.5148,
+      "step": 2912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019370907025259925,
+      "loss": 1.3585,
+      "step": 2913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001937036308236967,
+      "loss": 1.3705,
+      "step": 2914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936981891206571,
+      "loss": 1.1638,
+      "step": 2915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019369274514361248,
+      "loss": 1.3773,
+      "step": 2916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000193687298892695,
+      "loss": 1.4521,
+      "step": 2917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019368185036803677,
+      "loss": 1.1979,
+      "step": 2918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936763995697701,
+      "loss": 0.9865,
+      "step": 2919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936709464980272,
+      "loss": 1.0826,
+      "step": 2920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936654911529405,
+      "loss": 2.7219,
+      "step": 2921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019366003353464236,
+      "loss": 1.109,
+      "step": 2922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936545736432652,
+      "loss": 1.3311,
+      "step": 2923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936491114789416,
+      "loss": 1.2457,
+      "step": 2924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019364364704180408,
+      "loss": 1.3413,
+      "step": 2925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936381803319852,
+      "loss": 1.0308,
+      "step": 2926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019363271134961772,
+      "loss": 1.0723,
+      "step": 2927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936272400948344,
+      "loss": 1.2727,
+      "step": 2928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019362176656776794,
+      "loss": 1.1341,
+      "step": 2929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001936162907685512,
+      "loss": 1.2463,
+      "step": 2930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019361081269731706,
+      "loss": 2.9092,
+      "step": 2931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019360533235419854,
+      "loss": 1.4161,
+      "step": 2932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001935998497393286,
+      "loss": 1.356,
+      "step": 2933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019359436485284027,
+      "loss": 1.4006,
+      "step": 2934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019358887769486667,
+      "loss": 1.3442,
+      "step": 2935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019358338826554102,
+      "loss": 1.25,
+      "step": 2936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019357789656499654,
+      "loss": 1.3318,
+      "step": 2937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019357240259336648,
+      "loss": 1.3482,
+      "step": 2938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001935669063507842,
+      "loss": 1.3267,
+      "step": 2939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019356140783738303,
+      "loss": 1.4496,
+      "step": 2940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001935559070532965,
+      "loss": 1.1781,
+      "step": 2941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019355040399865806,
+      "loss": 1.1885,
+      "step": 2942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001935448986736013,
+      "loss": 1.415,
+      "step": 2943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019353939107825981,
+      "loss": 1.3293,
+      "step": 2944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019353388121276724,
+      "loss": 1.424,
+      "step": 2945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019352836907725736,
+      "loss": 1.071,
+      "step": 2946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001935228546718639,
+      "loss": 1.3443,
+      "step": 2947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019351733799672068,
+      "loss": 1.1903,
+      "step": 2948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019351181905196164,
+      "loss": 2.8906,
+      "step": 2949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001935062978377207,
+      "loss": 1.2411,
+      "step": 2950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019350077435413186,
+      "loss": 1.3925,
+      "step": 2951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019349524860132916,
+      "loss": 1.0601,
+      "step": 2952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019348972057944667,
+      "loss": 1.2748,
+      "step": 2953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019348419028861864,
+      "loss": 1.3116,
+      "step": 2954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019347865772897924,
+      "loss": 1.3104,
+      "step": 2955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019347312290066273,
+      "loss": 1.2953,
+      "step": 2956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019346758580380348,
+      "loss": 1.1143,
+      "step": 2957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001934620464385358,
+      "loss": 1.29,
+      "step": 2958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001934565048049942,
+      "loss": 1.2785,
+      "step": 2959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019345096090331315,
+      "loss": 1.2476,
+      "step": 2960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019344541473362718,
+      "loss": 1.3101,
+      "step": 2961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001934398662960709,
+      "loss": 1.1581,
+      "step": 2962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000193434315590779,
+      "loss": 1.2681,
+      "step": 2963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019342876261788612,
+      "loss": 1.375,
+      "step": 2964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001934232073775271,
+      "loss": 1.2673,
+      "step": 2965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019341764986983672,
+      "loss": 1.5259,
+      "step": 2966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019341209009494988,
+      "loss": 1.3649,
+      "step": 2967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019340652805300152,
+      "loss": 1.2762,
+      "step": 2968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001934009637441266,
+      "loss": 1.3722,
+      "step": 2969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019339539716846016,
+      "loss": 1.3025,
+      "step": 2970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001933898283261373,
+      "loss": 1.3426,
+      "step": 2971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019338425721729326,
+      "loss": 1.1761,
+      "step": 2972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019337868384206313,
+      "loss": 1.0924,
+      "step": 2973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019337310820058222,
+      "loss": 1.3459,
+      "step": 2974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019336753029298583,
+      "loss": 1.2212,
+      "step": 2975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019336195011940936,
+      "loss": 1.4,
+      "step": 2976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001933563676799882,
+      "loss": 1.3528,
+      "step": 2977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001933507829748579,
+      "loss": 2.5527,
+      "step": 2978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019334519600415395,
+      "loss": 1.1547,
+      "step": 2979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019333960676801192,
+      "loss": 1.3844,
+      "step": 2980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019333401526656753,
+      "loss": 1.3628,
+      "step": 2981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001933284214999564,
+      "loss": 1.2819,
+      "step": 2982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019332282546831434,
+      "loss": 1.496,
+      "step": 2983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019331722717177714,
+      "loss": 1.2398,
+      "step": 2984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001933116266104807,
+      "loss": 1.2057,
+      "step": 2985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001933060237845609,
+      "loss": 1.2693,
+      "step": 2986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019330041869415374,
+      "loss": 1.4993,
+      "step": 2987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019329481133939527,
+      "loss": 1.197,
+      "step": 2988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019328920172042153,
+      "loss": 1.4243,
+      "step": 2989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019328358983736874,
+      "loss": 1.3138,
+      "step": 2990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000193277975690373,
+      "loss": 1.3104,
+      "step": 2991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019327235927957061,
+      "loss": 1.4464,
+      "step": 2992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001932667406050979,
+      "loss": 1.1654,
+      "step": 2993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019326111966709122,
+      "loss": 1.5465,
+      "step": 2994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019325549646568695,
+      "loss": 1.3782,
+      "step": 2995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001932498710010216,
+      "loss": 1.2432,
+      "step": 2996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001932442432732317,
+      "loss": 2.5208,
+      "step": 2997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019323861328245377,
+      "loss": 1.3595,
+      "step": 2998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019323298102882452,
+      "loss": 1.3511,
+      "step": 2999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019322734651248063,
+      "loss": 1.2146,
+      "step": 3000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019322170973355883,
+      "loss": 1.3279,
+      "step": 3001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019321607069219593,
+      "loss": 1.2614,
+      "step": 3002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019321042938852875,
+      "loss": 1.1144,
+      "step": 3003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019320478582269429,
+      "loss": 1.2816,
+      "step": 3004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019319913999482941,
+      "loss": 1.1809,
+      "step": 3005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019319349190507118,
+      "loss": 1.2289,
+      "step": 3006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001931878415535567,
+      "loss": 1.2875,
+      "step": 3007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019318218894042306,
+      "loss": 1.2397,
+      "step": 3008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019317653406580746,
+      "loss": 1.4017,
+      "step": 3009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019317087692984717,
+      "loss": 1.3708,
+      "step": 3010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019316521753267948,
+      "loss": 1.2312,
+      "step": 3011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019315955587444168,
+      "loss": 2.7432,
+      "step": 3012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001931538919552712,
+      "loss": 1.1947,
+      "step": 3013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019314822577530554,
+      "loss": 1.2841,
+      "step": 3014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019314255733468222,
+      "loss": 1.4461,
+      "step": 3015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019313688663353873,
+      "loss": 1.3332,
+      "step": 3016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019313121367201276,
+      "loss": 1.3956,
+      "step": 3017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019312553845024199,
+      "loss": 1.3099,
+      "step": 3018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019311986096836413,
+      "loss": 1.3211,
+      "step": 3019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019311418122651698,
+      "loss": 1.2465,
+      "step": 3020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001931084992248384,
+      "loss": 1.4066,
+      "step": 3021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019310281496346624,
+      "loss": 1.4114,
+      "step": 3022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019309712844253854,
+      "loss": 1.3944,
+      "step": 3023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001930914396621932,
+      "loss": 1.1797,
+      "step": 3024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019308574862256838,
+      "loss": 1.2994,
+      "step": 3025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019308005532380214,
+      "loss": 1.386,
+      "step": 3026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019307435976603267,
+      "loss": 1.3459,
+      "step": 3027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019306866194939822,
+      "loss": 1.3039,
+      "step": 3028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019306296187403702,
+      "loss": 1.3429,
+      "step": 3029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019305725954008745,
+      "loss": 1.4727,
+      "step": 3030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019305155494768795,
+      "loss": 1.2373,
+      "step": 3031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019304584809697682,
+      "loss": 1.2511,
+      "step": 3032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019304013898809272,
+      "loss": 1.242,
+      "step": 3033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019303442762117412,
+      "loss": 1.0951,
+      "step": 3034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019302871399635966,
+      "loss": 1.1611,
+      "step": 3035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019302299811378796,
+      "loss": 1.055,
+      "step": 3036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019301727997359783,
+      "loss": 1.4735,
+      "step": 3037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019301155957592795,
+      "loss": 1.1312,
+      "step": 3038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019300583692091723,
+      "loss": 1.5543,
+      "step": 3039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001930001120087045,
+      "loss": 2.4978,
+      "step": 3040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019299438483942876,
+      "loss": 1.2729,
+      "step": 3041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019298865541322893,
+      "loss": 1.3171,
+      "step": 3042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019298292373024411,
+      "loss": 1.1776,
+      "step": 3043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019297718979061342,
+      "loss": 1.2242,
+      "step": 3044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019297145359447595,
+      "loss": 1.2286,
+      "step": 3045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000192965715141971,
+      "loss": 1.2406,
+      "step": 3046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019295997443323777,
+      "loss": 1.3036,
+      "step": 3047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001929542314684156,
+      "loss": 1.2491,
+      "step": 3048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019294848624764387,
+      "loss": 1.256,
+      "step": 3049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019294273877106205,
+      "loss": 1.4358,
+      "step": 3050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019293698903880962,
+      "loss": 1.2783,
+      "step": 3051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019293123705102608,
+      "loss": 2.5818,
+      "step": 3052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019292548280785106,
+      "loss": 1.446,
+      "step": 3053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001929197263094242,
+      "loss": 1.2718,
+      "step": 3054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019291396755588523,
+      "loss": 1.1883,
+      "step": 3055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019290820654737387,
+      "loss": 1.197,
+      "step": 3056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019290244328402993,
+      "loss": 1.1442,
+      "step": 3057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019289667776599336,
+      "loss": 1.3153,
+      "step": 3058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000192890909993404,
+      "loss": 1.0654,
+      "step": 3059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019288513996640192,
+      "loss": 2.5566,
+      "step": 3060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019287936768512708,
+      "loss": 1.3781,
+      "step": 3061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001928735931497196,
+      "loss": 1.0811,
+      "step": 3062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019286781636031956,
+      "loss": 1.2981,
+      "step": 3063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019286203731706727,
+      "loss": 1.1816,
+      "step": 3064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019285625602010294,
+      "loss": 1.3631,
+      "step": 3065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019285047246956686,
+      "loss": 1.2135,
+      "step": 3066
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001928446866655994,
+      "loss": 1.1756,
+      "step": 3067
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019283889860834098,
+      "loss": 1.4167,
+      "step": 3068
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001928331082979321,
+      "loss": 1.2128,
+      "step": 3069
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019282731573451323,
+      "loss": 2.54,
+      "step": 3070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000192821520918225,
+      "loss": 1.3203,
+      "step": 3071
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019281572384920805,
+      "loss": 1.2759,
+      "step": 3072
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019280992452760302,
+      "loss": 2.644,
+      "step": 3073
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019280412295355072,
+      "loss": 1.0391,
+      "step": 3074
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001927983191271919,
+      "loss": 1.4257,
+      "step": 3075
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019279251304866747,
+      "loss": 1.4281,
+      "step": 3076
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001927867047181183,
+      "loss": 1.2847,
+      "step": 3077
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019278089413568537,
+      "loss": 1.3128,
+      "step": 3078
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019277508130150967,
+      "loss": 1.4634,
+      "step": 3079
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019276926621573232,
+      "loss": 1.2445,
+      "step": 3080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001927634488784944,
+      "loss": 1.4386,
+      "step": 3081
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019275762928993716,
+      "loss": 1.292,
+      "step": 3082
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019275180745020176,
+      "loss": 1.4995,
+      "step": 3083
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019274598335942958,
+      "loss": 1.2125,
+      "step": 3084
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019274015701776186,
+      "loss": 1.3272,
+      "step": 3085
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001927343284253401,
+      "loss": 1.5164,
+      "step": 3086
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019272849758230574,
+      "loss": 2.4646,
+      "step": 3087
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019272266448880024,
+      "loss": 1.3984,
+      "step": 3088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019271682914496517,
+      "loss": 1.3059,
+      "step": 3089
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019271099155094223,
+      "loss": 1.2622,
+      "step": 3090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019270515170687302,
+      "loss": 1.3481,
+      "step": 3091
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926993096128993,
+      "loss": 1.2886,
+      "step": 3092
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019269346526916282,
+      "loss": 1.3636,
+      "step": 3093
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019268761867580544,
+      "loss": 1.2301,
+      "step": 3094
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926817698329691,
+      "loss": 1.3723,
+      "step": 3095
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926759187407957,
+      "loss": 1.3614,
+      "step": 3096
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019267006539942726,
+      "loss": 1.298,
+      "step": 3097
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019266420980900578,
+      "loss": 1.1834,
+      "step": 3098
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019265835196967344,
+      "loss": 1.352,
+      "step": 3099
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019265249188157242,
+      "loss": 1.3913,
+      "step": 3100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926466295448449,
+      "loss": 1.2202,
+      "step": 3101
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019264076495963313,
+      "loss": 2.5908,
+      "step": 3102
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926348981260795,
+      "loss": 1.1077,
+      "step": 3103
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019262902904432637,
+      "loss": 1.1396,
+      "step": 3104
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019262315771451615,
+      "loss": 1.2513,
+      "step": 3105
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926172841367914,
+      "loss": 1.2065,
+      "step": 3106
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926114083112946,
+      "loss": 1.3438,
+      "step": 3107
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001926055302381684,
+      "loss": 1.1984,
+      "step": 3108
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019259964991755546,
+      "loss": 1.0685,
+      "step": 3109
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019259376734959846,
+      "loss": 1.1802,
+      "step": 3110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019258788253444017,
+      "loss": 1.2039,
+      "step": 3111
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019258199547222344,
+      "loss": 1.3559,
+      "step": 3112
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019257610616309112,
+      "loss": 1.3184,
+      "step": 3113
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019257021460718617,
+      "loss": 1.199,
+      "step": 3114
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019256432080465153,
+      "loss": 1.2372,
+      "step": 3115
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019255842475563023,
+      "loss": 1.2614,
+      "step": 3116
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019255252646026546,
+      "loss": 1.0338,
+      "step": 3117
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019254662591870026,
+      "loss": 1.3512,
+      "step": 3118
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019254072313107788,
+      "loss": 1.3788,
+      "step": 3119
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019253481809754163,
+      "loss": 1.1539,
+      "step": 3120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001925289108182347,
+      "loss": 1.3257,
+      "step": 3121
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019252300129330056,
+      "loss": 1.3176,
+      "step": 3122
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019251708952288258,
+      "loss": 1.3433,
+      "step": 3123
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019251117550712427,
+      "loss": 1.3204,
+      "step": 3124
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019250525924616914,
+      "loss": 1.2037,
+      "step": 3125
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019249934074016075,
+      "loss": 1.2193,
+      "step": 3126
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924934199892428,
+      "loss": 1.3596,
+      "step": 3127
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924874969935589,
+      "loss": 1.3708,
+      "step": 3128
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019248157175325289,
+      "loss": 1.2339,
+      "step": 3129
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019247564426846855,
+      "loss": 1.4031,
+      "step": 3130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019246971453934966,
+      "loss": 1.2198,
+      "step": 3131
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924637825660402,
+      "loss": 1.2217,
+      "step": 3132
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019245784834868413,
+      "loss": 1.4128,
+      "step": 3133
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019245191188742546,
+      "loss": 2.6812,
+      "step": 3134
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924459731824083,
+      "loss": 1.3134,
+      "step": 3135
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019244003223377668,
+      "loss": 1.4156,
+      "step": 3136
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924340890416749,
+      "loss": 1.1868,
+      "step": 3137
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019242814360624711,
+      "loss": 1.1961,
+      "step": 3138
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019242219592763768,
+      "loss": 1.2606,
+      "step": 3139
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924162460059909,
+      "loss": 1.2303,
+      "step": 3140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001924102938414512,
+      "loss": 1.2137,
+      "step": 3141
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019240433943416304,
+      "loss": 2.6138,
+      "step": 3142
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001923983827842709,
+      "loss": 1.3201,
+      "step": 3143
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019239242389191934,
+      "loss": 1.4199,
+      "step": 3144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000192386462757253,
+      "loss": 1.2596,
+      "step": 3145
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019238049938041657,
+      "loss": 1.528,
+      "step": 3146
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019237453376155474,
+      "loss": 1.2059,
+      "step": 3147
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019236856590081235,
+      "loss": 1.3609,
+      "step": 3148
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019236259579833418,
+      "loss": 1.308,
+      "step": 3149
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019235662345426513,
+      "loss": 1.2863,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001923506488687502,
+      "loss": 1.2769,
+      "step": 3151
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019234467204193428,
+      "loss": 1.3102,
+      "step": 3152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019233869297396254,
+      "loss": 1.229,
+      "step": 3153
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019233271166498,
+      "loss": 1.2433,
+      "step": 3154
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019232672811513192,
+      "loss": 1.356,
+      "step": 3155
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019232074232456342,
+      "loss": 1.0909,
+      "step": 3156
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019231475429341984,
+      "loss": 1.1007,
+      "step": 3157
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019230876402184644,
+      "loss": 1.2175,
+      "step": 3158
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019230277150998867,
+      "loss": 1.5112,
+      "step": 3159
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019229677675799194,
+      "loss": 1.1822,
+      "step": 3160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001922907797660017,
+      "loss": 1.2914,
+      "step": 3161
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019228478053416357,
+      "loss": 1.2354,
+      "step": 3162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019227877906262308,
+      "loss": 1.0886,
+      "step": 3163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019227277535152594,
+      "loss": 1.3738,
+      "step": 3164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019226676940101777,
+      "loss": 1.3212,
+      "step": 3165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001922607612112444,
+      "loss": 1.2248,
+      "step": 3166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019225475078235167,
+      "loss": 1.1903,
+      "step": 3167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001922487381144854,
+      "loss": 1.4602,
+      "step": 3168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001922427232077915,
+      "loss": 2.5718,
+      "step": 3169
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000192236706062416,
+      "loss": 1.5455,
+      "step": 3170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001922306866785049,
+      "loss": 1.4031,
+      "step": 3171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001922246650562043,
+      "loss": 1.2863,
+      "step": 3172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001922186411956603,
+      "loss": 1.1541,
+      "step": 3173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001922126150970192,
+      "loss": 1.3976,
+      "step": 3174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019220658676042713,
+      "loss": 1.4728,
+      "step": 3175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019220055618603046,
+      "loss": 1.3168,
+      "step": 3176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019219452337397556,
+      "loss": 2.5964,
+      "step": 3177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019218848832440878,
+      "loss": 1.2302,
+      "step": 3178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019218245103747665,
+      "loss": 1.1286,
+      "step": 3179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019217641151332568,
+      "loss": 1.1972,
+      "step": 3180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019217036975210242,
+      "loss": 1.2897,
+      "step": 3181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001921643257539535,
+      "loss": 1.1892,
+      "step": 3182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019215827951902563,
+      "loss": 1.123,
+      "step": 3183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019215223104746554,
+      "loss": 1.2893,
+      "step": 3184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019214618033942005,
+      "loss": 2.5164,
+      "step": 3185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019214012739503594,
+      "loss": 2.5996,
+      "step": 3186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019213407221446016,
+      "loss": 1.2921,
+      "step": 3187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019212801479783967,
+      "loss": 1.3976,
+      "step": 3188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019212195514532146,
+      "loss": 1.1919,
+      "step": 3189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019211589325705258,
+      "loss": 1.2465,
+      "step": 3190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019210982913318018,
+      "loss": 1.1252,
+      "step": 3191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019210376277385143,
+      "loss": 1.1604,
+      "step": 3192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019209769417921356,
+      "loss": 1.3246,
+      "step": 3193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019209162334941384,
+      "loss": 1.1487,
+      "step": 3194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019208555028459961,
+      "loss": 1.309,
+      "step": 3195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019207947498491823,
+      "loss": 1.2218,
+      "step": 3196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001920733974505172,
+      "loss": 1.2195,
+      "step": 3197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019206731768154397,
+      "loss": 1.4423,
+      "step": 3198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019206123567814612,
+      "loss": 1.4267,
+      "step": 3199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019205515144047128,
+      "loss": 2.7476,
+      "step": 3200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019204906496866705,
+      "loss": 1.1732,
+      "step": 3201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019204297626288118,
+      "loss": 1.1855,
+      "step": 3202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019203688532326143,
+      "loss": 1.3084,
+      "step": 3203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019203079214995565,
+      "loss": 1.202,
+      "step": 3204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019202469674311167,
+      "loss": 1.1855,
+      "step": 3205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001920185991028775,
+      "loss": 1.212,
+      "step": 3206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019201249922940107,
+      "loss": 1.2028,
+      "step": 3207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019200639712283038,
+      "loss": 1.2956,
+      "step": 3208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001920002927833136,
+      "loss": 1.3481,
+      "step": 3209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019199418621099884,
+      "loss": 2.6765,
+      "step": 3210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019198807740603436,
+      "loss": 1.227,
+      "step": 3211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019198196636856831,
+      "loss": 1.2358,
+      "step": 3212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001919758530987491,
+      "loss": 1.3159,
+      "step": 3213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019196973759672505,
+      "loss": 1.2798,
+      "step": 3214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001919636198626446,
+      "loss": 1.3054,
+      "step": 3215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001919574998966562,
+      "loss": 2.2815,
+      "step": 3216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001919513776989084,
+      "loss": 1.2614,
+      "step": 3217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019194525326954976,
+      "loss": 1.3908,
+      "step": 3218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019193912660872897,
+      "loss": 1.3578,
+      "step": 3219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019193299771659466,
+      "loss": 1.3186,
+      "step": 3220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019192686659329558,
+      "loss": 1.2113,
+      "step": 3221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019192073323898057,
+      "loss": 1.2283,
+      "step": 3222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019191459765379845,
+      "loss": 1.4315,
+      "step": 3223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019190845983789814,
+      "loss": 1.2984,
+      "step": 3224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019190231979142861,
+      "loss": 1.3082,
+      "step": 3225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019189617751453884,
+      "loss": 1.1394,
+      "step": 3226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019189003300737793,
+      "loss": 1.1774,
+      "step": 3227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000191883886270095,
+      "loss": 2.4646,
+      "step": 3228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019187773730283922,
+      "loss": 1.4316,
+      "step": 3229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019187158610575983,
+      "loss": 1.2668,
+      "step": 3230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019186543267900614,
+      "loss": 1.1816,
+      "step": 3231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019185927702272744,
+      "loss": 1.3376,
+      "step": 3232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019185311913707317,
+      "loss": 2.5,
+      "step": 3233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019184695902219273,
+      "loss": 1.2654,
+      "step": 3234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019184079667823568,
+      "loss": 1.2825,
+      "step": 3235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001918346321053515,
+      "loss": 1.2879,
+      "step": 3236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001918284653036899,
+      "loss": 1.5315,
+      "step": 3237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019182229627340047,
+      "loss": 1.3379,
+      "step": 3238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019181612501463296,
+      "loss": 1.2955,
+      "step": 3239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001918099515275371,
+      "loss": 1.3336,
+      "step": 3240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001918037758122628,
+      "loss": 1.1674,
+      "step": 3241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019179759786895986,
+      "loss": 1.3463,
+      "step": 3242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001917914176977783,
+      "loss": 1.3823,
+      "step": 3243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019178523529886802,
+      "loss": 1.4173,
+      "step": 3244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019177905067237907,
+      "loss": 1.3904,
+      "step": 3245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019177286381846163,
+      "loss": 1.2525,
+      "step": 3246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019176667473726577,
+      "loss": 1.2933,
+      "step": 3247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019176048342894176,
+      "loss": 1.3049,
+      "step": 3248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019175428989363978,
+      "loss": 1.2151,
+      "step": 3249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019174809413151025,
+      "loss": 1.2673,
+      "step": 3250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019174189614270347,
+      "loss": 1.2591,
+      "step": 3251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019173569592736983,
+      "loss": 1.3896,
+      "step": 3252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019172949348565985,
+      "loss": 1.2103,
+      "step": 3253
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001917232888177241,
+      "loss": 1.3153,
+      "step": 3254
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019171708192371306,
+      "loss": 1.2769,
+      "step": 3255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001917108728037775,
+      "loss": 1.3383,
+      "step": 3256
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019170466145806799,
+      "loss": 1.3412,
+      "step": 3257
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019169844788673532,
+      "loss": 1.1565,
+      "step": 3258
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019169223208993032,
+      "loss": 1.1665,
+      "step": 3259
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001916860140678038,
+      "loss": 2.6494,
+      "step": 3260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019167979382050672,
+      "loss": 1.3312,
+      "step": 3261
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019167357134818998,
+      "loss": 1.3167,
+      "step": 3262
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019166734665100466,
+      "loss": 1.3718,
+      "step": 3263
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019166111972910178,
+      "loss": 1.3937,
+      "step": 3264
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019165489058263247,
+      "loss": 1.1912,
+      "step": 3265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001916486592117479,
+      "loss": 1.3516,
+      "step": 3266
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019164242561659934,
+      "loss": 1.3446,
+      "step": 3267
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019163618979733808,
+      "loss": 1.2361,
+      "step": 3268
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019162995175411545,
+      "loss": 1.3069,
+      "step": 3269
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019162371148708277,
+      "loss": 1.23,
+      "step": 3270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001916174689963916,
+      "loss": 1.1918,
+      "step": 3271
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019161122428219338,
+      "loss": 1.3702,
+      "step": 3272
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019160497734463966,
+      "loss": 1.191,
+      "step": 3273
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019159872818388208,
+      "loss": 1.2097,
+      "step": 3274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019159247680007228,
+      "loss": 1.2503,
+      "step": 3275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000191586223193362,
+      "loss": 1.3156,
+      "step": 3276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019157996736390297,
+      "loss": 1.2162,
+      "step": 3277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019157370931184708,
+      "loss": 1.3331,
+      "step": 3278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019156744903734617,
+      "loss": 2.5327,
+      "step": 3279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019156118654055216,
+      "loss": 1.2233,
+      "step": 3280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019155492182161706,
+      "loss": 1.2967,
+      "step": 3281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019154865488069293,
+      "loss": 1.2135,
+      "step": 3282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019154238571793183,
+      "loss": 1.1387,
+      "step": 3283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001915361143334859,
+      "loss": 1.1697,
+      "step": 3284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001915298407275074,
+      "loss": 1.3037,
+      "step": 3285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019152356490014853,
+      "loss": 1.411,
+      "step": 3286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019151728685156162,
+      "loss": 1.2603,
+      "step": 3287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019151100658189906,
+      "loss": 1.3073,
+      "step": 3288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019150472409131322,
+      "loss": 2.4568,
+      "step": 3289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019149843937995664,
+      "loss": 1.4844,
+      "step": 3290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019149215244798176,
+      "loss": 1.3233,
+      "step": 3291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019148586329554123,
+      "loss": 1.3724,
+      "step": 3292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019147957192278765,
+      "loss": 1.3191,
+      "step": 3293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019147327832987374,
+      "loss": 1.4336,
+      "step": 3294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001914669825169522,
+      "loss": 1.3616,
+      "step": 3295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019146068448417584,
+      "loss": 2.5903,
+      "step": 3296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019145438423169754,
+      "loss": 1.2759,
+      "step": 3297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019144808175967017,
+      "loss": 1.2562,
+      "step": 3298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001914417770682467,
+      "loss": 2.2769,
+      "step": 3299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019143547015758013,
+      "loss": 1.1683,
+      "step": 3300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019142916102782355,
+      "loss": 1.3281,
+      "step": 3301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019142284967913007,
+      "loss": 1.3159,
+      "step": 3302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019141653611165285,
+      "loss": 1.1254,
+      "step": 3303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019141022032554512,
+      "loss": 1.1859,
+      "step": 3304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019140390232096017,
+      "loss": 1.0641,
+      "step": 3305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019139758209805132,
+      "loss": 1.3044,
+      "step": 3306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000191391259656972,
+      "loss": 1.407,
+      "step": 3307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019138493499787557,
+      "loss": 2.6348,
+      "step": 3308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019137860812091563,
+      "loss": 1.2364,
+      "step": 3309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019137227902624564,
+      "loss": 1.2118,
+      "step": 3310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019136594771401922,
+      "loss": 1.2656,
+      "step": 3311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001913596141843901,
+      "loss": 1.4041,
+      "step": 3312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001913532784375119,
+      "loss": 1.1811,
+      "step": 3313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019134694047353844,
+      "loss": 1.4551,
+      "step": 3314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019134060029262355,
+      "loss": 1.0893,
+      "step": 3315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019133425789492106,
+      "loss": 1.3596,
+      "step": 3316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019132791328058487,
+      "loss": 1.2664,
+      "step": 3317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019132156644976906,
+      "loss": 1.2536,
+      "step": 3318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001913152174026276,
+      "loss": 1.3173,
+      "step": 3319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019130886613931455,
+      "loss": 1.4785,
+      "step": 3320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019130251265998412,
+      "loss": 1.2477,
+      "step": 3321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019129615696479046,
+      "loss": 1.2343,
+      "step": 3322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001912897990538878,
+      "loss": 1.2642,
+      "step": 3323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019128343892743055,
+      "loss": 1.2121,
+      "step": 3324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019127707658557296,
+      "loss": 1.1569,
+      "step": 3325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019127071202846945,
+      "loss": 1.0814,
+      "step": 3326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019126434525627451,
+      "loss": 1.5298,
+      "step": 3327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001912579762691427,
+      "loss": 1.2815,
+      "step": 3328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019125160506722847,
+      "loss": 1.3136,
+      "step": 3329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019124523165068658,
+      "loss": 1.3807,
+      "step": 3330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019123885601967164,
+      "loss": 1.3314,
+      "step": 3331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001912324781743384,
+      "loss": 1.384,
+      "step": 3332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019122609811484162,
+      "loss": 1.0303,
+      "step": 3333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001912197158413362,
+      "loss": 1.3369,
+      "step": 3334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019121333135397697,
+      "loss": 1.3171,
+      "step": 3335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001912069446529189,
+      "loss": 1.4723,
+      "step": 3336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000191200555738317,
+      "loss": 1.2814,
+      "step": 3337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019119416461032635,
+      "loss": 1.3094,
+      "step": 3338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019118777126910202,
+      "loss": 1.2862,
+      "step": 3339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019118137571479918,
+      "loss": 1.2247,
+      "step": 3340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019117497794757306,
+      "loss": 1.1465,
+      "step": 3341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019116857796757893,
+      "loss": 1.0865,
+      "step": 3342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019116217577497206,
+      "loss": 1.2367,
+      "step": 3343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019115577136990793,
+      "loss": 1.2961,
+      "step": 3344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001911493647525419,
+      "loss": 1.4006,
+      "step": 3345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019114295592302945,
+      "loss": 1.1816,
+      "step": 3346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019113654488152614,
+      "loss": 1.4471,
+      "step": 3347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001911301316281876,
+      "loss": 1.2159,
+      "step": 3348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019112371616316937,
+      "loss": 1.252,
+      "step": 3349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019111729848662727,
+      "loss": 1.4623,
+      "step": 3350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019111087859871698,
+      "loss": 1.2191,
+      "step": 3351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019110445649959433,
+      "loss": 1.3155,
+      "step": 3352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019109803218941523,
+      "loss": 1.291,
+      "step": 3353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019109160566833548,
+      "loss": 1.3602,
+      "step": 3354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019108517693651114,
+      "loss": 1.0659,
+      "step": 3355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019107874599409822,
+      "loss": 1.2584,
+      "step": 3356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019107231284125274,
+      "loss": 1.4763,
+      "step": 3357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019106587747813093,
+      "loss": 1.2961,
+      "step": 3358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019105943990488884,
+      "loss": 1.1957,
+      "step": 3359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001910530001216828,
+      "loss": 1.2534,
+      "step": 3360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019104655812866912,
+      "loss": 1.4979,
+      "step": 3361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019104011392600406,
+      "loss": 1.2004,
+      "step": 3362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019103366751384407,
+      "loss": 1.4113,
+      "step": 3363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001910272188923456,
+      "loss": 1.4816,
+      "step": 3364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019102076806166513,
+      "loss": 1.2052,
+      "step": 3365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019101431502195927,
+      "loss": 1.1547,
+      "step": 3366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019100785977338455,
+      "loss": 1.1561,
+      "step": 3367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019100140231609772,
+      "loss": 1.4637,
+      "step": 3368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019099494265025543,
+      "loss": 1.4088,
+      "step": 3369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001909884807760145,
+      "loss": 1.2069,
+      "step": 3370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019098201669353175,
+      "loss": 1.2646,
+      "step": 3371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019097555040296403,
+      "loss": 1.109,
+      "step": 3372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019096908190446827,
+      "loss": 1.2195,
+      "step": 3373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001909626111982015,
+      "loss": 1.2643,
+      "step": 3374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019095613828432075,
+      "loss": 1.1699,
+      "step": 3375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001909496631629831,
+      "loss": 1.092,
+      "step": 3376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001909431858343457,
+      "loss": 1.1951,
+      "step": 3377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019093670629856578,
+      "loss": 1.2747,
+      "step": 3378
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001909302245558005,
+      "loss": 1.392,
+      "step": 3379
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001909237406062073,
+      "loss": 1.3613,
+      "step": 3380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019091725444994343,
+      "loss": 2.5872,
+      "step": 3381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001909107660871664,
+      "loss": 1.3145,
+      "step": 3382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001909042755180336,
+      "loss": 1.2083,
+      "step": 3383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001908977827427026,
+      "loss": 1.1708,
+      "step": 3384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019089128776133096,
+      "loss": 1.2674,
+      "step": 3385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001908847905740763,
+      "loss": 1.4489,
+      "step": 3386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019087829118109634,
+      "loss": 1.2819,
+      "step": 3387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019087178958254877,
+      "loss": 1.3199,
+      "step": 3388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001908652857785914,
+      "loss": 1.2429,
+      "step": 3389
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001908587797693821,
+      "loss": 2.6555,
+      "step": 3390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019085227155507872,
+      "loss": 1.4751,
+      "step": 3391
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019084576113583923,
+      "loss": 1.3594,
+      "step": 3392
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019083924851182166,
+      "loss": 1.2258,
+      "step": 3393
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019083273368318405,
+      "loss": 1.201,
+      "step": 3394
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001908262166500845,
+      "loss": 1.1381,
+      "step": 3395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019081969741268116,
+      "loss": 1.3314,
+      "step": 3396
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001908131759711323,
+      "loss": 1.0735,
+      "step": 3397
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019080665232559614,
+      "loss": 1.1245,
+      "step": 3398
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019080012647623102,
+      "loss": 1.315,
+      "step": 3399
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019079359842319536,
+      "loss": 1.5391,
+      "step": 3400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019078706816664753,
+      "loss": 1.0852,
+      "step": 3401
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019078053570674607,
+      "loss": 1.1805,
+      "step": 3402
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001907740010436495,
+      "loss": 1.3401,
+      "step": 3403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019076746417751638,
+      "loss": 1.2426,
+      "step": 3404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001907609251085054,
+      "loss": 1.2594,
+      "step": 3405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019075438383677523,
+      "loss": 1.3774,
+      "step": 3406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019074784036248463,
+      "loss": 1.256,
+      "step": 3407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019074129468579245,
+      "loss": 1.1701,
+      "step": 3408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019073474680685748,
+      "loss": 1.4025,
+      "step": 3409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019072819672583867,
+      "loss": 1.4823,
+      "step": 3410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019072164444289502,
+      "loss": 1.4177,
+      "step": 3411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019071508995818548,
+      "loss": 1.385,
+      "step": 3412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019070853327186917,
+      "loss": 1.2245,
+      "step": 3413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001907019743841052,
+      "loss": 1.2041,
+      "step": 3414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019069541329505273,
+      "loss": 1.3773,
+      "step": 3415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019068885000487107,
+      "loss": 1.2438,
+      "step": 3416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019068228451371943,
+      "loss": 1.2094,
+      "step": 3417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019067571682175718,
+      "loss": 1.4181,
+      "step": 3418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019066914692914374,
+      "loss": 1.2189,
+      "step": 3419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001906625748360385,
+      "loss": 1.3424,
+      "step": 3420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019065600054260097,
+      "loss": 1.3877,
+      "step": 3421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019064942404899076,
+      "loss": 1.3803,
+      "step": 3422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019064284535536745,
+      "loss": 1.289,
+      "step": 3423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001906362644618907,
+      "loss": 2.6409,
+      "step": 3424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001906296813687202,
+      "loss": 1.3451,
+      "step": 3425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019062309607601575,
+      "loss": 1.3105,
+      "step": 3426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019061650858393715,
+      "loss": 1.2456,
+      "step": 3427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019060991889264427,
+      "loss": 1.21,
+      "step": 3428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001906033270022971,
+      "loss": 1.1198,
+      "step": 3429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019059673291305554,
+      "loss": 1.2916,
+      "step": 3430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019059013662507966,
+      "loss": 1.2079,
+      "step": 3431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019058353813852956,
+      "loss": 1.322,
+      "step": 3432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001905769374535654,
+      "loss": 1.2951,
+      "step": 3433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001905703345703473,
+      "loss": 1.18,
+      "step": 3434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019056372948903557,
+      "loss": 1.2874,
+      "step": 3435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001905571222097905,
+      "loss": 1.2501,
+      "step": 3436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019055051273277241,
+      "loss": 1.1066,
+      "step": 3437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001905439010581418,
+      "loss": 1.4363,
+      "step": 3438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019053728718605904,
+      "loss": 1.3006,
+      "step": 3439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019053067111668467,
+      "loss": 1.4146,
+      "step": 3440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019052405285017928,
+      "loss": 1.1255,
+      "step": 3441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019051743238670342,
+      "loss": 1.2093,
+      "step": 3442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019051080972641788,
+      "loss": 1.3155,
+      "step": 3443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019050418486948333,
+      "loss": 1.1445,
+      "step": 3444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019049755781606054,
+      "loss": 1.3104,
+      "step": 3445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019049092856631035,
+      "loss": 1.4012,
+      "step": 3446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019048429712039365,
+      "loss": 2.6392,
+      "step": 3447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019047766347847142,
+      "loss": 1.4667,
+      "step": 3448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019047102764070458,
+      "loss": 1.3842,
+      "step": 3449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019046438960725425,
+      "loss": 1.343,
+      "step": 3450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019045774937828148,
+      "loss": 1.3317,
+      "step": 3451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019045110695394747,
+      "loss": 1.2622,
+      "step": 3452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001904444623344134,
+      "loss": 1.1755,
+      "step": 3453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019043781551984052,
+      "loss": 1.4536,
+      "step": 3454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019043116651039016,
+      "loss": 1.1573,
+      "step": 3455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001904245153062237,
+      "loss": 1.3362,
+      "step": 3456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019041786190750255,
+      "loss": 0.9344,
+      "step": 3457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019041120631438818,
+      "loss": 1.2927,
+      "step": 3458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019040454852704212,
+      "loss": 2.5791,
+      "step": 3459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019039788854562597,
+      "loss": 1.4547,
+      "step": 3460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019039122637030134,
+      "loss": 1.2669,
+      "step": 3461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019038456200122988,
+      "loss": 1.3839,
+      "step": 3462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019037789543857345,
+      "loss": 2.196,
+      "step": 3463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001903712266824937,
+      "loss": 1.1269,
+      "step": 3464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001903645557331526,
+      "loss": 2.4556,
+      "step": 3465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019035788259071196,
+      "loss": 1.0775,
+      "step": 3466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001903512072553338,
+      "loss": 1.2441,
+      "step": 3467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019034452972718007,
+      "loss": 1.1223,
+      "step": 3468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001903378500064129,
+      "loss": 1.1977,
+      "step": 3469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001903311680931943,
+      "loss": 1.3586,
+      "step": 3470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019032448398768653,
+      "loss": 1.3422,
+      "step": 3471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019031779769005178,
+      "loss": 1.2485,
+      "step": 3472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001903111092004523,
+      "loss": 1.2546,
+      "step": 3473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019030441851905044,
+      "loss": 1.4144,
+      "step": 3474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019029772564600858,
+      "loss": 1.2646,
+      "step": 3475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019029103058148915,
+      "loss": 1.1463,
+      "step": 3476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019028433332565463,
+      "loss": 1.3682,
+      "step": 3477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019027763387866758,
+      "loss": 1.3809,
+      "step": 3478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019027093224069055,
+      "loss": 1.356,
+      "step": 3479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001902642284118862,
+      "loss": 1.2908,
+      "step": 3480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019025752239241727,
+      "loss": 1.3459,
+      "step": 3481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019025081418244644,
+      "loss": 1.2943,
+      "step": 3482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019024410378213656,
+      "loss": 1.213,
+      "step": 3483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001902373911916505,
+      "loss": 1.1099,
+      "step": 3484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019023067641115116,
+      "loss": 2.4977,
+      "step": 3485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019022395944080146,
+      "loss": 1.2853,
+      "step": 3486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019021724028076448,
+      "loss": 1.2478,
+      "step": 3487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019021051893120325,
+      "loss": 1.591,
+      "step": 3488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019020379539228094,
+      "loss": 1.2007,
+      "step": 3489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019019706966416067,
+      "loss": 1.282,
+      "step": 3490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019019034174700566,
+      "loss": 2.4585,
+      "step": 3491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019018361164097928,
+      "loss": 1.4508,
+      "step": 3492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001901768793462448,
+      "loss": 1.5037,
+      "step": 3493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019017014486296562,
+      "loss": 1.3134,
+      "step": 3494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001901634081913052,
+      "loss": 1.2982,
+      "step": 3495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000190156669331427,
+      "loss": 1.392,
+      "step": 3496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019014992828349456,
+      "loss": 1.3546,
+      "step": 3497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019014318504767153,
+      "loss": 1.3574,
+      "step": 3498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019013643962412154,
+      "loss": 1.3184,
+      "step": 3499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019012969201300835,
+      "loss": 2.3586,
+      "step": 3500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019012294221449561,
+      "loss": 1.3649,
+      "step": 3501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019011619022874724,
+      "loss": 1.3023,
+      "step": 3502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019010943605592706,
+      "loss": 2.2952,
+      "step": 3503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000190102679696199,
+      "loss": 1.298,
+      "step": 3504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000190095921149727,
+      "loss": 1.4204,
+      "step": 3505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019008916041667512,
+      "loss": 1.2485,
+      "step": 3506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019008239749720746,
+      "loss": 1.2257,
+      "step": 3507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001900756323914881,
+      "loss": 1.2755,
+      "step": 3508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019006886509968126,
+      "loss": 1.2758,
+      "step": 3509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019006209562195115,
+      "loss": 1.2134,
+      "step": 3510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019005532395846212,
+      "loss": 1.3024,
+      "step": 3511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019004855010937844,
+      "loss": 1.2699,
+      "step": 3512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001900417740748646,
+      "loss": 1.3382,
+      "step": 3513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019003499585508492,
+      "loss": 1.3599,
+      "step": 3514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019002821545020403,
+      "loss": 2.4404,
+      "step": 3515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019002143286038644,
+      "loss": 1.357,
+      "step": 3516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001900146480857967,
+      "loss": 1.0832,
+      "step": 3517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001900078611265996,
+      "loss": 1.3846,
+      "step": 3518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019000107198295974,
+      "loss": 1.1003,
+      "step": 3519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018999428065504196,
+      "loss": 1.2978,
+      "step": 3520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000189987487143011,
+      "loss": 1.3734,
+      "step": 3521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018998069144703187,
+      "loss": 1.2734,
+      "step": 3522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018997389356726935,
+      "loss": 2.5713,
+      "step": 3523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018996709350388853,
+      "loss": 1.2174,
+      "step": 3524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001899602912570544,
+      "loss": 2.5654,
+      "step": 3525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000189953486826932,
+      "loss": 1.4142,
+      "step": 3526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018994668021368657,
+      "loss": 1.3551,
+      "step": 3527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018993987141748323,
+      "loss": 1.4528,
+      "step": 3528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018993306043848726,
+      "loss": 1.233,
+      "step": 3529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018992624727686395,
+      "loss": 1.3309,
+      "step": 3530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018991943193277864,
+      "loss": 1.3427,
+      "step": 3531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018991261440639672,
+      "loss": 1.1272,
+      "step": 3532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018990579469788373,
+      "loss": 1.3295,
+      "step": 3533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018989897280740508,
+      "loss": 1.4041,
+      "step": 3534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018989214873512636,
+      "loss": 1.4437,
+      "step": 3535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018988532248121324,
+      "loss": 1.3249,
+      "step": 3536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001898784940458313,
+      "loss": 1.2452,
+      "step": 3537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018987166342914638,
+      "loss": 1.3193,
+      "step": 3538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018986483063132418,
+      "loss": 1.3071,
+      "step": 3539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001898579956525305,
+      "loss": 2.6433,
+      "step": 3540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018985115849293126,
+      "loss": 1.0946,
+      "step": 3541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001898443191526924,
+      "loss": 1.2938,
+      "step": 3542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001898374776319799,
+      "loss": 1.4089,
+      "step": 3543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018983063393095983,
+      "loss": 1.3451,
+      "step": 3544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018982378804979826,
+      "loss": 1.3911,
+      "step": 3545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018981693998866128,
+      "loss": 1.2213,
+      "step": 3546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018981008974771517,
+      "loss": 1.2739,
+      "step": 3547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018980323732712613,
+      "loss": 1.2215,
+      "step": 3548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897963827270605,
+      "loss": 2.8076,
+      "step": 3549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018978952594768462,
+      "loss": 1.1907,
+      "step": 3550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897826669891649,
+      "loss": 1.0454,
+      "step": 3551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018977580585166778,
+      "loss": 1.1849,
+      "step": 3552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018976894253535984,
+      "loss": 0.992,
+      "step": 3553
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897620770404076,
+      "loss": 1.311,
+      "step": 3554
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897552093669777,
+      "loss": 1.182,
+      "step": 3555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897483395152368,
+      "loss": 2.3875,
+      "step": 3556
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018974146748535162,
+      "loss": 1.2572,
+      "step": 3557
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018973459327748898,
+      "loss": 2.3977,
+      "step": 3558
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018972771689181568,
+      "loss": 1.0816,
+      "step": 3559
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897208383284986,
+      "loss": 1.3605,
+      "step": 3560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001897139575877047,
+      "loss": 1.3542,
+      "step": 3561
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000189707074669601,
+      "loss": 1.2533,
+      "step": 3562
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018970018957435448,
+      "loss": 2.5127,
+      "step": 3563
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018969330230213225,
+      "loss": 1.4461,
+      "step": 3564
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001896864128531015,
+      "loss": 1.4196,
+      "step": 3565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001896795212274294,
+      "loss": 1.2214,
+      "step": 3566
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018967262742528322,
+      "loss": 1.2758,
+      "step": 3567
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001896657314468303,
+      "loss": 1.2483,
+      "step": 3568
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018965883329223794,
+      "loss": 1.2432,
+      "step": 3569
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001896519329616736,
+      "loss": 1.3067,
+      "step": 3570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001896450304553047,
+      "loss": 1.196,
+      "step": 3571
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018963812577329882,
+      "loss": 1.1255,
+      "step": 3572
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018963121891582349,
+      "loss": 1.1638,
+      "step": 3573
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018962430988304636,
+      "loss": 1.2097,
+      "step": 3574
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018961739867513506,
+      "loss": 1.3056,
+      "step": 3575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001896104852922574,
+      "loss": 1.3244,
+      "step": 3576
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018960356973458112,
+      "loss": 1.1227,
+      "step": 3577
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018959665200227405,
+      "loss": 2.5906,
+      "step": 3578
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001895897320955041,
+      "loss": 1.1902,
+      "step": 3579
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001895828100144392,
+      "loss": 2.4956,
+      "step": 3580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018957588575924735,
+      "loss": 1.2584,
+      "step": 3581
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018956895933009657,
+      "loss": 1.4371,
+      "step": 3582
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000189562030727155,
+      "loss": 1.3083,
+      "step": 3583
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001895550999505908,
+      "loss": 1.3232,
+      "step": 3584
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001895481670005721,
+      "loss": 1.3088,
+      "step": 3585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00018954123187726727,
+      "loss": 1.2271,
+      "step": 3586
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018953429458084455,
+      "loss": 1.2055,
+      "step": 3587
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018952735511147234,
+      "loss": 1.4459,
+      "step": 3588
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018952041346931902,
+      "loss": 1.207,
+      "step": 3589
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018951346965455308,
+      "loss": 1.1974,
+      "step": 3590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018950652366734303,
+      "loss": 1.149,
+      "step": 3591
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018949957550785747,
+      "loss": 1.3295,
+      "step": 3592
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018949262517626504,
+      "loss": 1.1158,
+      "step": 3593
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018948567267273433,
+      "loss": 1.1493,
+      "step": 3594
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018947871799743416,
+      "loss": 1.3639,
+      "step": 3595
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018947176115053334,
+      "loss": 1.2646,
+      "step": 3596
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018946480213220061,
+      "loss": 1.1927,
+      "step": 3597
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018945784094260497,
+      "loss": 1.1838,
+      "step": 3598
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018945087758191526,
+      "loss": 1.2962,
+      "step": 3599
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018944391205030057,
+      "loss": 1.4633,
+      "step": 3600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001894369443479299,
+      "loss": 1.2887,
+      "step": 3601
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001894299744749723,
+      "loss": 1.4655,
+      "step": 3602
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018942300243159706,
+      "loss": 1.4883,
+      "step": 3603
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001894160282179733,
+      "loss": 1.2068,
+      "step": 3604
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018940905183427025,
+      "loss": 1.2927,
+      "step": 3605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018940207328065729,
+      "loss": 1.1559,
+      "step": 3606
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018939509255730378,
+      "loss": 1.4415,
+      "step": 3607
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018938810966437908,
+      "loss": 1.1522,
+      "step": 3608
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001893811246020527,
+      "loss": 1.3078,
+      "step": 3609
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001893741373704942,
+      "loss": 1.4973,
+      "step": 3610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001893671479698731,
+      "loss": 1.1694,
+      "step": 3611
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018936015640035907,
+      "loss": 1.2455,
+      "step": 3612
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018935316266212174,
+      "loss": 1.2075,
+      "step": 3613
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001893461667553309,
+      "loss": 2.5864,
+      "step": 3614
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018933916868015633,
+      "loss": 1.3064,
+      "step": 3615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018933216843676785,
+      "loss": 1.2336,
+      "step": 3616
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018932516602533533,
+      "loss": 1.2537,
+      "step": 3617
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018931816144602875,
+      "loss": 1.3332,
+      "step": 3618
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018931115469901809,
+      "loss": 1.252,
+      "step": 3619
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001893041457844734,
+      "loss": 1.2418,
+      "step": 3620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001892971347025648,
+      "loss": 1.1271,
+      "step": 3621
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018929012145346244,
+      "loss": 1.4055,
+      "step": 3622
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018928310603733651,
+      "loss": 1.3402,
+      "step": 3623
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018927608845435728,
+      "loss": 1.4018,
+      "step": 3624
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018926906870469508,
+      "loss": 1.3315,
+      "step": 3625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018926204678852026,
+      "loss": 1.3154,
+      "step": 3626
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018925502270600325,
+      "loss": 1.2227,
+      "step": 3627
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018924799645731447,
+      "loss": 1.3481,
+      "step": 3628
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018924096804262452,
+      "loss": 1.3079,
+      "step": 3629
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018923393746210393,
+      "loss": 1.1863,
+      "step": 3630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018922690471592335,
+      "loss": 1.2751,
+      "step": 3631
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018921986980425344,
+      "loss": 1.1024,
+      "step": 3632
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018921283272726493,
+      "loss": 1.3162,
+      "step": 3633
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018920579348512864,
+      "loss": 1.4242,
+      "step": 3634
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018919875207801536,
+      "loss": 1.1459,
+      "step": 3635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018919170850609603,
+      "loss": 2.7473,
+      "step": 3636
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001891846627695416,
+      "loss": 1.2577,
+      "step": 3637
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018917761486852297,
+      "loss": 1.1369,
+      "step": 3638
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001891705648032113,
+      "loss": 1.4869,
+      "step": 3639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018916351257377764,
+      "loss": 1.1951,
+      "step": 3640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018915645818039315,
+      "loss": 1.2499,
+      "step": 3641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018914940162322906,
+      "loss": 1.2574,
+      "step": 3642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001891423429024566,
+      "loss": 1.2473,
+      "step": 3643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018913528201824705,
+      "loss": 1.2874,
+      "step": 3644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018912821897077185,
+      "loss": 1.1417,
+      "step": 3645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018912115376020237,
+      "loss": 1.2449,
+      "step": 3646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018911408638671011,
+      "loss": 1.1967,
+      "step": 3647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018910701685046654,
+      "loss": 1.3525,
+      "step": 3648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018909994515164326,
+      "loss": 1.3135,
+      "step": 3649
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018909287129041191,
+      "loss": 1.1746,
+      "step": 3650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018908579526694417,
+      "loss": 1.3904,
+      "step": 3651
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018907871708141177,
+      "loss": 1.3224,
+      "step": 3652
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018907163673398647,
+      "loss": 1.267,
+      "step": 3653
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001890645542248401,
+      "loss": 1.2868,
+      "step": 3654
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018905746955414458,
+      "loss": 1.2756,
+      "step": 3655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018905038272207185,
+      "loss": 1.4152,
+      "step": 3656
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001890432937287939,
+      "loss": 1.1046,
+      "step": 3657
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018903620257448273,
+      "loss": 1.2546,
+      "step": 3658
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001890291092593105,
+      "loss": 1.1093,
+      "step": 3659
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018902201378344938,
+      "loss": 1.1531,
+      "step": 3660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018901491614707147,
+      "loss": 1.0375,
+      "step": 3661
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018900781635034912,
+      "loss": 1.0703,
+      "step": 3662
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018900071439345462,
+      "loss": 1.3046,
+      "step": 3663
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018899361027656028,
+      "loss": 1.41,
+      "step": 3664
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001889865039998386,
+      "loss": 2.6575,
+      "step": 3665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018897939556346196,
+      "loss": 1.283,
+      "step": 3666
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001889722849676029,
+      "loss": 2.7393,
+      "step": 3667
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018896517221243404,
+      "loss": 1.3596,
+      "step": 3668
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018895805729812794,
+      "loss": 2.4856,
+      "step": 3669
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018895094022485732,
+      "loss": 1.0589,
+      "step": 3670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018894382099279492,
+      "loss": 1.2659,
+      "step": 3671
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018893669960211342,
+      "loss": 1.2643,
+      "step": 3672
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018892957605298576,
+      "loss": 1.3258,
+      "step": 3673
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018892245034558478,
+      "loss": 1.1821,
+      "step": 3674
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018891532248008344,
+      "loss": 1.3805,
+      "step": 3675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001889081924566547,
+      "loss": 1.3422,
+      "step": 3676
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001889010602754716,
+      "loss": 1.4489,
+      "step": 3677
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018889392593670728,
+      "loss": 1.1155,
+      "step": 3678
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018888678944053484,
+      "loss": 1.3203,
+      "step": 3679
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001888796507871275,
+      "loss": 1.298,
+      "step": 3680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018887250997665848,
+      "loss": 1.1666,
+      "step": 3681
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018886536700930115,
+      "loss": 0.9549,
+      "step": 3682
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001888582218852288,
+      "loss": 1.2834,
+      "step": 3683
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018885107460461488,
+      "loss": 1.2543,
+      "step": 3684
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018884392516763283,
+      "loss": 1.137,
+      "step": 3685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018883677357445618,
+      "loss": 1.209,
+      "step": 3686
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018882961982525845,
+      "loss": 1.0058,
+      "step": 3687
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001888224639202133,
+      "loss": 1.4321,
+      "step": 3688
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00018881530585949442,
+      "loss": 1.425,
+      "step": 3689
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001888081456432755,
+      "loss": 1.2085,
+      "step": 3690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001888009832717303,
+      "loss": 1.1925,
+      "step": 3691
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018879381874503266,
+      "loss": 1.4023,
+      "step": 3692
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018878665206335647,
+      "loss": 1.1737,
+      "step": 3693
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018877948322687567,
+      "loss": 1.2133,
+      "step": 3694
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001887723122357642,
+      "loss": 1.2985,
+      "step": 3695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018876513909019615,
+      "loss": 1.4461,
+      "step": 3696
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001887579637903456,
+      "loss": 1.3091,
+      "step": 3697
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018875078633638666,
+      "loss": 1.3782,
+      "step": 3698
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018874360672849352,
+      "loss": 1.2991,
+      "step": 3699
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018873642496684045,
+      "loss": 1.2021,
+      "step": 3700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018872924105160175,
+      "loss": 1.3427,
+      "step": 3701
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018872205498295176,
+      "loss": 1.064,
+      "step": 3702
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018871486676106487,
+      "loss": 1.4042,
+      "step": 3703
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018870767638611555,
+      "loss": 1.1428,
+      "step": 3704
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001887004838582783,
+      "loss": 2.6926,
+      "step": 3705
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001886932891777277,
+      "loss": 1.476,
+      "step": 3706
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018868609234463834,
+      "loss": 1.2234,
+      "step": 3707
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018867889335918487,
+      "loss": 1.3915,
+      "step": 3708
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018867169222154203,
+      "loss": 1.4348,
+      "step": 3709
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018866448893188457,
+      "loss": 1.179,
+      "step": 3710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018865728349038734,
+      "loss": 1.4115,
+      "step": 3711
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001886500758972252,
+      "loss": 1.162,
+      "step": 3712
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018864286615257302,
+      "loss": 1.3469,
+      "step": 3713
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018863565425660588,
+      "loss": 1.3507,
+      "step": 3714
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001886284402094987,
+      "loss": 1.2932,
+      "step": 3715
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018862122401142666,
+      "loss": 1.2007,
+      "step": 3716
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018861400566256483,
+      "loss": 1.2912,
+      "step": 3717
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018860678516308837,
+      "loss": 1.3232,
+      "step": 3718
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001885995625131726,
+      "loss": 1.3584,
+      "step": 3719
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018859233771299278,
+      "loss": 1.1966,
+      "step": 3720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018858511076272423,
+      "loss": 1.3813,
+      "step": 3721
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018857788166254237,
+      "loss": 1.221,
+      "step": 3722
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001885706504126226,
+      "loss": 1.3016,
+      "step": 3723
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018856341701314045,
+      "loss": 1.3657,
+      "step": 3724
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018855618146427147,
+      "loss": 1.3439,
+      "step": 3725
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001885489437661913,
+      "loss": 1.2179,
+      "step": 3726
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018854170391907553,
+      "loss": 1.0123,
+      "step": 3727
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001885344619230999,
+      "loss": 1.2746,
+      "step": 3728
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018852721777844018,
+      "loss": 1.1832,
+      "step": 3729
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018851997148527217,
+      "loss": 1.3897,
+      "step": 3730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001885127230437717,
+      "loss": 1.1848,
+      "step": 3731
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018850547245411473,
+      "loss": 1.2678,
+      "step": 3732
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018849821971647721,
+      "loss": 1.1172,
+      "step": 3733
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001884909648310352,
+      "loss": 1.2291,
+      "step": 3734
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001884837077979647,
+      "loss": 1.1938,
+      "step": 3735
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018847644861744186,
+      "loss": 1.2782,
+      "step": 3736
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018846918728964292,
+      "loss": 1.1609,
+      "step": 3737
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018846192381474403,
+      "loss": 1.1786,
+      "step": 3738
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001884546581929215,
+      "loss": 1.3052,
+      "step": 3739
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018844739042435166,
+      "loss": 1.2022,
+      "step": 3740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018844012050921088,
+      "loss": 1.2617,
+      "step": 3741
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001884328484476756,
+      "loss": 2.8032,
+      "step": 3742
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018842557423992236,
+      "loss": 1.2739,
+      "step": 3743
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018841829788612761,
+      "loss": 1.4056,
+      "step": 3744
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018841101938646803,
+      "loss": 1.3384,
+      "step": 3745
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018840373874112024,
+      "loss": 1.2062,
+      "step": 3746
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883964559502609,
+      "loss": 1.4895,
+      "step": 3747
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883891710140668,
+      "loss": 1.4086,
+      "step": 3748
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018838188393271475,
+      "loss": 1.1541,
+      "step": 3749
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018837459470638158,
+      "loss": 1.3777,
+      "step": 3750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018836730333524418,
+      "loss": 1.1892,
+      "step": 3751
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883600098194795,
+      "loss": 1.2649,
+      "step": 3752
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883527141592646,
+      "loss": 1.124,
+      "step": 3753
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018834541635477653,
+      "loss": 1.1396,
+      "step": 3754
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018833811640619236,
+      "loss": 1.3563,
+      "step": 3755
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883308143136893,
+      "loss": 1.2566,
+      "step": 3756
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883235100774446,
+      "loss": 1.3075,
+      "step": 3757
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883162036976354,
+      "loss": 1.2449,
+      "step": 3758
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018830889517443917,
+      "loss": 1.4694,
+      "step": 3759
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001883015845080332,
+      "loss": 1.3599,
+      "step": 3760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018829427169859492,
+      "loss": 1.196,
+      "step": 3761
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018828695674630183,
+      "loss": 1.3246,
+      "step": 3762
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018827963965133149,
+      "loss": 1.4609,
+      "step": 3763
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001882723204138614,
+      "loss": 1.3424,
+      "step": 3764
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018826499903406928,
+      "loss": 1.3115,
+      "step": 3765
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018825767551213275,
+      "loss": 2.6846,
+      "step": 3766
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001882503498482296,
+      "loss": 1.383,
+      "step": 3767
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018824302204253757,
+      "loss": 1.2097,
+      "step": 3768
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018823569209523453,
+      "loss": 1.2801,
+      "step": 3769
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018822836000649838,
+      "loss": 1.3572,
+      "step": 3770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018822102577650704,
+      "loss": 1.1831,
+      "step": 3771
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018821368940543853,
+      "loss": 1.4697,
+      "step": 3772
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018820635089347086,
+      "loss": 1.183,
+      "step": 3773
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018819901024078224,
+      "loss": 1.418,
+      "step": 3774
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001881916674475507,
+      "loss": 1.331,
+      "step": 3775
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001881843225139545,
+      "loss": 1.1386,
+      "step": 3776
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001881769754401719,
+      "loss": 1.287,
+      "step": 3777
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018816962622638118,
+      "loss": 1.1293,
+      "step": 3778
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018816227487276077,
+      "loss": 1.3423,
+      "step": 3779
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000188154921379489,
+      "loss": 1.2022,
+      "step": 3780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018814756574674437,
+      "loss": 1.2262,
+      "step": 3781
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001881402079747054,
+      "loss": 1.2859,
+      "step": 3782
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018813284806355068,
+      "loss": 1.1541,
+      "step": 3783
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018812548601345879,
+      "loss": 1.3276,
+      "step": 3784
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018811812182460843,
+      "loss": 1.3487,
+      "step": 3785
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018811075549717834,
+      "loss": 1.25,
+      "step": 3786
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018810338703134725,
+      "loss": 1.4185,
+      "step": 3787
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000188096016427294,
+      "loss": 1.2371,
+      "step": 3788
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001880886436851975,
+      "loss": 1.2237,
+      "step": 3789
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018808126880523666,
+      "loss": 1.2545,
+      "step": 3790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018807389178759047,
+      "loss": 1.2605,
+      "step": 3791
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000188066512632438,
+      "loss": 1.4651,
+      "step": 3792
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018805913133995827,
+      "loss": 1.1482,
+      "step": 3793
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00018805174791033043,
+      "loss": 1.2742,
+      "step": 3794
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018804436234373374,
+      "loss": 1.3109,
+      "step": 3795
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018803697464034742,
+      "loss": 1.3061,
+      "step": 3796
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018802958480035068,
+      "loss": 1.1464,
+      "step": 3797
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000188022192823923,
+      "loss": 1.5181,
+      "step": 3798
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018801479871124367,
+      "loss": 1.3794,
+      "step": 3799
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001880074024624922,
+      "loss": 1.1863,
+      "step": 3800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018800000407784807,
+      "loss": 1.3914,
+      "step": 3801
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018799260355749088,
+      "loss": 1.3618,
+      "step": 3802
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018798520090160015,
+      "loss": 1.2311,
+      "step": 3803
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001879777961103556,
+      "loss": 1.3954,
+      "step": 3804
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018797038918393693,
+      "loss": 1.0887,
+      "step": 3805
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001879629801225239,
+      "loss": 2.4739,
+      "step": 3806
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018795556892629633,
+      "loss": 2.8247,
+      "step": 3807
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018794815559543406,
+      "loss": 1.3152,
+      "step": 3808
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018794074013011707,
+      "loss": 1.2267,
+      "step": 3809
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018793332253052525,
+      "loss": 1.3151,
+      "step": 3810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018792590279683867,
+      "loss": 1.1918,
+      "step": 3811
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018791848092923738,
+      "loss": 1.2617,
+      "step": 3812
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018791105692790154,
+      "loss": 1.1557,
+      "step": 3813
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018790363079301128,
+      "loss": 1.2604,
+      "step": 3814
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018789620252474684,
+      "loss": 2.4895,
+      "step": 3815
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018788877212328852,
+      "loss": 1.2602,
+      "step": 3816
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018788133958881666,
+      "loss": 1.2112,
+      "step": 3817
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018787390492151163,
+      "loss": 1.1943,
+      "step": 3818
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018786646812155383,
+      "loss": 1.2107,
+      "step": 3819
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001878590291891238,
+      "loss": 1.3518,
+      "step": 3820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018785158812440207,
+      "loss": 1.343,
+      "step": 3821
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018784414492756919,
+      "loss": 1.1709,
+      "step": 3822
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018783669959880584,
+      "loss": 1.2881,
+      "step": 3823
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001878292521382927,
+      "loss": 1.291,
+      "step": 3824
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018782180254621055,
+      "loss": 1.2346,
+      "step": 3825
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018781435082274012,
+      "loss": 1.0268,
+      "step": 3826
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001878068969680623,
+      "loss": 1.259,
+      "step": 3827
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018779944098235802,
+      "loss": 1.3338,
+      "step": 3828
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018779198286580817,
+      "loss": 1.3988,
+      "step": 3829
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018778452261859378,
+      "loss": 1.3054,
+      "step": 3830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018777706024089596,
+      "loss": 1.3351,
+      "step": 3831
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018776959573289572,
+      "loss": 1.4355,
+      "step": 3832
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001877621290947743,
+      "loss": 2.7556,
+      "step": 3833
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018775466032671287,
+      "loss": 1.3541,
+      "step": 3834
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001877471894288927,
+      "loss": 1.3577,
+      "step": 3835
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018773971640149512,
+      "loss": 1.2559,
+      "step": 3836
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018773224124470149,
+      "loss": 1.1335,
+      "step": 3837
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018772476395869322,
+      "loss": 1.2496,
+      "step": 3838
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018771728454365175,
+      "loss": 1.1233,
+      "step": 3839
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018770980299975866,
+      "loss": 1.2746,
+      "step": 3840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001877023193271955,
+      "loss": 1.4583,
+      "step": 3841
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018769483352614388,
+      "loss": 1.4196,
+      "step": 3842
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001876873455967855,
+      "loss": 1.0377,
+      "step": 3843
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018767985553930209,
+      "loss": 1.3605,
+      "step": 3844
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001876723633538754,
+      "loss": 1.1395,
+      "step": 3845
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018766486904068727,
+      "loss": 1.2369,
+      "step": 3846
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001876573725999196,
+      "loss": 1.4203,
+      "step": 3847
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018764987403175433,
+      "loss": 1.2378,
+      "step": 3848
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001876423733363734,
+      "loss": 1.3911,
+      "step": 3849
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001876348705139589,
+      "loss": 2.6499,
+      "step": 3850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018762736556469288,
+      "loss": 1.226,
+      "step": 3851
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018761985848875752,
+      "loss": 1.2758,
+      "step": 3852
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000187612349286335,
+      "loss": 1.3182,
+      "step": 3853
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018760483795760753,
+      "loss": 1.329,
+      "step": 3854
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018759732450275744,
+      "loss": 1.2058,
+      "step": 3855
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018758980892196706,
+      "loss": 1.0573,
+      "step": 3856
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018758229121541883,
+      "loss": 1.2261,
+      "step": 3857
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018757477138329516,
+      "loss": 1.1515,
+      "step": 3858
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018756724942577856,
+      "loss": 1.2419,
+      "step": 3859
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018755972534305158,
+      "loss": 1.2242,
+      "step": 3860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018755219913529686,
+      "loss": 1.3026,
+      "step": 3861
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018754467080269696,
+      "loss": 1.287,
+      "step": 3862
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018753714034543473,
+      "loss": 1.2426,
+      "step": 3863
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018752960776369283,
+      "loss": 1.3555,
+      "step": 3864
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018752207305765414,
+      "loss": 1.2577,
+      "step": 3865
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001875145362275014,
+      "loss": 1.1166,
+      "step": 3866
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018750699727341767,
+      "loss": 1.2262,
+      "step": 3867
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018749945619558585,
+      "loss": 1.1273,
+      "step": 3868
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001874919129941889,
+      "loss": 1.2584,
+      "step": 3869
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018748436766941002,
+      "loss": 2.7285,
+      "step": 3870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018747682022143228,
+      "loss": 1.3067,
+      "step": 3871
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018746927065043875,
+      "loss": 1.2814,
+      "step": 3872
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001874617189566128,
+      "loss": 1.477,
+      "step": 3873
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018745416514013763,
+      "loss": 1.4856,
+      "step": 3874
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001874466092011966,
+      "loss": 1.4928,
+      "step": 3875
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018743905113997303,
+      "loss": 1.1263,
+      "step": 3876
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018743149095665042,
+      "loss": 1.2492,
+      "step": 3877
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001874239286514122,
+      "loss": 1.277,
+      "step": 3878
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018741636422444198,
+      "loss": 1.217,
+      "step": 3879
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018740879767592322,
+      "loss": 1.2427,
+      "step": 3880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018740122900603968,
+      "loss": 1.4071,
+      "step": 3881
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018739365821497496,
+      "loss": 1.2506,
+      "step": 3882
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018738608530291284,
+      "loss": 1.5781,
+      "step": 3883
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001873785102700371,
+      "loss": 1.2199,
+      "step": 3884
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001873709331165316,
+      "loss": 1.3613,
+      "step": 3885
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018736335384258022,
+      "loss": 1.2899,
+      "step": 3886
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018735577244836685,
+      "loss": 1.3238,
+      "step": 3887
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001873481889340756,
+      "loss": 1.2292,
+      "step": 3888
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018734060329989044,
+      "loss": 1.1637,
+      "step": 3889
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018733301554599548,
+      "loss": 1.2644,
+      "step": 3890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001873254256725749,
+      "loss": 1.3753,
+      "step": 3891
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018731783367981284,
+      "loss": 1.4385,
+      "step": 3892
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018731023956789362,
+      "loss": 1.3317,
+      "step": 3893
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018730264333700154,
+      "loss": 1.2944,
+      "step": 3894
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001872950449873209,
+      "loss": 1.1934,
+      "step": 3895
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018728744451903615,
+      "loss": 1.3774,
+      "step": 3896
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00018727984193233173,
+      "loss": 1.245,
+      "step": 3897
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001872722372273922,
+      "loss": 1.3637,
+      "step": 3898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018726463040440202,
+      "loss": 1.1564,
+      "step": 3899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018725702146354595,
+      "loss": 1.2255,
+      "step": 3900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018724941040500852,
+      "loss": 1.2219,
+      "step": 3901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018724179722897453,
+      "loss": 1.15,
+      "step": 3902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001872341819356287,
+      "loss": 1.4843,
+      "step": 3903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018722656452515587,
+      "loss": 1.2598,
+      "step": 3904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018721894499774088,
+      "loss": 1.4706,
+      "step": 3905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018721132335356873,
+      "loss": 1.3468,
+      "step": 3906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001872036995928243,
+      "loss": 1.2176,
+      "step": 3907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001871960737156927,
+      "loss": 2.5435,
+      "step": 3908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018718844572235892,
+      "loss": 1.3108,
+      "step": 3909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018718081561300816,
+      "loss": 1.2659,
+      "step": 3910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018717318338782555,
+      "loss": 1.2762,
+      "step": 3911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018716554904699636,
+      "loss": 1.1414,
+      "step": 3912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001871579125907058,
+      "loss": 1.2047,
+      "step": 3913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018715027401913932,
+      "loss": 1.4135,
+      "step": 3914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001871426333324822,
+      "loss": 1.1992,
+      "step": 3915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001871349905309199,
+      "loss": 1.2045,
+      "step": 3916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018712734561463794,
+      "loss": 1.079,
+      "step": 3917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001871196985838218,
+      "loss": 1.0759,
+      "step": 3918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018711204943865715,
+      "loss": 1.3476,
+      "step": 3919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001871043981793296,
+      "loss": 1.5623,
+      "step": 3920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018709674480602478,
+      "loss": 1.32,
+      "step": 3921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001870890893189285,
+      "loss": 1.329,
+      "step": 3922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018708143171822656,
+      "loss": 1.3152,
+      "step": 3923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001870737720041048,
+      "loss": 1.2997,
+      "step": 3924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018706611017674905,
+      "loss": 2.7175,
+      "step": 3925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018705844623634533,
+      "loss": 1.0753,
+      "step": 3926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001870507801830796,
+      "loss": 1.1267,
+      "step": 3927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018704311201713796,
+      "loss": 1.2761,
+      "step": 3928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018703544173870643,
+      "loss": 1.4219,
+      "step": 3929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018702776934797127,
+      "loss": 1.5098,
+      "step": 3930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001870200948451186,
+      "loss": 1.3806,
+      "step": 3931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018701241823033474,
+      "loss": 1.1859,
+      "step": 3932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018700473950380593,
+      "loss": 1.2293,
+      "step": 3933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018699705866571855,
+      "loss": 1.105,
+      "step": 3934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018698937571625906,
+      "loss": 1.1272,
+      "step": 3935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018698169065561383,
+      "loss": 1.3618,
+      "step": 3936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001869740034839695,
+      "loss": 1.5037,
+      "step": 3937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018696631420151252,
+      "loss": 1.2535,
+      "step": 3938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018695862280842954,
+      "loss": 1.1482,
+      "step": 3939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018695092930490721,
+      "loss": 1.2825,
+      "step": 3940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001869432336911323,
+      "loss": 1.2342,
+      "step": 3941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018693553596729152,
+      "loss": 2.5557,
+      "step": 3942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018692783613357173,
+      "loss": 1.4771,
+      "step": 3943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018692013419015978,
+      "loss": 2.5105,
+      "step": 3944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001869124301372426,
+      "loss": 2.5625,
+      "step": 3945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018690472397500717,
+      "loss": 1.399,
+      "step": 3946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018689701570364048,
+      "loss": 1.3907,
+      "step": 3947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018688930532332967,
+      "loss": 1.3434,
+      "step": 3948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001868815928342618,
+      "loss": 1.3549,
+      "step": 3949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018687387823662407,
+      "loss": 1.2114,
+      "step": 3950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018686616153060372,
+      "loss": 1.3604,
+      "step": 3951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018685844271638805,
+      "loss": 1.2104,
+      "step": 3952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018685072179416436,
+      "loss": 1.2247,
+      "step": 3953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018684299876412002,
+      "loss": 1.2736,
+      "step": 3954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001868352736264425,
+      "loss": 1.3318,
+      "step": 3955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018682754638131926,
+      "loss": 1.2708,
+      "step": 3956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018681981702893782,
+      "loss": 1.1015,
+      "step": 3957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018681208556948584,
+      "loss": 1.2663,
+      "step": 3958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018680435200315088,
+      "loss": 1.186,
+      "step": 3959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018679661633012065,
+      "loss": 1.146,
+      "step": 3960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001867888785505829,
+      "loss": 1.2803,
+      "step": 3961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018678113866472547,
+      "loss": 1.3281,
+      "step": 3962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018677339667273613,
+      "loss": 1.4597,
+      "step": 3963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018676565257480277,
+      "loss": 1.2567,
+      "step": 3964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018675790637111338,
+      "loss": 1.3439,
+      "step": 3965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018675015806185593,
+      "loss": 1.3709,
+      "step": 3966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018674240764721852,
+      "loss": 2.9211,
+      "step": 3967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018673465512738916,
+      "loss": 1.3544,
+      "step": 3968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018672690050255608,
+      "loss": 1.1452,
+      "step": 3969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001867191437729074,
+      "loss": 1.113,
+      "step": 3970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018671138493863148,
+      "loss": 1.2415,
+      "step": 3971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018670362399991652,
+      "loss": 1.226,
+      "step": 3972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001866958609569509,
+      "loss": 1.3176,
+      "step": 3973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018668809580992308,
+      "loss": 1.146,
+      "step": 3974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018668032855902146,
+      "loss": 1.1676,
+      "step": 3975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018667255920443453,
+      "loss": 2.7637,
+      "step": 3976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018666478774635094,
+      "loss": 1.3723,
+      "step": 3977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001866570141849592,
+      "loss": 1.1577,
+      "step": 3978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018664923852044802,
+      "loss": 2.5559,
+      "step": 3979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001866414607530061,
+      "loss": 1.2314,
+      "step": 3980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001866336808828222,
+      "loss": 1.2146,
+      "step": 3981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018662589891008512,
+      "loss": 1.2109,
+      "step": 3982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001866181148349838,
+      "loss": 1.0527,
+      "step": 3983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018661032865770704,
+      "loss": 1.2882,
+      "step": 3984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018660254037844388,
+      "loss": 1.1744,
+      "step": 3985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001865947499973833,
+      "loss": 1.4803,
+      "step": 3986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018658695751471441,
+      "loss": 1.2975,
+      "step": 3987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001865791629306263,
+      "loss": 1.4342,
+      "step": 3988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018657136624530818,
+      "loss": 1.3636,
+      "step": 3989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001865635674589492,
+      "loss": 1.2378,
+      "step": 3990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018655576657173868,
+      "loss": 1.283,
+      "step": 3991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018654796358386593,
+      "loss": 1.3845,
+      "step": 3992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018654015849552032,
+      "loss": 1.1737,
+      "step": 3993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018653235130689134,
+      "loss": 1.2037,
+      "step": 3994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018652454201816836,
+      "loss": 1.1172,
+      "step": 3995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018651673062954096,
+      "loss": 1.2636,
+      "step": 3996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018650891714119873,
+      "loss": 2.8342,
+      "step": 3997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001865011015533313,
+      "loss": 1.4731,
+      "step": 3998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001864932838661283,
+      "loss": 1.3232,
+      "step": 3999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001864854640797795,
+      "loss": 1.2665,
+      "step": 4000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018647764219447473,
+      "loss": 2.5115,
+      "step": 4001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00018646981821040373,
+      "loss": 2.6438,
+      "step": 4002
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018646199212775643,
+      "loss": 1.2591,
+      "step": 4003
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018645416394672278,
+      "loss": 1.2867,
+      "step": 4004
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018644633366749272,
+      "loss": 2.416,
+      "step": 4005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018643850129025633,
+      "loss": 1.3334,
+      "step": 4006
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018643066681520368,
+      "loss": 1.0925,
+      "step": 4007
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001864228302425249,
+      "loss": 1.2931,
+      "step": 4008
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018641499157241018,
+      "loss": 1.2842,
+      "step": 4009
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018640715080504977,
+      "loss": 1.3093,
+      "step": 4010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000186399307940634,
+      "loss": 1.3707,
+      "step": 4011
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018639146297935315,
+      "loss": 1.3904,
+      "step": 4012
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018638361592139763,
+      "loss": 1.077,
+      "step": 4013
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001863757667669579,
+      "loss": 1.3517,
+      "step": 4014
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001863679155162244,
+      "loss": 1.2264,
+      "step": 4015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018636006216938777,
+      "loss": 2.6045,
+      "step": 4016
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018635220672663855,
+      "loss": 1.2523,
+      "step": 4017
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018634434918816738,
+      "loss": 1.3348,
+      "step": 4018
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018633648955416497,
+      "loss": 1.3873,
+      "step": 4019
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001863286278248221,
+      "loss": 1.3248,
+      "step": 4020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001863207640003295,
+      "loss": 1.3179,
+      "step": 4021
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018631289808087808,
+      "loss": 1.4031,
+      "step": 4022
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018630503006665872,
+      "loss": 1.1999,
+      "step": 4023
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001862971599578624,
+      "loss": 1.2087,
+      "step": 4024
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018628928775468007,
+      "loss": 1.3196,
+      "step": 4025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018628141345730282,
+      "loss": 1.2439,
+      "step": 4026
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018627353706592177,
+      "loss": 1.389,
+      "step": 4027
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018626565858072803,
+      "loss": 1.2671,
+      "step": 4028
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018625777800191284,
+      "loss": 2.3254,
+      "step": 4029
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018624989532966746,
+      "loss": 1.2672,
+      "step": 4030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018624201056418316,
+      "loss": 1.2594,
+      "step": 4031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018623412370565135,
+      "loss": 1.299,
+      "step": 4032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001862262347542634,
+      "loss": 1.2711,
+      "step": 4033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018621834371021078,
+      "loss": 1.3459,
+      "step": 4034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018621045057368502,
+      "loss": 1.3828,
+      "step": 4035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018620255534487767,
+      "loss": 1.3994,
+      "step": 4036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018619465802398034,
+      "loss": 1.2611,
+      "step": 4037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018618675861118472,
+      "loss": 1.2355,
+      "step": 4038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018617885710668248,
+      "loss": 1.2874,
+      "step": 4039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001861709535106654,
+      "loss": 1.0361,
+      "step": 4040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001861630478233253,
+      "loss": 1.2174,
+      "step": 4041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018615514004485407,
+      "loss": 1.1612,
+      "step": 4042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001861472301754436,
+      "loss": 1.4566,
+      "step": 4043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018613931821528587,
+      "loss": 1.1038,
+      "step": 4044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001861314041645729,
+      "loss": 1.3322,
+      "step": 4045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018612348802349673,
+      "loss": 1.2518,
+      "step": 4046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001861155697922495,
+      "loss": 1.2783,
+      "step": 4047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001861076494710234,
+      "loss": 1.1884,
+      "step": 4048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018609972706001065,
+      "loss": 1.1787,
+      "step": 4049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018609180255940347,
+      "loss": 2.5442,
+      "step": 4050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018608387596939427,
+      "loss": 1.356,
+      "step": 4051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018607594729017538,
+      "loss": 1.1703,
+      "step": 4052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018606801652193918,
+      "loss": 1.3442,
+      "step": 4053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001860600836648782,
+      "loss": 0.9997,
+      "step": 4054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000186052148719185,
+      "loss": 1.4649,
+      "step": 4055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001860442116850521,
+      "loss": 1.2196,
+      "step": 4056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018603627256267208,
+      "loss": 1.2222,
+      "step": 4057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018602833135223772,
+      "loss": 1.3712,
+      "step": 4058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001860203880539417,
+      "loss": 1.3433,
+      "step": 4059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018601244266797679,
+      "loss": 1.3076,
+      "step": 4060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018600449519453588,
+      "loss": 2.4939,
+      "step": 4061
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018599654563381177,
+      "loss": 1.247,
+      "step": 4062
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018598859398599743,
+      "loss": 1.2631,
+      "step": 4063
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018598064025128586,
+      "loss": 1.3202,
+      "step": 4064
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018597268442987005,
+      "loss": 1.305,
+      "step": 4065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018596472652194313,
+      "loss": 1.224,
+      "step": 4066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001859567665276982,
+      "loss": 1.2512,
+      "step": 4067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018594880444732846,
+      "loss": 1.2233,
+      "step": 4068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018594084028102713,
+      "loss": 1.2757,
+      "step": 4069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018593287402898754,
+      "loss": 1.3285,
+      "step": 4070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018592490569140297,
+      "loss": 1.4082,
+      "step": 4071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018591693526846684,
+      "loss": 1.2526,
+      "step": 4072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018590896276037255,
+      "loss": 1.2512,
+      "step": 4073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018590098816731365,
+      "loss": 1.5297,
+      "step": 4074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018589301148948364,
+      "loss": 1.2422,
+      "step": 4075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018588503272707612,
+      "loss": 1.2794,
+      "step": 4076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018587705188028472,
+      "loss": 1.3289,
+      "step": 4077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018586906894930313,
+      "loss": 1.3397,
+      "step": 4078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018586108393432512,
+      "loss": 1.3411,
+      "step": 4079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018585309683554445,
+      "loss": 1.288,
+      "step": 4080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018584510765315497,
+      "loss": 1.1324,
+      "step": 4081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001858371163873506,
+      "loss": 0.933,
+      "step": 4082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018582912303832525,
+      "loss": 1.3782,
+      "step": 4083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001858211276062729,
+      "loss": 1.3107,
+      "step": 4084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018581313009138765,
+      "loss": 1.2389,
+      "step": 4085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018580513049386353,
+      "loss": 1.3849,
+      "step": 4086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018579712881389477,
+      "loss": 1.3462,
+      "step": 4087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018578912505167548,
+      "loss": 1.258,
+      "step": 4088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001857811192074,
+      "loss": 1.2285,
+      "step": 4089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018577311128126252,
+      "loss": 1.3386,
+      "step": 4090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018576510127345746,
+      "loss": 1.3521,
+      "step": 4091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001857570891841792,
+      "loss": 1.2873,
+      "step": 4092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018574907501362219,
+      "loss": 1.1929,
+      "step": 4093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001857410587619809,
+      "loss": 1.2874,
+      "step": 4094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018573304042944995,
+      "loss": 1.2186,
+      "step": 4095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018572502001622387,
+      "loss": 1.414,
+      "step": 4096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018571699752249735,
+      "loss": 1.2784,
+      "step": 4097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018570897294846507,
+      "loss": 1.3108,
+      "step": 4098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018570094629432183,
+      "loss": 2.5276,
+      "step": 4099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018569291756026236,
+      "loss": 1.3339,
+      "step": 4100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001856848867464816,
+      "loss": 1.1533,
+      "step": 4101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018567685385317435,
+      "loss": 1.3866,
+      "step": 4102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018566881888053564,
+      "loss": 1.2246,
+      "step": 4103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018566078182876043,
+      "loss": 1.1376,
+      "step": 4104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018565274269804382,
+      "loss": 1.2492,
+      "step": 4105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018564470148858087,
+      "loss": 1.2744,
+      "step": 4106
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018563665820056679,
+      "loss": 1.1626,
+      "step": 4107
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001856286128341967,
+      "loss": 1.3871,
+      "step": 4108
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000185620565389666,
+      "loss": 1.2249,
+      "step": 4109
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018561251586716985,
+      "loss": 1.2194,
+      "step": 4110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018560446426690364,
+      "loss": 1.2971,
+      "step": 4111
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001855964105890629,
+      "loss": 1.4031,
+      "step": 4112
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001855883548338429,
+      "loss": 1.3341,
+      "step": 4113
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018558029700143925,
+      "loss": 2.5874,
+      "step": 4114
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018557223709204755,
+      "loss": 1.1747,
+      "step": 4115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018556417510586334,
+      "loss": 1.3431,
+      "step": 4116
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018555611104308228,
+      "loss": 1.2605,
+      "step": 4117
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018554804490390013,
+      "loss": 2.4783,
+      "step": 4118
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001855399766885126,
+      "loss": 1.3765,
+      "step": 4119
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018553190639711552,
+      "loss": 1.4837,
+      "step": 4120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018552383402990476,
+      "loss": 1.2274,
+      "step": 4121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018551575958707624,
+      "loss": 1.2521,
+      "step": 4122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018550768306882588,
+      "loss": 1.2407,
+      "step": 4123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018549960447534976,
+      "loss": 2.4343,
+      "step": 4124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001854915238068439,
+      "loss": 1.1823,
+      "step": 4125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018548344106350436,
+      "loss": 1.2168,
+      "step": 4126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018547535624552743,
+      "loss": 1.2314,
+      "step": 4127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018546726935310924,
+      "loss": 1.1833,
+      "step": 4128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018545918038644606,
+      "loss": 1.1666,
+      "step": 4129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018545108934573423,
+      "loss": 1.2972,
+      "step": 4130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018544299623117009,
+      "loss": 1.3588,
+      "step": 4131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018543490104295008,
+      "loss": 1.2892,
+      "step": 4132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001854268037812706,
+      "loss": 1.2213,
+      "step": 4133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001854187044463283,
+      "loss": 1.2045,
+      "step": 4134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018541060303831964,
+      "loss": 1.1612,
+      "step": 4135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018540249955744122,
+      "loss": 1.1812,
+      "step": 4136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018539439400388978,
+      "loss": 1.3762,
+      "step": 4137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018538628637786202,
+      "loss": 1.2857,
+      "step": 4138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018537817667955468,
+      "loss": 1.4405,
+      "step": 4139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001853700649091646,
+      "loss": 1.3516,
+      "step": 4140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018536195106688862,
+      "loss": 1.4111,
+      "step": 4141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018535383515292368,
+      "loss": 1.2545,
+      "step": 4142
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001853457171674668,
+      "loss": 1.3848,
+      "step": 4143
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001853375971107149,
+      "loss": 2.6907,
+      "step": 4144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001853294749828651,
+      "loss": 1.4055,
+      "step": 4145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018532135078411455,
+      "loss": 1.3138,
+      "step": 4146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018531322451466035,
+      "loss": 1.2733,
+      "step": 4147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018530509617469977,
+      "loss": 1.2901,
+      "step": 4148
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018529696576443007,
+      "loss": 1.3929,
+      "step": 4149
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018528883328404857,
+      "loss": 1.2039,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018528069873375262,
+      "loss": 2.3938,
+      "step": 4151
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001852725621137397,
+      "loss": 1.226,
+      "step": 4152
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001852644234242072,
+      "loss": 1.2478,
+      "step": 4153
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001852562826653527,
+      "loss": 1.1875,
+      "step": 4154
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018524813983737378,
+      "loss": 1.4359,
+      "step": 4155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000185239994940468,
+      "loss": 2.5811,
+      "step": 4156
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001852318479748331,
+      "loss": 1.2686,
+      "step": 4157
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018522369894066672,
+      "loss": 1.3112,
+      "step": 4158
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018521554783816675,
+      "loss": 1.2646,
+      "step": 4159
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018520739466753092,
+      "loss": 1.3535,
+      "step": 4160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018519923942895714,
+      "loss": 1.2544,
+      "step": 4161
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001851910821226433,
+      "loss": 1.2259,
+      "step": 4162
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018518292274878742,
+      "loss": 1.2578,
+      "step": 4163
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018517476130758747,
+      "loss": 1.1213,
+      "step": 4164
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018516659779924162,
+      "loss": 1.3482,
+      "step": 4165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018515843222394787,
+      "loss": 1.2438,
+      "step": 4166
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018515026458190448,
+      "loss": 1.2893,
+      "step": 4167
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018514209487330963,
+      "loss": 1.3737,
+      "step": 4168
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018513392309836162,
+      "loss": 1.3037,
+      "step": 4169
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018512574925725876,
+      "loss": 1.2536,
+      "step": 4170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018511757335019945,
+      "loss": 1.3012,
+      "step": 4171
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001851093953773821,
+      "loss": 2.5405,
+      "step": 4172
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018510121533900516,
+      "loss": 1.2423,
+      "step": 4173
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018509303323526718,
+      "loss": 1.3466,
+      "step": 4174
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018508484906636677,
+      "loss": 1.3778,
+      "step": 4175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018507666283250246,
+      "loss": 1.4573,
+      "step": 4176
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000185068474533873,
+      "loss": 1.0665,
+      "step": 4177
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001850602841706771,
+      "loss": 1.2966,
+      "step": 4178
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018505209174311354,
+      "loss": 1.3697,
+      "step": 4179
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018504389725138115,
+      "loss": 1.2497,
+      "step": 4180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018503570069567879,
+      "loss": 1.3143,
+      "step": 4181
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001850275020762054,
+      "loss": 1.4177,
+      "step": 4182
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018501930139315994,
+      "loss": 1.2542,
+      "step": 4183
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018501109864674145,
+      "loss": 2.5339,
+      "step": 4184
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018500289383714897,
+      "loss": 1.4043,
+      "step": 4185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001849946869645817,
+      "loss": 1.2858,
+      "step": 4186
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018498647802923875,
+      "loss": 1.1939,
+      "step": 4187
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018497826703131937,
+      "loss": 1.2531,
+      "step": 4188
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018497005397102284,
+      "loss": 1.371,
+      "step": 4189
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018496183884854846,
+      "loss": 2.5405,
+      "step": 4190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018495362166409568,
+      "loss": 1.241,
+      "step": 4191
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018494540241786384,
+      "loss": 1.2259,
+      "step": 4192
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018493718111005247,
+      "loss": 1.15,
+      "step": 4193
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018492895774086104,
+      "loss": 1.1798,
+      "step": 4194
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001849207323104892,
+      "loss": 1.17,
+      "step": 4195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018491250481913655,
+      "loss": 1.412,
+      "step": 4196
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018490427526700274,
+      "loss": 1.1539,
+      "step": 4197
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018489604365428755,
+      "loss": 1.2796,
+      "step": 4198
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001848878099811907,
+      "loss": 1.3322,
+      "step": 4199
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018487957424791204,
+      "loss": 1.2066,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018487133645465144,
+      "loss": 1.4061,
+      "step": 4201
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018486309660160886,
+      "loss": 1.3058,
+      "step": 4202
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018485485468898422,
+      "loss": 1.2216,
+      "step": 4203
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001848466107169776,
+      "loss": 1.2551,
+      "step": 4204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018483836468578905,
+      "loss": 1.2997,
+      "step": 4205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018483011659561872,
+      "loss": 1.3686,
+      "step": 4206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018482186644666672,
+      "loss": 1.194,
+      "step": 4207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018481361423913336,
+      "loss": 1.2226,
+      "step": 4208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001848053599732189,
+      "loss": 1.2361,
+      "step": 4209
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018479710364912363,
+      "loss": 1.0687,
+      "step": 4210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018478884526704792,
+      "loss": 1.2018,
+      "step": 4211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018478058482719224,
+      "loss": 1.2793,
+      "step": 4212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018477232232975705,
+      "loss": 1.2654,
+      "step": 4213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018476405777494287,
+      "loss": 1.1835,
+      "step": 4214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018475579116295027,
+      "loss": 1.2671,
+      "step": 4215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001847475224939799,
+      "loss": 1.2313,
+      "step": 4216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018473925176823244,
+      "loss": 1.3718,
+      "step": 4217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001847309789859086,
+      "loss": 1.1393,
+      "step": 4218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001847227041472091,
+      "loss": 1.2064,
+      "step": 4219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018471442725233488,
+      "loss": 2.6389,
+      "step": 4220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018470614830148674,
+      "loss": 1.4277,
+      "step": 4221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001846978672948656,
+      "loss": 1.2975,
+      "step": 4222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018468958423267247,
+      "loss": 1.3857,
+      "step": 4223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018468129911510838,
+      "loss": 1.4036,
+      "step": 4224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001846730119423744,
+      "loss": 1.2344,
+      "step": 4225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001846647227146716,
+      "loss": 1.226,
+      "step": 4226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018465643143220125,
+      "loss": 1.3797,
+      "step": 4227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001846481380951645,
+      "loss": 1.301,
+      "step": 4228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018463984270376265,
+      "loss": 1.132,
+      "step": 4229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018463154525819698,
+      "loss": 1.2822,
+      "step": 4230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018462324575866897,
+      "loss": 1.1126,
+      "step": 4231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018461494420537997,
+      "loss": 1.312,
+      "step": 4232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018460664059853144,
+      "loss": 1.403,
+      "step": 4233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018459833493832497,
+      "loss": 1.3812,
+      "step": 4234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018459002722496208,
+      "loss": 1.1606,
+      "step": 4235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001845817174586444,
+      "loss": 1.4174,
+      "step": 4236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018457340563957363,
+      "loss": 1.1412,
+      "step": 4237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018456509176795148,
+      "loss": 1.218,
+      "step": 4238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001845567758439797,
+      "loss": 1.3826,
+      "step": 4239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018454845786786012,
+      "loss": 1.2023,
+      "step": 4240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018454013783979464,
+      "loss": 1.1841,
+      "step": 4241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018453181575998518,
+      "loss": 1.39,
+      "step": 4242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018452349162863369,
+      "loss": 1.2634,
+      "step": 4243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018451516544594222,
+      "loss": 1.2979,
+      "step": 4244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018450683721211279,
+      "loss": 1.2151,
+      "step": 4245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018449850692734757,
+      "loss": 2.4128,
+      "step": 4246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018449017459184873,
+      "loss": 1.5251,
+      "step": 4247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018448184020581848,
+      "loss": 1.3586,
+      "step": 4248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018447350376945906,
+      "loss": 1.3263,
+      "step": 4249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018446516528297286,
+      "loss": 1.0996,
+      "step": 4250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018445682474656222,
+      "loss": 1.2172,
+      "step": 4251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018444848216042948,
+      "loss": 2.5315,
+      "step": 4252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018444013752477725,
+      "loss": 1.3136,
+      "step": 4253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018443179083980798,
+      "loss": 1.3054,
+      "step": 4254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018442344210572423,
+      "loss": 1.2305,
+      "step": 4255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001844150913227286,
+      "loss": 1.1816,
+      "step": 4256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018440673849102384,
+      "loss": 1.2614,
+      "step": 4257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018439838361081255,
+      "loss": 1.1998,
+      "step": 4258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018439002668229764,
+      "loss": 1.1538,
+      "step": 4259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018438166770568182,
+      "loss": 1.2433,
+      "step": 4260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000184373306681168,
+      "loss": 1.0963,
+      "step": 4261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018436494360895908,
+      "loss": 1.1497,
+      "step": 4262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018435657848925806,
+      "loss": 1.1855,
+      "step": 4263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018434821132226793,
+      "loss": 1.2273,
+      "step": 4264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018433984210819174,
+      "loss": 1.1882,
+      "step": 4265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018433147084723266,
+      "loss": 1.5126,
+      "step": 4266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001843230975395938,
+      "loss": 1.3416,
+      "step": 4267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001843147221854784,
+      "loss": 1.2293,
+      "step": 4268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018430634478508974,
+      "loss": 1.2808,
+      "step": 4269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001842979653386311,
+      "loss": 1.2088,
+      "step": 4270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001842895838463059,
+      "loss": 1.2748,
+      "step": 4271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001842812003083175,
+      "loss": 1.4194,
+      "step": 4272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018427281472486938,
+      "loss": 1.1697,
+      "step": 4273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018426442709616506,
+      "loss": 1.2019,
+      "step": 4274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001842560374224081,
+      "loss": 1.2565,
+      "step": 4275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018424764570380213,
+      "loss": 1.2935,
+      "step": 4276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018423925194055077,
+      "loss": 1.4125,
+      "step": 4277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018423085613285777,
+      "loss": 1.2834,
+      "step": 4278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018422245828092688,
+      "loss": 1.2039,
+      "step": 4279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001842140583849619,
+      "loss": 1.0606,
+      "step": 4280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018420565644516673,
+      "loss": 1.3246,
+      "step": 4281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001841972524617452,
+      "loss": 2.4014,
+      "step": 4282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018418884643490137,
+      "loss": 1.2247,
+      "step": 4283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001841804383648392,
+      "loss": 1.131,
+      "step": 4284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018417202825176274,
+      "loss": 1.2474,
+      "step": 4285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001841636160958761,
+      "loss": 1.2341,
+      "step": 4286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018415520189738347,
+      "loss": 1.3026,
+      "step": 4287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018414678565648905,
+      "loss": 1.4667,
+      "step": 4288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018413836737339703,
+      "loss": 1.0367,
+      "step": 4289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018412994704831181,
+      "loss": 1.3206,
+      "step": 4290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001841215246814377,
+      "loss": 1.1805,
+      "step": 4291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018411310027297915,
+      "loss": 1.3097,
+      "step": 4292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018410467382314055,
+      "loss": 1.4247,
+      "step": 4293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018409624533212643,
+      "loss": 1.076,
+      "step": 4294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018408781480014138,
+      "loss": 1.2116,
+      "step": 4295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018407938222738993,
+      "loss": 1.1289,
+      "step": 4296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018407094761407682,
+      "loss": 1.5832,
+      "step": 4297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001840625109604067,
+      "loss": 1.2258,
+      "step": 4298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018405407226658436,
+      "loss": 1.26,
+      "step": 4299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018404563153281455,
+      "loss": 1.4301,
+      "step": 4300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018403718875930215,
+      "loss": 2.551,
+      "step": 4301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001840287439462521,
+      "loss": 1.3353,
+      "step": 4302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018402029709386927,
+      "loss": 1.1371,
+      "step": 4303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018401184820235873,
+      "loss": 1.4072,
+      "step": 4304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001840033972719255,
+      "loss": 1.3607,
+      "step": 4305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001839949443027747,
+      "loss": 1.3036,
+      "step": 4306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018398648929511145,
+      "loss": 1.0854,
+      "step": 4307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018397803224914095,
+      "loss": 1.3599,
+      "step": 4308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001839695731650685,
+      "loss": 1.3088,
+      "step": 4309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018396111204309928,
+      "loss": 1.347,
+      "step": 4310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001839526488834388,
+      "loss": 1.3509,
+      "step": 4311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018394418368629235,
+      "loss": 1.2851,
+      "step": 4312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018393571645186537,
+      "loss": 1.3578,
+      "step": 4313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018392724718036338,
+      "loss": 1.3728,
+      "step": 4314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018391877587199194,
+      "loss": 1.2886,
+      "step": 4315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018391030252695666,
+      "loss": 1.4025,
+      "step": 4316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001839018271454631,
+      "loss": 1.1333,
+      "step": 4317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018389334972771702,
+      "loss": 1.2881,
+      "step": 4318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001838848702739242,
+      "loss": 1.2228,
+      "step": 4319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018387638878429035,
+      "loss": 1.3888,
+      "step": 4320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001838679052590213,
+      "loss": 2.5352,
+      "step": 4321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018385941969832303,
+      "loss": 1.2513,
+      "step": 4322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018385093210240142,
+      "loss": 1.1915,
+      "step": 4323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018384244247146252,
+      "loss": 1.2672,
+      "step": 4324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018383395080571227,
+      "loss": 1.1829,
+      "step": 4325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001838254571053568,
+      "loss": 1.2729,
+      "step": 4326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001838169613706023,
+      "loss": 1.523,
+      "step": 4327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018380846360165488,
+      "loss": 1.3571,
+      "step": 4328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018379996379872085,
+      "loss": 1.239,
+      "step": 4329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018379146196200642,
+      "loss": 2.4927,
+      "step": 4330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018378295809171798,
+      "loss": 1.2684,
+      "step": 4331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018377445218806186,
+      "loss": 1.2861,
+      "step": 4332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001837659442512446,
+      "loss": 1.2459,
+      "step": 4333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018375743428147257,
+      "loss": 1.2969,
+      "step": 4334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001837489222789523,
+      "loss": 1.2951,
+      "step": 4335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001837404082438905,
+      "loss": 1.1468,
+      "step": 4336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018373189217649367,
+      "loss": 1.2237,
+      "step": 4337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018372337407696857,
+      "loss": 1.3165,
+      "step": 4338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018371485394552185,
+      "loss": 1.2271,
+      "step": 4339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018370633178236035,
+      "loss": 1.2606,
+      "step": 4340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001836978075876909,
+      "loss": 1.1774,
+      "step": 4341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001836892813617204,
+      "loss": 1.315,
+      "step": 4342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018368075310465568,
+      "loss": 1.31,
+      "step": 4343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018367222281670382,
+      "loss": 1.2394,
+      "step": 4344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018366369049807176,
+      "loss": 1.1923,
+      "step": 4345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018365515614896666,
+      "loss": 1.1838,
+      "step": 4346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018364661976959557,
+      "loss": 1.3078,
+      "step": 4347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001836380813601657,
+      "loss": 1.2725,
+      "step": 4348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018362954092088427,
+      "loss": 1.2204,
+      "step": 4349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018362099845195857,
+      "loss": 1.1826,
+      "step": 4350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018361245395359584,
+      "loss": 1.2391,
+      "step": 4351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001836039074260036,
+      "loss": 1.1932,
+      "step": 4352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001835953588693891,
+      "loss": 1.4272,
+      "step": 4353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018358680828395993,
+      "loss": 2.7292,
+      "step": 4354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018357825566992357,
+      "loss": 1.0781,
+      "step": 4355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018356970102748757,
+      "loss": 1.149,
+      "step": 4356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018356114435685953,
+      "loss": 1.2701,
+      "step": 4357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001835525856582472,
+      "loss": 2.7791,
+      "step": 4358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001835440249318582,
+      "loss": 1.138,
+      "step": 4359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018353546217790036,
+      "loss": 1.2335,
+      "step": 4360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001835268973965815,
+      "loss": 1.1874,
+      "step": 4361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001835183305881094,
+      "loss": 1.1532,
+      "step": 4362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018350976175269206,
+      "loss": 2.4641,
+      "step": 4363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018350119089053738,
+      "loss": 1.3212,
+      "step": 4364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018349261800185338,
+      "loss": 1.3629,
+      "step": 4365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018348404308684817,
+      "loss": 1.2821,
+      "step": 4366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018347546614572982,
+      "loss": 1.2319,
+      "step": 4367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018346688717870647,
+      "loss": 1.2434,
+      "step": 4368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018345830618598637,
+      "loss": 1.023,
+      "step": 4369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018344972316777772,
+      "loss": 1.2613,
+      "step": 4370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018344113812428892,
+      "loss": 1.2169,
+      "step": 4371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001834325510557282,
+      "loss": 1.0514,
+      "step": 4372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018342396196230408,
+      "loss": 1.2698,
+      "step": 4373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018341537084422496,
+      "loss": 2.4187,
+      "step": 4374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001834067777016993,
+      "loss": 1.2617,
+      "step": 4375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018339818253493572,
+      "loss": 1.1643,
+      "step": 4376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018338958534414276,
+      "loss": 1.2147,
+      "step": 4377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018338098612952913,
+      "loss": 1.3048,
+      "step": 4378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001833723848913035,
+      "loss": 1.1597,
+      "step": 4379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001833637816296746,
+      "loss": 1.2036,
+      "step": 4380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018335517634485127,
+      "loss": 1.1502,
+      "step": 4381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018334656903704234,
+      "loss": 1.039,
+      "step": 4382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018333795970645666,
+      "loss": 1.1447,
+      "step": 4383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018332934835330324,
+      "loss": 1.2407,
+      "step": 4384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018332073497779106,
+      "loss": 1.162,
+      "step": 4385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001833121195801291,
+      "loss": 1.1432,
+      "step": 4386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001833035021605265,
+      "loss": 1.2517,
+      "step": 4387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001832948827191924,
+      "loss": 1.3191,
+      "step": 4388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000183286261256336,
+      "loss": 1.2487,
+      "step": 4389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001832776377721665,
+      "loss": 1.2595,
+      "step": 4390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001832690122668932,
+      "loss": 1.1155,
+      "step": 4391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018326038474072548,
+      "loss": 1.3795,
+      "step": 4392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018325175519387269,
+      "loss": 1.3608,
+      "step": 4393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018324312362654426,
+      "loss": 1.1874,
+      "step": 4394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001832344900389497,
+      "loss": 1.2847,
+      "step": 4395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018322585443129847,
+      "loss": 1.1404,
+      "step": 4396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018321721680380023,
+      "loss": 1.3137,
+      "step": 4397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001832085771566646,
+      "loss": 1.3038,
+      "step": 4398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018319993549010123,
+      "loss": 1.2249,
+      "step": 4399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018319129180431986,
+      "loss": 1.4332,
+      "step": 4400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001831826460995303,
+      "loss": 1.3129,
+      "step": 4401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018317399837594232,
+      "loss": 1.2675,
+      "step": 4402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001831653486337658,
+      "loss": 1.2877,
+      "step": 4403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018315669687321071,
+      "loss": 1.2549,
+      "step": 4404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018314804309448702,
+      "loss": 1.2213,
+      "step": 4405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018313938729780472,
+      "loss": 1.454,
+      "step": 4406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001831307294833739,
+      "loss": 1.1964,
+      "step": 4407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018312206965140465,
+      "loss": 1.2776,
+      "step": 4408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001831134078021072,
+      "loss": 1.356,
+      "step": 4409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001831047439356917,
+      "loss": 1.4883,
+      "step": 4410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001830960780523685,
+      "loss": 1.2471,
+      "step": 4411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018308741015234784,
+      "loss": 1.3311,
+      "step": 4412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001830787402358401,
+      "loss": 1.2117,
+      "step": 4413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018307006830305572,
+      "loss": 1.2524,
+      "step": 4414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018306139435420518,
+      "loss": 1.413,
+      "step": 4415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001830527183894989,
+      "loss": 1.289,
+      "step": 4416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018304404040914754,
+      "loss": 1.2907,
+      "step": 4417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018303536041336166,
+      "loss": 1.4526,
+      "step": 4418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018302667840235196,
+      "loss": 1.4261,
+      "step": 4419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001830179943763291,
+      "loss": 1.3736,
+      "step": 4420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018300930833550387,
+      "loss": 1.083,
+      "step": 4421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018300062028008704,
+      "loss": 1.2066,
+      "step": 4422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018299193021028952,
+      "loss": 1.156,
+      "step": 4423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018298323812632213,
+      "loss": 1.1242,
+      "step": 4424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001829745440283959,
+      "loss": 1.3016,
+      "step": 4425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018296584791672183,
+      "loss": 2.6648,
+      "step": 4426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018295714979151088,
+      "loss": 1.1506,
+      "step": 4427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018294844965297428,
+      "loss": 1.3646,
+      "step": 4428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018293974750132306,
+      "loss": 1.1882,
+      "step": 4429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001829310433367685,
+      "loss": 1.2217,
+      "step": 4430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001829223371595218,
+      "loss": 1.29,
+      "step": 4431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018291362896979426,
+      "loss": 1.182,
+      "step": 4432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018290491876779723,
+      "loss": 1.315,
+      "step": 4433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018289620655374212,
+      "loss": 1.1449,
+      "step": 4434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018288749232784032,
+      "loss": 2.531,
+      "step": 4435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018287877609030338,
+      "loss": 1.0788,
+      "step": 4436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828700578413428,
+      "loss": 1.2259,
+      "step": 4437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018286133758117017,
+      "loss": 1.2522,
+      "step": 4438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018285261530999716,
+      "loss": 1.1808,
+      "step": 4439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828438910280354,
+      "loss": 1.2521,
+      "step": 4440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018283516473549665,
+      "loss": 1.2089,
+      "step": 4441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018282643643259273,
+      "loss": 1.3307,
+      "step": 4442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828177061195354,
+      "loss": 1.2312,
+      "step": 4443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018280897379653656,
+      "loss": 1.2526,
+      "step": 4444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828002394638082,
+      "loss": 1.1131,
+      "step": 4445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018279150312156223,
+      "loss": 1.346,
+      "step": 4446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827827647700107,
+      "loss": 1.4158,
+      "step": 4447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827740244093657,
+      "loss": 1.3699,
+      "step": 4448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827652820398393,
+      "loss": 1.1556,
+      "step": 4449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018275653766164376,
+      "loss": 2.6021,
+      "step": 4450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018274779127499123,
+      "loss": 1.2305,
+      "step": 4451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000182739042880094,
+      "loss": 1.1214,
+      "step": 4452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018273029247716438,
+      "loss": 1.2025,
+      "step": 4453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018272154006641474,
+      "loss": 1.2644,
+      "step": 4454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018271278564805754,
+      "loss": 1.3535,
+      "step": 4455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827040292223052,
+      "loss": 1.3598,
+      "step": 4456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018269527078937023,
+      "loss": 1.3152,
+      "step": 4457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826865103494652,
+      "loss": 1.1967,
+      "step": 4458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018267774790280274,
+      "loss": 1.3006,
+      "step": 4459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018266898344959548,
+      "loss": 1.205,
+      "step": 4460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018266021699005615,
+      "loss": 1.2114,
+      "step": 4461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018265144852439754,
+      "loss": 1.119,
+      "step": 4462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826426780528324,
+      "loss": 1.188,
+      "step": 4463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018263390557557357,
+      "loss": 1.3618,
+      "step": 4464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018262513109283402,
+      "loss": 2.7585,
+      "step": 4465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018261635460482668,
+      "loss": 1.2761,
+      "step": 4466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826075761117645,
+      "loss": 1.1613,
+      "step": 4467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018259879561386056,
+      "loss": 1.3148,
+      "step": 4468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018259001311132802,
+      "loss": 2.3745,
+      "step": 4469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018258122860437994,
+      "loss": 1.3237,
+      "step": 4470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018257244209322958,
+      "loss": 1.124,
+      "step": 4471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018256365357809012,
+      "loss": 1.1706,
+      "step": 4472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001825548630591749,
+      "loss": 1.2423,
+      "step": 4473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018254607053669726,
+      "loss": 1.3156,
+      "step": 4474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018253727601087052,
+      "loss": 1.2181,
+      "step": 4475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018252847948190822,
+      "loss": 1.3042,
+      "step": 4476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018251968095002382,
+      "loss": 1.0798,
+      "step": 4477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018251088041543078,
+      "loss": 1.1875,
+      "step": 4478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001825020778783428,
+      "loss": 0.9409,
+      "step": 4479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001824932733389734,
+      "loss": 1.4928,
+      "step": 4480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018248446679753637,
+      "loss": 1.4161,
+      "step": 4481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018247565825424539,
+      "loss": 1.1631,
+      "step": 4482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018246684770931417,
+      "loss": 1.2819,
+      "step": 4483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018245803516295665,
+      "loss": 1.1653,
+      "step": 4484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018244922061538664,
+      "loss": 1.2087,
+      "step": 4485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018244040406681809,
+      "loss": 1.1195,
+      "step": 4486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018243158551746496,
+      "loss": 1.2346,
+      "step": 4487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018242276496754128,
+      "loss": 1.086,
+      "step": 4488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001824139424172611,
+      "loss": 1.0855,
+      "step": 4489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001824051178668386,
+      "loss": 2.688,
+      "step": 4490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018239629131648787,
+      "loss": 1.2904,
+      "step": 4491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018238746276642316,
+      "loss": 1.149,
+      "step": 4492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018237863221685872,
+      "loss": 1.2461,
+      "step": 4493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018236979966800888,
+      "loss": 1.3106,
+      "step": 4494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018236096512008802,
+      "loss": 1.4755,
+      "step": 4495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018235212857331052,
+      "loss": 1.1265,
+      "step": 4496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018234329002789082,
+      "loss": 1.2126,
+      "step": 4497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018233444948404347,
+      "loss": 1.2197,
+      "step": 4498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018232560694198303,
+      "loss": 1.0753,
+      "step": 4499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000182316762401924,
+      "loss": 1.3884,
+      "step": 4500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823079158640812,
+      "loss": 2.572,
+      "step": 4501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822990673286692,
+      "loss": 1.3453,
+      "step": 4502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822902167959028,
+      "loss": 1.252,
+      "step": 4503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822813642659968,
+      "loss": 1.3026,
+      "step": 4504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000182272509739166,
+      "loss": 1.176,
+      "step": 4505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018226365321562536,
+      "loss": 1.1281,
+      "step": 4506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822547946955898,
+      "loss": 1.2857,
+      "step": 4507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822459341792743,
+      "loss": 2.5483,
+      "step": 4508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822370716668939,
+      "loss": 1.2838,
+      "step": 4509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822282071586637,
+      "loss": 1.0998,
+      "step": 4510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018221934065479881,
+      "loss": 1.2946,
+      "step": 4511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018221047215551445,
+      "loss": 1.2011,
+      "step": 4512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018220160166102585,
+      "loss": 1.152,
+      "step": 4513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018219272917154825,
+      "loss": 1.1954,
+      "step": 4514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018218385468729705,
+      "loss": 1.2819,
+      "step": 4515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018217497820848758,
+      "loss": 1.3094,
+      "step": 4516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001821660997353353,
+      "loss": 1.3314,
+      "step": 4517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018215721926805563,
+      "loss": 1.3315,
+      "step": 4518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018214833680686414,
+      "loss": 1.2444,
+      "step": 4519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018213945235197638,
+      "loss": 1.1904,
+      "step": 4520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018213056590360798,
+      "loss": 1.2972,
+      "step": 4521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018212167746197463,
+      "loss": 1.2363,
+      "step": 4522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018211278702729203,
+      "loss": 1.2443,
+      "step": 4523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018210389459977596,
+      "loss": 1.3085,
+      "step": 4524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001820950001796422,
+      "loss": 1.3084,
+      "step": 4525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018208610376710662,
+      "loss": 1.123,
+      "step": 4526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018207720536238516,
+      "loss": 1.1943,
+      "step": 4527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018206830496569375,
+      "loss": 1.2779,
+      "step": 4528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018205940257724846,
+      "loss": 1.3208,
+      "step": 4529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018205049819726524,
+      "loss": 1.4536,
+      "step": 4530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018204159182596028,
+      "loss": 1.2581,
+      "step": 4531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001820326834635497,
+      "loss": 1.1517,
+      "step": 4532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001820237731102497,
+      "loss": 1.3843,
+      "step": 4533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018201486076627654,
+      "loss": 2.4417,
+      "step": 4534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018200594643184653,
+      "loss": 1.2386,
+      "step": 4535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018199703010717596,
+      "loss": 1.1363,
+      "step": 4536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018198811179248128,
+      "loss": 1.275,
+      "step": 4537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018197919148797893,
+      "loss": 1.0395,
+      "step": 4538
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001819702691938854,
+      "loss": 1.2721,
+      "step": 4539
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018196134491041722,
+      "loss": 1.3954,
+      "step": 4540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018195241863779095,
+      "loss": 1.2928,
+      "step": 4541
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018194349037622325,
+      "loss": 1.455,
+      "step": 4542
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001819345601259308,
+      "loss": 1.2484,
+      "step": 4543
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018192562788713035,
+      "loss": 1.3705,
+      "step": 4544
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018191669366003868,
+      "loss": 1.3204,
+      "step": 4545
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001819077574448726,
+      "loss": 1.3166,
+      "step": 4546
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018189881924184897,
+      "loss": 1.2061,
+      "step": 4547
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018188987905118475,
+      "loss": 1.2529,
+      "step": 4548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018188093687309694,
+      "loss": 1.2398,
+      "step": 4549
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001818719927078025,
+      "loss": 1.063,
+      "step": 4550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001818630465555185,
+      "loss": 2.6831,
+      "step": 4551
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001818540984164621,
+      "loss": 1.2376,
+      "step": 4552
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018184514829085044,
+      "loss": 1.1699,
+      "step": 4553
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001818361961789008,
+      "loss": 1.0948,
+      "step": 4554
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018182724208083034,
+      "loss": 2.5181,
+      "step": 4555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018181828599685642,
+      "loss": 1.1147,
+      "step": 4556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018180932792719642,
+      "loss": 1.2859,
+      "step": 4557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001818003678720677,
+      "loss": 1.3975,
+      "step": 4558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018179140583168775,
+      "loss": 1.1643,
+      "step": 4559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018178244180627406,
+      "loss": 1.3746,
+      "step": 4560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018177347579604417,
+      "loss": 1.0405,
+      "step": 4561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018176450780121574,
+      "loss": 1.3315,
+      "step": 4562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018175553782200638,
+      "loss": 2.4705,
+      "step": 4563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018174656585863373,
+      "loss": 1.1962,
+      "step": 4564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018173759191131565,
+      "loss": 1.3934,
+      "step": 4565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018172861598026984,
+      "loss": 1.0374,
+      "step": 4566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018171963806571417,
+      "loss": 1.2007,
+      "step": 4567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018171065816786653,
+      "loss": 1.295,
+      "step": 4568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001817016762869449,
+      "loss": 1.2253,
+      "step": 4569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018169269242316718,
+      "loss": 0.9804,
+      "step": 4570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018168370657675143,
+      "loss": 1.4005,
+      "step": 4571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018167471874791578,
+      "loss": 1.2893,
+      "step": 4572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018166572893687832,
+      "loss": 1.2053,
+      "step": 4573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018165673714385725,
+      "loss": 1.2499,
+      "step": 4574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018164774336907076,
+      "loss": 1.3019,
+      "step": 4575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018163874761273715,
+      "loss": 1.2495,
+      "step": 4576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018162974987507476,
+      "loss": 1.1315,
+      "step": 4577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018162075015630192,
+      "loss": 1.1604,
+      "step": 4578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001816117484566371,
+      "loss": 1.2328,
+      "step": 4579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018160274477629869,
+      "loss": 1.1074,
+      "step": 4580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018159373911550528,
+      "loss": 1.1389,
+      "step": 4581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018158473147447538,
+      "loss": 1.2473,
+      "step": 4582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018157572185342763,
+      "loss": 1.0763,
+      "step": 4583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018156671025258068,
+      "loss": 1.3396,
+      "step": 4584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018155769667215323,
+      "loss": 1.5256,
+      "step": 4585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018154868111236404,
+      "loss": 1.252,
+      "step": 4586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018153966357343192,
+      "loss": 1.2749,
+      "step": 4587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018153064405557575,
+      "loss": 1.3879,
+      "step": 4588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018152162255901435,
+      "loss": 1.1833,
+      "step": 4589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018151259908396673,
+      "loss": 2.4888,
+      "step": 4590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018150357363065186,
+      "loss": 2.5925,
+      "step": 4591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001814945461992888,
+      "loss": 2.4668,
+      "step": 4592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001814855167900966,
+      "loss": 1.222,
+      "step": 4593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018147648540329446,
+      "loss": 1.3167,
+      "step": 4594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018146745203910153,
+      "loss": 1.2141,
+      "step": 4595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018145841669773706,
+      "loss": 1.3616,
+      "step": 4596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018144937937942033,
+      "loss": 1.2654,
+      "step": 4597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018144034008437063,
+      "loss": 1.1566,
+      "step": 4598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018143129881280742,
+      "loss": 1.0704,
+      "step": 4599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018142225556495002,
+      "loss": 1.16,
+      "step": 4600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018141321034101802,
+      "loss": 1.2216,
+      "step": 4601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018140416314123088,
+      "loss": 1.091,
+      "step": 4602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018139511396580814,
+      "loss": 1.1707,
+      "step": 4603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001813860628149695,
+      "loss": 1.1002,
+      "step": 4604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018137700968893456,
+      "loss": 1.1387,
+      "step": 4605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018136795458792307,
+      "loss": 1.0607,
+      "step": 4606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018135889751215477,
+      "loss": 1.1873,
+      "step": 4607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018134983846184949,
+      "loss": 1.3745,
+      "step": 4608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018134077743722708,
+      "loss": 1.1798,
+      "step": 4609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018133171443850745,
+      "loss": 1.0919,
+      "step": 4610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018132264946591056,
+      "loss": 1.3052,
+      "step": 4611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018131358251965638,
+      "loss": 1.2416,
+      "step": 4612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000181304513599965,
+      "loss": 2.5935,
+      "step": 4613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018129544270705647,
+      "loss": 1.2535,
+      "step": 4614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000181286369841151,
+      "loss": 2.459,
+      "step": 4615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018127729500246872,
+      "loss": 1.3074,
+      "step": 4616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018126821819122992,
+      "loss": 1.1566,
+      "step": 4617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018125913940765487,
+      "loss": 1.3179,
+      "step": 4618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018125005865196392,
+      "loss": 1.0826,
+      "step": 4619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001812409759243774,
+      "loss": 1.217,
+      "step": 4620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001812318912251158,
+      "loss": 1.2248,
+      "step": 4621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018122280455439962,
+      "loss": 2.6545,
+      "step": 4622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018121371591244933,
+      "loss": 1.2983,
+      "step": 4623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018120462529948555,
+      "loss": 1.3825,
+      "step": 4624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018119553271572886,
+      "loss": 1.0729,
+      "step": 4625
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018118643816139994,
+      "loss": 1.2653,
+      "step": 4626
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018117734163671957,
+      "loss": 1.1632,
+      "step": 4627
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001811682431419085,
+      "loss": 1.084,
+      "step": 4628
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018115914267718747,
+      "loss": 1.2572,
+      "step": 4629
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018115004024277738,
+      "loss": 1.2458,
+      "step": 4630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001811409358388992,
+      "loss": 1.253,
+      "step": 4631
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018113182946577383,
+      "loss": 1.1561,
+      "step": 4632
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001811227211236223,
+      "loss": 1.2388,
+      "step": 4633
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018111361081266564,
+      "loss": 1.2288,
+      "step": 4634
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018110449853312496,
+      "loss": 1.416,
+      "step": 4635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018109538428522144,
+      "loss": 1.4027,
+      "step": 4636
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018108626806917624,
+      "loss": 1.2042,
+      "step": 4637
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018107714988521064,
+      "loss": 1.3162,
+      "step": 4638
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018106802973354592,
+      "loss": 1.2581,
+      "step": 4639
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001810589076144034,
+      "loss": 2.6338,
+      "step": 4640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001810497835280045,
+      "loss": 1.3109,
+      "step": 4641
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018104065747457062,
+      "loss": 1.2551,
+      "step": 4642
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001810315294543233,
+      "loss": 1.4789,
+      "step": 4643
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000181022399467484,
+      "loss": 1.2293,
+      "step": 4644
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001810132675142744,
+      "loss": 2.2611,
+      "step": 4645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018100413359491602,
+      "loss": 1.1393,
+      "step": 4646
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018099499770963057,
+      "loss": 1.3246,
+      "step": 4647
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018098585985863982,
+      "loss": 1.4816,
+      "step": 4648
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018097672004216547,
+      "loss": 1.2609,
+      "step": 4649
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018096757826042943,
+      "loss": 1.1611,
+      "step": 4650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018095843451365346,
+      "loss": 1.2659,
+      "step": 4651
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018094928880205954,
+      "loss": 1.1772,
+      "step": 4652
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001809401411258696,
+      "loss": 1.3387,
+      "step": 4653
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001809309914853057,
+      "loss": 1.2341,
+      "step": 4654
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018092183988058986,
+      "loss": 1.1457,
+      "step": 4655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018091268631194416,
+      "loss": 1.167,
+      "step": 4656
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018090353077959082,
+      "loss": 1.4254,
+      "step": 4657
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018089437328375194,
+      "loss": 1.2516,
+      "step": 4658
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018088521382464987,
+      "loss": 1.349,
+      "step": 4659
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018087605240250682,
+      "loss": 1.309,
+      "step": 4660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018086688901754522,
+      "loss": 1.1831,
+      "step": 4661
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001808577236699874,
+      "loss": 1.3011,
+      "step": 4662
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018084855636005582,
+      "loss": 1.3492,
+      "step": 4663
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018083938708797294,
+      "loss": 1.422,
+      "step": 4664
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018083021585396135,
+      "loss": 1.1771,
+      "step": 4665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018082104265824356,
+      "loss": 1.2713,
+      "step": 4666
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018081186750104223,
+      "loss": 1.3495,
+      "step": 4667
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018080269038258005,
+      "loss": 1.4083,
+      "step": 4668
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001807935113030797,
+      "loss": 1.3602,
+      "step": 4669
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018078433026276402,
+      "loss": 1.2152,
+      "step": 4670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018077514726185578,
+      "loss": 1.1612,
+      "step": 4671
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018076596230057787,
+      "loss": 1.3479,
+      "step": 4672
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018075677537915315,
+      "loss": 1.2896,
+      "step": 4673
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018074758649780468,
+      "loss": 1.4333,
+      "step": 4674
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018073839565675538,
+      "loss": 1.2203,
+      "step": 4675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018072920285622837,
+      "loss": 1.0832,
+      "step": 4676
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018072000809644668,
+      "loss": 1.1722,
+      "step": 4677
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018071081137763356,
+      "loss": 1.2091,
+      "step": 4678
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001807016127000121,
+      "loss": 1.2955,
+      "step": 4679
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018069241206380564,
+      "loss": 1.2903,
+      "step": 4680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018068320946923742,
+      "loss": 1.2245,
+      "step": 4681
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001806740049165308,
+      "loss": 1.2317,
+      "step": 4682
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001806647984059092,
+      "loss": 1.1252,
+      "step": 4683
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000180655589937596,
+      "loss": 1.3097,
+      "step": 4684
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018064637951181468,
+      "loss": 1.1793,
+      "step": 4685
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018063716712878885,
+      "loss": 1.1744,
+      "step": 4686
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018062795278874202,
+      "loss": 1.262,
+      "step": 4687
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018061873649189786,
+      "loss": 1.3196,
+      "step": 4688
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018060951823848,
+      "loss": 1.1842,
+      "step": 4689
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018060029802871222,
+      "loss": 1.2152,
+      "step": 4690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018059107586281822,
+      "loss": 1.3018,
+      "step": 4691
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018058185174102187,
+      "loss": 1.2851,
+      "step": 4692
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000180572625663547,
+      "loss": 1.3111,
+      "step": 4693
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018056339763061755,
+      "loss": 1.2426,
+      "step": 4694
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018055416764245749,
+      "loss": 1.19,
+      "step": 4695
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018054493569929077,
+      "loss": 1.1048,
+      "step": 4696
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001805357018013415,
+      "loss": 2.4312,
+      "step": 4697
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018052646594883377,
+      "loss": 1.0594,
+      "step": 4698
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001805172281419917,
+      "loss": 1.2065,
+      "step": 4699
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018050798838103953,
+      "loss": 1.3022,
+      "step": 4700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018049874666620146,
+      "loss": 1.1608,
+      "step": 4701
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001804895029977018,
+      "loss": 1.1961,
+      "step": 4702
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001804802573757649,
+      "loss": 1.3647,
+      "step": 4703
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018047100980061515,
+      "loss": 2.5247,
+      "step": 4704
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018046176027247698,
+      "loss": 1.2829,
+      "step": 4705
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018045250879157482,
+      "loss": 1.0993,
+      "step": 4706
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018044325535813326,
+      "loss": 1.1412,
+      "step": 4707
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001804339999723769,
+      "loss": 1.2777,
+      "step": 4708
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018042474263453028,
+      "loss": 1.3776,
+      "step": 4709
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018041548334481812,
+      "loss": 2.5803,
+      "step": 4710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018040622210346513,
+      "loss": 1.1347,
+      "step": 4711
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018039695891069606,
+      "loss": 1.1892,
+      "step": 4712
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018038769376673574,
+      "loss": 1.1447,
+      "step": 4713
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018037842667180908,
+      "loss": 1.3474,
+      "step": 4714
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018036915762614087,
+      "loss": 1.1419,
+      "step": 4715
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018035988662995617,
+      "loss": 1.1614,
+      "step": 4716
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018035061368347992,
+      "loss": 2.4558,
+      "step": 4717
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018034133878693718,
+      "loss": 1.2032,
+      "step": 4718
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018033206194055306,
+      "loss": 1.2906,
+      "step": 4719
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018032278314455272,
+      "loss": 1.1633,
+      "step": 4720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018031350239916134,
+      "loss": 1.1451,
+      "step": 4721
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018030421970460417,
+      "loss": 1.2511,
+      "step": 4722
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018029493506110646,
+      "loss": 1.0956,
+      "step": 4723
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018028564846889355,
+      "loss": 1.2123,
+      "step": 4724
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018027635992819084,
+      "loss": 2.3745,
+      "step": 4725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018026706943922373,
+      "loss": 1.1963,
+      "step": 4726
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018025777700221777,
+      "loss": 1.276,
+      "step": 4727
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018024848261739837,
+      "loss": 1.3116,
+      "step": 4728
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018023918628499116,
+      "loss": 1.3586,
+      "step": 4729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018022988800522178,
+      "loss": 1.3053,
+      "step": 4730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018022058777831588,
+      "loss": 1.3007,
+      "step": 4731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018021128560449914,
+      "loss": 1.2006,
+      "step": 4732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018020198148399735,
+      "loss": 1.1622,
+      "step": 4733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018019267541703632,
+      "loss": 1.3148,
+      "step": 4734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018018336740384183,
+      "loss": 1.0943,
+      "step": 4735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018017405744463988,
+      "loss": 1.2109,
+      "step": 4736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001801647455396564,
+      "loss": 1.249,
+      "step": 4737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018015543168911733,
+      "loss": 1.2505,
+      "step": 4738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018014611589324875,
+      "loss": 1.1997,
+      "step": 4739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018013679815227676,
+      "loss": 1.1075,
+      "step": 4740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018012747846642743,
+      "loss": 1.0889,
+      "step": 4741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018011815683592706,
+      "loss": 1.2938,
+      "step": 4742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001801088332610018,
+      "loss": 1.1384,
+      "step": 4743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001800995077418779,
+      "loss": 1.2066,
+      "step": 4744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018009018027878178,
+      "loss": 0.9839,
+      "step": 4745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018008085087193974,
+      "loss": 1.1944,
+      "step": 4746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001800715195215782,
+      "loss": 1.3065,
+      "step": 4747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018006218622792366,
+      "loss": 1.1248,
+      "step": 4748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018005285099120263,
+      "loss": 1.2227,
+      "step": 4749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018004351381164168,
+      "loss": 1.1402,
+      "step": 4750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018003417468946738,
+      "loss": 1.4142,
+      "step": 4751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018002483362490642,
+      "loss": 2.5862,
+      "step": 4752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018001549061818546,
+      "loss": 1.24,
+      "step": 4753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001800061456695313,
+      "loss": 1.2006,
+      "step": 4754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017999679877917074,
+      "loss": 1.1344,
+      "step": 4755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017998744994733056,
+      "loss": 1.2103,
+      "step": 4756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001799780991742377,
+      "loss": 1.1371,
+      "step": 4757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001799687464601191,
+      "loss": 1.2944,
+      "step": 4758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017995939180520174,
+      "loss": 1.3019,
+      "step": 4759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017995003520971262,
+      "loss": 2.5107,
+      "step": 4760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001799406766738789,
+      "loss": 1.1736,
+      "step": 4761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001799313161979276,
+      "loss": 1.209,
+      "step": 4762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017992195378208596,
+      "loss": 1.3328,
+      "step": 4763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001799125894265812,
+      "loss": 1.2178,
+      "step": 4764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017990322313164055,
+      "loss": 1.1984,
+      "step": 4765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001798938548974914,
+      "loss": 1.1906,
+      "step": 4766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017988448472436105,
+      "loss": 1.1281,
+      "step": 4767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001798751126124769,
+      "loss": 1.2349,
+      "step": 4768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017986573856206643,
+      "loss": 1.2,
+      "step": 4769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017985636257335713,
+      "loss": 1.3022,
+      "step": 4770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001798469846465766,
+      "loss": 1.107,
+      "step": 4771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017983760478195238,
+      "loss": 1.3556,
+      "step": 4772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017982822297971213,
+      "loss": 1.2481,
+      "step": 4773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017981883924008356,
+      "loss": 1.3665,
+      "step": 4774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017980945356329438,
+      "loss": 1.0321,
+      "step": 4775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001798000659495724,
+      "loss": 1.3419,
+      "step": 4776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017979067639914545,
+      "loss": 1.17,
+      "step": 4777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017978128491224137,
+      "loss": 1.2532,
+      "step": 4778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001797718914890881,
+      "loss": 1.2444,
+      "step": 4779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017976249612991366,
+      "loss": 1.3083,
+      "step": 4780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017975309883494606,
+      "loss": 1.2928,
+      "step": 4781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017974369960441333,
+      "loss": 1.3077,
+      "step": 4782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001797342984385436,
+      "loss": 1.3962,
+      "step": 4783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017972489533756502,
+      "loss": 1.2293,
+      "step": 4784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017971549030170583,
+      "loss": 1.1004,
+      "step": 4785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017970608333119428,
+      "loss": 1.3021,
+      "step": 4786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017969667442625862,
+      "loss": 1.0981,
+      "step": 4787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017968726358712727,
+      "loss": 1.3701,
+      "step": 4788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017967785081402857,
+      "loss": 1.1038,
+      "step": 4789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000179668436107191,
+      "loss": 2.5195,
+      "step": 4790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017965901946684305,
+      "loss": 1.1348,
+      "step": 4791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017964960089321324,
+      "loss": 1.4185,
+      "step": 4792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017964018038653016,
+      "loss": 1.4378,
+      "step": 4793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001796307579470224,
+      "loss": 1.1268,
+      "step": 4794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001796213335749187,
+      "loss": 1.0638,
+      "step": 4795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017961190727044777,
+      "loss": 1.1515,
+      "step": 4796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017960247903383837,
+      "loss": 1.0757,
+      "step": 4797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001795930488653193,
+      "loss": 1.1828,
+      "step": 4798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017958361676511948,
+      "loss": 1.1481,
+      "step": 4799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017957418273346773,
+      "loss": 1.0924,
+      "step": 4800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017956474677059312,
+      "loss": 1.264,
+      "step": 4801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017955530887672457,
+      "loss": 1.3942,
+      "step": 4802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001795458690520912,
+      "loss": 1.1107,
+      "step": 4803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017953642729692204,
+      "loss": 1.1631,
+      "step": 4804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017952698361144628,
+      "loss": 0.9755,
+      "step": 4805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001795175379958931,
+      "loss": 1.3739,
+      "step": 4806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017950809045049179,
+      "loss": 1.3575,
+      "step": 4807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017949864097547153,
+      "loss": 1.1686,
+      "step": 4808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017948918957106172,
+      "loss": 1.1044,
+      "step": 4809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017947973623749176,
+      "loss": 1.2418,
+      "step": 4810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017947028097499104,
+      "loss": 1.2531,
+      "step": 4811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017946082378378906,
+      "loss": 1.4573,
+      "step": 4812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001794513646641153,
+      "loss": 1.4615,
+      "step": 4813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017944190361619935,
+      "loss": 1.2487,
+      "step": 4814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017943244064027087,
+      "loss": 1.2213,
+      "step": 4815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017942297573655947,
+      "loss": 1.1517,
+      "step": 4816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017941350890529484,
+      "loss": 1.1906,
+      "step": 4817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017940404014670677,
+      "loss": 1.4294,
+      "step": 4818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017939456946102509,
+      "loss": 1.281,
+      "step": 4819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001793850968484796,
+      "loss": 1.2856,
+      "step": 4820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017937562230930018,
+      "loss": 1.2869,
+      "step": 4821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017936614584371683,
+      "loss": 1.4813,
+      "step": 4822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017935666745195951,
+      "loss": 1.157,
+      "step": 4823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017934718713425823,
+      "loss": 1.2219,
+      "step": 4824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001793377048908431,
+      "loss": 1.0168,
+      "step": 4825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001793282207219443,
+      "loss": 1.2925,
+      "step": 4826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001793187346277919,
+      "loss": 1.4013,
+      "step": 4827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017930924660861618,
+      "loss": 1.2743,
+      "step": 4828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017929975666464745,
+      "loss": 1.2133,
+      "step": 4829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017929026479611591,
+      "loss": 1.1686,
+      "step": 4830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017928077100325205,
+      "loss": 1.1542,
+      "step": 4831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001792712752862862,
+      "loss": 1.2172,
+      "step": 4832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017926177764544883,
+      "loss": 1.1667,
+      "step": 4833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017925227808097047,
+      "loss": 1.1055,
+      "step": 4834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017924277659308165,
+      "loss": 1.3247,
+      "step": 4835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017923327318201295,
+      "loss": 1.2492,
+      "step": 4836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017922376784799501,
+      "loss": 1.2209,
+      "step": 4837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017921426059125856,
+      "loss": 1.281,
+      "step": 4838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017920475141203435,
+      "loss": 1.2251,
+      "step": 4839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017919524031055305,
+      "loss": 1.2097,
+      "step": 4840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017918572728704564,
+      "loss": 2.5007,
+      "step": 4841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001791762123417429,
+      "loss": 1.0864,
+      "step": 4842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017916669547487577,
+      "loss": 1.0656,
+      "step": 4843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001791571766866752,
+      "loss": 1.1931,
+      "step": 4844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017914765597737225,
+      "loss": 1.2167,
+      "step": 4845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017913813334719797,
+      "loss": 1.2891,
+      "step": 4846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017912860879638345,
+      "loss": 1.3732,
+      "step": 4847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017911908232515986,
+      "loss": 1.2522,
+      "step": 4848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001791095539337584,
+      "loss": 1.355,
+      "step": 4849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017910002362241033,
+      "loss": 1.2665,
+      "step": 4850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017909049139134693,
+      "loss": 1.3008,
+      "step": 4851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001790809572407995,
+      "loss": 1.3561,
+      "step": 4852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017907142117099952,
+      "loss": 1.2656,
+      "step": 4853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001790618831821784,
+      "loss": 1.2599,
+      "step": 4854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017905234327456756,
+      "loss": 1.2654,
+      "step": 4855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001790428014483986,
+      "loss": 1.3374,
+      "step": 4856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017903325770390308,
+      "loss": 1.3329,
+      "step": 4857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017902371204131257,
+      "loss": 1.3463,
+      "step": 4858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001790141644608588,
+      "loss": 1.0854,
+      "step": 4859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017900461496277346,
+      "loss": 1.1717,
+      "step": 4860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001789950635472883,
+      "loss": 1.007,
+      "step": 4861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017898551021463513,
+      "loss": 1.4093,
+      "step": 4862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017897595496504582,
+      "loss": 1.2648,
+      "step": 4863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017896639779875227,
+      "loss": 1.1849,
+      "step": 4864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017895683871598642,
+      "loss": 1.1494,
+      "step": 4865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017894727771698026,
+      "loss": 1.2122,
+      "step": 4866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017893771480196585,
+      "loss": 1.1569,
+      "step": 4867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017892814997117525,
+      "loss": 1.192,
+      "step": 4868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001789185832248406,
+      "loss": 1.2419,
+      "step": 4869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017890901456319412,
+      "loss": 1.3867,
+      "step": 4870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017889944398646798,
+      "loss": 1.5249,
+      "step": 4871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017888987149489448,
+      "loss": 1.119,
+      "step": 4872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017888029708870593,
+      "loss": 1.3112,
+      "step": 4873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017887072076813468,
+      "loss": 1.1034,
+      "step": 4874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001788611425334132,
+      "loss": 1.0829,
+      "step": 4875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017885156238477387,
+      "loss": 1.1127,
+      "step": 4876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017884198032244925,
+      "loss": 1.3685,
+      "step": 4877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001788323963466719,
+      "loss": 1.1877,
+      "step": 4878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001788228104576744,
+      "loss": 1.1631,
+      "step": 4879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017881322265568935,
+      "loss": 2.3718,
+      "step": 4880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017880363294094952,
+      "loss": 1.3611,
+      "step": 4881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017879404131368757,
+      "loss": 1.1454,
+      "step": 4882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017878444777413635,
+      "loss": 1.2454,
+      "step": 4883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017877485232252864,
+      "loss": 1.3089,
+      "step": 4884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017876525495909737,
+      "loss": 1.2087,
+      "step": 4885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001787556556840754,
+      "loss": 1.3026,
+      "step": 4886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017874605449769572,
+      "loss": 1.2478,
+      "step": 4887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001787364514001914,
+      "loss": 1.2662,
+      "step": 4888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017872684639179545,
+      "loss": 1.2986,
+      "step": 4889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017871723947274096,
+      "loss": 2.5222,
+      "step": 4890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017870763064326112,
+      "loss": 1.2714,
+      "step": 4891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017869801990358911,
+      "loss": 1.296,
+      "step": 4892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001786884072539582,
+      "loss": 1.1993,
+      "step": 4893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017867879269460168,
+      "loss": 1.3103,
+      "step": 4894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001786691762257529,
+      "loss": 1.2178,
+      "step": 4895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017865955784764523,
+      "loss": 1.3096,
+      "step": 4896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017864993756051207,
+      "loss": 1.3184,
+      "step": 4897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017864031536458694,
+      "loss": 1.3608,
+      "step": 4898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017863069126010337,
+      "loss": 1.067,
+      "step": 4899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001786210652472949,
+      "loss": 1.1636,
+      "step": 4900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017861143732639522,
+      "loss": 1.335,
+      "step": 4901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001786018074976379,
+      "loss": 1.324,
+      "step": 4902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017859217576125672,
+      "loss": 1.2895,
+      "step": 4903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017858254211748536,
+      "loss": 1.0926,
+      "step": 4904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017857290656655771,
+      "loss": 1.2427,
+      "step": 4905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017856326910870759,
+      "loss": 1.2022,
+      "step": 4906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001785536297441689,
+      "loss": 1.1647,
+      "step": 4907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017854398847317555,
+      "loss": 1.2259,
+      "step": 4908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017853434529596152,
+      "loss": 2.5115,
+      "step": 4909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017852470021276093,
+      "loss": 1.1683,
+      "step": 4910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001785150532238078,
+      "loss": 1.1857,
+      "step": 4911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017850540432933623,
+      "loss": 1.3529,
+      "step": 4912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017849575352958045,
+      "loss": 1.3562,
+      "step": 4913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017848610082477463,
+      "loss": 1.0614,
+      "step": 4914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017847644621515313,
+      "loss": 1.1069,
+      "step": 4915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017846678970095012,
+      "loss": 1.2166,
+      "step": 4916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017845713128240007,
+      "loss": 1.2159,
+      "step": 4917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017844747095973736,
+      "loss": 1.2639,
+      "step": 4918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001784378087331964,
+      "loss": 1.0421,
+      "step": 4919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017842814460301173,
+      "loss": 1.1391,
+      "step": 4920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017841847856941786,
+      "loss": 1.2744,
+      "step": 4921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017840881063264943,
+      "loss": 1.3152,
+      "step": 4922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000178399140792941,
+      "loss": 1.3114,
+      "step": 4923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017838946905052736,
+      "loss": 1.1848,
+      "step": 4924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017837979540564312,
+      "loss": 1.1565,
+      "step": 4925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017837011985852313,
+      "loss": 1.3014,
+      "step": 4926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017836044240940218,
+      "loss": 1.1466,
+      "step": 4927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017835076305851516,
+      "loss": 1.0968,
+      "step": 4928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017834108180609696,
+      "loss": 0.9884,
+      "step": 4929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017833139865238256,
+      "loss": 1.1212,
+      "step": 4930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017832171359760695,
+      "loss": 1.3814,
+      "step": 4931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017831202664200517,
+      "loss": 2.5798,
+      "step": 4932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017830233778581232,
+      "loss": 1.2048,
+      "step": 4933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001782926470292636,
+      "loss": 1.2321,
+      "step": 4934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017828295437259407,
+      "loss": 1.1795,
+      "step": 4935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017827325981603913,
+      "loss": 1.2025,
+      "step": 4936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00017826356335983395,
+      "loss": 1.09,
+      "step": 4937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017825386500421386,
+      "loss": 1.1802,
+      "step": 4938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001782441647494143,
+      "loss": 1.4703,
+      "step": 4939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017823446259567063,
+      "loss": 1.3671,
+      "step": 4940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017822475854321836,
+      "loss": 1.2064,
+      "step": 4941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017821505259229297,
+      "loss": 1.2993,
+      "step": 4942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017820534474313003,
+      "loss": 1.132,
+      "step": 4943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017819563499596515,
+      "loss": 1.2168,
+      "step": 4944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017818592335103395,
+      "loss": 1.2043,
+      "step": 4945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017817620980857218,
+      "loss": 2.4875,
+      "step": 4946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017816649436881553,
+      "loss": 1.1082,
+      "step": 4947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001781567770319998,
+      "loss": 1.1116,
+      "step": 4948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017814705779836088,
+      "loss": 1.1208,
+      "step": 4949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017813733666813457,
+      "loss": 1.2225,
+      "step": 4950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017812761364155682,
+      "loss": 2.542,
+      "step": 4951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017811788871886365,
+      "loss": 1.1586,
+      "step": 4952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017810816190029103,
+      "loss": 1.2814,
+      "step": 4953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017809843318607502,
+      "loss": 1.1068,
+      "step": 4954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001780887025764518,
+      "loss": 1.1902,
+      "step": 4955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017807897007165741,
+      "loss": 1.2004,
+      "step": 4956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017806923567192817,
+      "loss": 1.3802,
+      "step": 4957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017805949937750026,
+      "loss": 1.1528,
+      "step": 4958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017804976118860998,
+      "loss": 1.3311,
+      "step": 4959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001780400211054937,
+      "loss": 1.1152,
+      "step": 4960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017803027912838778,
+      "loss": 1.2188,
+      "step": 4961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017802053525752868,
+      "loss": 1.1503,
+      "step": 4962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017801078949315283,
+      "loss": 1.1019,
+      "step": 4963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001780010418354968,
+      "loss": 1.1659,
+      "step": 4964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017799129228479715,
+      "loss": 1.1483,
+      "step": 4965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001779815408412905,
+      "loss": 1.2309,
+      "step": 4966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017797178750521349,
+      "loss": 1.147,
+      "step": 4967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017796203227680283,
+      "loss": 1.3759,
+      "step": 4968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001779522751562953,
+      "loss": 1.1629,
+      "step": 4969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001779425161439277,
+      "loss": 1.22,
+      "step": 4970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017793275523993684,
+      "loss": 1.2336,
+      "step": 4971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017792299244455965,
+      "loss": 1.3472,
+      "step": 4972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017791322775803306,
+      "loss": 1.0853,
+      "step": 4973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017790346118059402,
+      "loss": 1.1859,
+      "step": 4974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001778936927124796,
+      "loss": 1.2879,
+      "step": 4975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017788392235392687,
+      "loss": 1.14,
+      "step": 4976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017787415010517293,
+      "loss": 1.1849,
+      "step": 4977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017786437596645495,
+      "loss": 1.2556,
+      "step": 4978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017785459993801018,
+      "loss": 1.0016,
+      "step": 4979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017784482202007581,
+      "loss": 1.4297,
+      "step": 4980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017783504221288924,
+      "loss": 1.179,
+      "step": 4981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017782526051668773,
+      "loss": 1.0756,
+      "step": 4982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017781547693170873,
+      "loss": 1.2127,
+      "step": 4983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017780569145818968,
+      "loss": 1.2107,
+      "step": 4984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017779590409636802,
+      "loss": 1.33,
+      "step": 4985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017778611484648135,
+      "loss": 1.2189,
+      "step": 4986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001777763237087672,
+      "loss": 1.3614,
+      "step": 4987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017776653068346323,
+      "loss": 1.0948,
+      "step": 4988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017775673577080707,
+      "loss": 1.3187,
+      "step": 4989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001777469389710365,
+      "loss": 1.105,
+      "step": 4990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017773714028438922,
+      "loss": 1.0214,
+      "step": 4991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017772733971110305,
+      "loss": 1.3344,
+      "step": 4992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001777175372514159,
+      "loss": 1.2345,
+      "step": 4993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001777077329055656,
+      "loss": 2.532,
+      "step": 4994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017769792667379013,
+      "loss": 1.0828,
+      "step": 4995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001776881185563275,
+      "loss": 1.3387,
+      "step": 4996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001776783085534157,
+      "loss": 1.1658,
+      "step": 4997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017766849666529284,
+      "loss": 1.2048,
+      "step": 4998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017765868289219708,
+      "loss": 1.3596,
+      "step": 4999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017764886723436654,
+      "loss": 1.1642,
+      "step": 5000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017763904969203944,
+      "loss": 1.2358,
+      "step": 5001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017762923026545411,
+      "loss": 1.0023,
+      "step": 5002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001776194089548488,
+      "loss": 1.2586,
+      "step": 5003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001776095857604619,
+      "loss": 1.2857,
+      "step": 5004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017759976068253185,
+      "loss": 1.1405,
+      "step": 5005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000177589933721297,
+      "loss": 1.2166,
+      "step": 5006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001775801048769959,
+      "loss": 1.0283,
+      "step": 5007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001775702741498671,
+      "loss": 1.1489,
+      "step": 5008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001775604415401492,
+      "loss": 2.4233,
+      "step": 5009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001775506070480808,
+      "loss": 1.2968,
+      "step": 5010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017754077067390055,
+      "loss": 1.1621,
+      "step": 5011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017753093241784727,
+      "loss": 1.3516,
+      "step": 5012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001775210922801596,
+      "loss": 1.2715,
+      "step": 5013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017751125026107647,
+      "loss": 1.2516,
+      "step": 5014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001775014063608367,
+      "loss": 1.2218,
+      "step": 5015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001774915605796792,
+      "loss": 1.1947,
+      "step": 5016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017748171291784285,
+      "loss": 1.2653,
+      "step": 5017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017747186337556677,
+      "loss": 1.1926,
+      "step": 5018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017746201195308994,
+      "loss": 2.3804,
+      "step": 5019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017745215865065143,
+      "loss": 1.0728,
+      "step": 5020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001774423034684904,
+      "loss": 1.0073,
+      "step": 5021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017743244640684605,
+      "loss": 1.2455,
+      "step": 5022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017742258746595758,
+      "loss": 1.0388,
+      "step": 5023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017741272664606423,
+      "loss": 2.2722,
+      "step": 5024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017740286394740538,
+      "loss": 1.2589,
+      "step": 5025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017739299937022035,
+      "loss": 1.2747,
+      "step": 5026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017738313291474858,
+      "loss": 1.1163,
+      "step": 5027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017737326458122947,
+      "loss": 2.3464,
+      "step": 5028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017736339436990258,
+      "loss": 1.0964,
+      "step": 5029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017735352228100744,
+      "loss": 2.29,
+      "step": 5030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001773436483147836,
+      "loss": 1.2328,
+      "step": 5031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017733377247147075,
+      "loss": 1.1314,
+      "step": 5032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017732389475130854,
+      "loss": 1.0545,
+      "step": 5033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017731401515453667,
+      "loss": 1.1084,
+      "step": 5034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017730413368139498,
+      "loss": 1.2141,
+      "step": 5035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017729425033212326,
+      "loss": 1.0933,
+      "step": 5036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017728436510696134,
+      "loss": 1.3353,
+      "step": 5037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017727447800614917,
+      "loss": 1.2734,
+      "step": 5038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001772645890299267,
+      "loss": 1.3134,
+      "step": 5039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017725469817853393,
+      "loss": 1.3723,
+      "step": 5040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00017724480545221088,
+      "loss": 1.1776,
+      "step": 5041
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017723491085119765,
+      "loss": 1.2766,
+      "step": 5042
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001772250143757344,
+      "loss": 1.2275,
+      "step": 5043
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017721511602606128,
+      "loss": 1.3222,
+      "step": 5044
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017720521580241857,
+      "loss": 1.0667,
+      "step": 5045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017719531370504648,
+      "loss": 1.1788,
+      "step": 5046
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017718540973418536,
+      "loss": 1.1768,
+      "step": 5047
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001771755038900756,
+      "loss": 2.4761,
+      "step": 5048
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017716559617295753,
+      "loss": 1.3394,
+      "step": 5049
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001771556865830717,
+      "loss": 1.096,
+      "step": 5050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017714577512065852,
+      "loss": 1.2444,
+      "step": 5051
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017713586178595863,
+      "loss": 1.2661,
+      "step": 5052
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017712594657921256,
+      "loss": 2.6042,
+      "step": 5053
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017711602950066095,
+      "loss": 1.0422,
+      "step": 5054
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001771061105505445,
+      "loss": 1.199,
+      "step": 5055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017709618972910391,
+      "loss": 1.3625,
+      "step": 5056
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017708626703658,
+      "loss": 1.2856,
+      "step": 5057
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017707634247321355,
+      "loss": 1.233,
+      "step": 5058
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001770664160392454,
+      "loss": 1.2917,
+      "step": 5059
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001770564877349165,
+      "loss": 1.1915,
+      "step": 5060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017704655756046783,
+      "loss": 1.0899,
+      "step": 5061
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017703662551614037,
+      "loss": 1.2087,
+      "step": 5062
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017702669160217513,
+      "loss": 1.0781,
+      "step": 5063
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017701675581881324,
+      "loss": 1.2853,
+      "step": 5064
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017700681816629584,
+      "loss": 1.3268,
+      "step": 5065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017699687864486404,
+      "loss": 1.2586,
+      "step": 5066
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017698693725475914,
+      "loss": 1.3249,
+      "step": 5067
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017697699399622243,
+      "loss": 1.2216,
+      "step": 5068
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017696704886949514,
+      "loss": 1.2875,
+      "step": 5069
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017695710187481871,
+      "loss": 1.155,
+      "step": 5070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001769471530124345,
+      "loss": 1.2833,
+      "step": 5071
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017693720228258402,
+      "loss": 1.2112,
+      "step": 5072
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001769272496855087,
+      "loss": 1.1834,
+      "step": 5073
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001769172952214501,
+      "loss": 1.0183,
+      "step": 5074
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017690733889064988,
+      "loss": 1.3522,
+      "step": 5075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001768973806933496,
+      "loss": 2.5676,
+      "step": 5076
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017688742062979093,
+      "loss": 2.3945,
+      "step": 5077
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017687745870021565,
+      "loss": 1.4025,
+      "step": 5078
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017686749490486553,
+      "loss": 1.4308,
+      "step": 5079
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017685752924398234,
+      "loss": 1.1623,
+      "step": 5080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017684756171780798,
+      "loss": 1.3129,
+      "step": 5081
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017683759232658432,
+      "loss": 2.5193,
+      "step": 5082
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017682762107055334,
+      "loss": 1.2718,
+      "step": 5083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017681764794995702,
+      "loss": 1.0986,
+      "step": 5084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017680767296503747,
+      "loss": 1.2725,
+      "step": 5085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017679769611603666,
+      "loss": 1.1292,
+      "step": 5086
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001767877174031968,
+      "loss": 1.0091,
+      "step": 5087
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017677773682676007,
+      "loss": 1.2596,
+      "step": 5088
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017676775438696868,
+      "loss": 1.1442,
+      "step": 5089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017675777008406484,
+      "loss": 2.4412,
+      "step": 5090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017674778391829095,
+      "loss": 1.3383,
+      "step": 5091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017673779588988934,
+      "loss": 1.254,
+      "step": 5092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001767278059991024,
+      "loss": 2.574,
+      "step": 5093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017671781424617258,
+      "loss": 1.3533,
+      "step": 5094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001767078206313424,
+      "loss": 1.252,
+      "step": 5095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017669782515485438,
+      "loss": 1.2492,
+      "step": 5096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017668782781695113,
+      "loss": 1.3542,
+      "step": 5097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017667782861787522,
+      "loss": 1.1425,
+      "step": 5098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017666782755786938,
+      "loss": 1.3012,
+      "step": 5099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001766578246371763,
+      "loss": 1.2585,
+      "step": 5100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001766478198560388,
+      "loss": 1.1802,
+      "step": 5101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017663781321469962,
+      "loss": 2.3525,
+      "step": 5102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017662780471340166,
+      "loss": 1.1637,
+      "step": 5103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017661779435238777,
+      "loss": 1.1638,
+      "step": 5104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017660778213190098,
+      "loss": 2.394,
+      "step": 5105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001765977680521842,
+      "loss": 1.246,
+      "step": 5106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001765877521134805,
+      "loss": 1.2434,
+      "step": 5107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000176577734316033,
+      "loss": 1.2886,
+      "step": 5108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017656771466008475,
+      "loss": 2.4124,
+      "step": 5109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017655769314587897,
+      "loss": 1.2516,
+      "step": 5110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017654766977365886,
+      "loss": 1.1719,
+      "step": 5111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017653764454366773,
+      "loss": 1.3271,
+      "step": 5112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017652761745614882,
+      "loss": 1.2554,
+      "step": 5113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017651758851134548,
+      "loss": 1.4925,
+      "step": 5114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017650755770950118,
+      "loss": 1.3661,
+      "step": 5115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017649752505085928,
+      "loss": 1.272,
+      "step": 5116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017648749053566332,
+      "loss": 1.1516,
+      "step": 5117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017647745416415683,
+      "loss": 1.3256,
+      "step": 5118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017646741593658336,
+      "loss": 2.1472,
+      "step": 5119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017645737585318653,
+      "loss": 1.1205,
+      "step": 5120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017644733391421007,
+      "loss": 1.3761,
+      "step": 5121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017643729011989763,
+      "loss": 1.3146,
+      "step": 5122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017642724447049295,
+      "loss": 2.4099,
+      "step": 5123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001764171969662399,
+      "loss": 1.2777,
+      "step": 5124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001764071476073823,
+      "loss": 1.2491,
+      "step": 5125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000176397096394164,
+      "loss": 1.119,
+      "step": 5126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000176387043326829,
+      "loss": 1.3102,
+      "step": 5127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017637698840562127,
+      "loss": 1.0787,
+      "step": 5128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017636693163078482,
+      "loss": 1.0281,
+      "step": 5129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001763568730025637,
+      "loss": 1.24,
+      "step": 5130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001763468125212021,
+      "loss": 1.2199,
+      "step": 5131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001763367501869441,
+      "loss": 1.2211,
+      "step": 5132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017632668600003394,
+      "loss": 1.1369,
+      "step": 5133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017631661996071589,
+      "loss": 1.0967,
+      "step": 5134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017630655206923426,
+      "loss": 1.1154,
+      "step": 5135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017629648232583335,
+      "loss": 2.5447,
+      "step": 5136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017628641073075754,
+      "loss": 1.1483,
+      "step": 5137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017627633728425131,
+      "loss": 1.1663,
+      "step": 5138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001762662619865591,
+      "loss": 1.0736,
+      "step": 5139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017625618483792544,
+      "loss": 1.3201,
+      "step": 5140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001762461058385949,
+      "loss": 1.3834,
+      "step": 5141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017623602498881215,
+      "loss": 1.2968,
+      "step": 5142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017622594228882172,
+      "loss": 2.3218,
+      "step": 5143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017621585773886845,
+      "loss": 1.3421,
+      "step": 5144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00017620577133919697,
+      "loss": 1.1198,
+      "step": 5145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017619568309005214,
+      "loss": 1.514,
+      "step": 5146
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017618559299167877,
+      "loss": 1.2353,
+      "step": 5147
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017617550104432174,
+      "loss": 1.1321,
+      "step": 5148
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017616540724822602,
+      "loss": 1.1806,
+      "step": 5149
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017615531160363653,
+      "loss": 1.3136,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001761452141107983,
+      "loss": 1.0933,
+      "step": 5151
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017613511476995638,
+      "loss": 1.1779,
+      "step": 5152
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017612501358135591,
+      "loss": 1.1827,
+      "step": 5153
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017611491054524202,
+      "loss": 1.2513,
+      "step": 5154
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017610480566185988,
+      "loss": 1.1771,
+      "step": 5155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017609469893145477,
+      "loss": 1.2446,
+      "step": 5156
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017608459035427197,
+      "loss": 1.1899,
+      "step": 5157
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017607447993055677,
+      "loss": 1.2948,
+      "step": 5158
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001760643676605546,
+      "loss": 2.3792,
+      "step": 5159
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017605425354451083,
+      "loss": 1.2231,
+      "step": 5160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017604413758267096,
+      "loss": 1.137,
+      "step": 5161
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001760340197752805,
+      "loss": 1.4907,
+      "step": 5162
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017602390012258495,
+      "loss": 1.2289,
+      "step": 5163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017601377862482999,
+      "loss": 1.3265,
+      "step": 5164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017600365528226122,
+      "loss": 1.0425,
+      "step": 5165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001759935300951243,
+      "loss": 1.2407,
+      "step": 5166
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017598340306366502,
+      "loss": 1.1964,
+      "step": 5167
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017597327418812913,
+      "loss": 1.1759,
+      "step": 5168
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017596314346876243,
+      "loss": 1.2366,
+      "step": 5169
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017595301090581083,
+      "loss": 1.4747,
+      "step": 5170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001759428764995202,
+      "loss": 1.1595,
+      "step": 5171
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017593274025013656,
+      "loss": 1.2864,
+      "step": 5172
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017592260215790583,
+      "loss": 1.2802,
+      "step": 5173
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017591246222307413,
+      "loss": 2.3772,
+      "step": 5174
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001759023204458875,
+      "loss": 1.2167,
+      "step": 5175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017589217682659208,
+      "loss": 1.1535,
+      "step": 5176
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017588203136543408,
+      "loss": 1.3127,
+      "step": 5177
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001758718840626597,
+      "loss": 1.3194,
+      "step": 5178
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001758617349185152,
+      "loss": 1.2758,
+      "step": 5179
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017585158393324695,
+      "loss": 1.1988,
+      "step": 5180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017584143110710128,
+      "loss": 1.2991,
+      "step": 5181
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017583127644032453,
+      "loss": 1.1924,
+      "step": 5182
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017582111993316322,
+      "loss": 1.3438,
+      "step": 5183
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017581096158586384,
+      "loss": 1.0718,
+      "step": 5184
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017580080139867292,
+      "loss": 1.2335,
+      "step": 5185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017579063937183702,
+      "loss": 1.1085,
+      "step": 5186
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001757804755056028,
+      "loss": 1.2062,
+      "step": 5187
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001757703098002169,
+      "loss": 1.1609,
+      "step": 5188
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017576014225592602,
+      "loss": 1.3203,
+      "step": 5189
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000175749972872977,
+      "loss": 1.3167,
+      "step": 5190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017573980165161657,
+      "loss": 1.1061,
+      "step": 5191
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001757296285920916,
+      "loss": 1.1916,
+      "step": 5192
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017571945369464903,
+      "loss": 1.227,
+      "step": 5193
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017570927695953573,
+      "loss": 1.2552,
+      "step": 5194
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017569909838699874,
+      "loss": 1.4153,
+      "step": 5195
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017568891797728505,
+      "loss": 1.0921,
+      "step": 5196
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017567873573064174,
+      "loss": 1.1415,
+      "step": 5197
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017566855164731596,
+      "loss": 1.1208,
+      "step": 5198
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001756583657275548,
+      "loss": 1.1176,
+      "step": 5199
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017564817797160553,
+      "loss": 1.2933,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017563798837971542,
+      "loss": 1.1808,
+      "step": 5201
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017562779695213168,
+      "loss": 1.1878,
+      "step": 5202
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017561760368910173,
+      "loss": 1.3654,
+      "step": 5203
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001756074085908729,
+      "loss": 1.2805,
+      "step": 5204
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017559721165769267,
+      "loss": 1.3679,
+      "step": 5205
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001755870128898085,
+      "loss": 1.2051,
+      "step": 5206
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017557681228746788,
+      "loss": 1.2127,
+      "step": 5207
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001755666098509184,
+      "loss": 1.1964,
+      "step": 5208
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017555640558040764,
+      "loss": 1.3053,
+      "step": 5209
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001755461994761833,
+      "loss": 1.1155,
+      "step": 5210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017553599153849304,
+      "loss": 1.3019,
+      "step": 5211
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001755257817675846,
+      "loss": 1.2158,
+      "step": 5212
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017551557016370577,
+      "loss": 1.3304,
+      "step": 5213
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001755053567271044,
+      "loss": 1.3587,
+      "step": 5214
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017549514145802836,
+      "loss": 1.2199,
+      "step": 5215
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017548492435672552,
+      "loss": 0.9611,
+      "step": 5216
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001754747054234439,
+      "loss": 2.4688,
+      "step": 5217
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001754644846584315,
+      "loss": 1.3337,
+      "step": 5218
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017545426206193638,
+      "loss": 1.0891,
+      "step": 5219
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001754440376342066,
+      "loss": 1.1449,
+      "step": 5220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001754338113754903,
+      "loss": 1.2681,
+      "step": 5221
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017542358328603573,
+      "loss": 0.9637,
+      "step": 5222
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017541335336609105,
+      "loss": 1.2767,
+      "step": 5223
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017540312161590458,
+      "loss": 1.1705,
+      "step": 5224
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001753928880357246,
+      "loss": 1.1837,
+      "step": 5225
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001753826526257995,
+      "loss": 1.2622,
+      "step": 5226
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001753724153863777,
+      "loss": 1.2894,
+      "step": 5227
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017536217631770762,
+      "loss": 1.1603,
+      "step": 5228
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017535193542003778,
+      "loss": 1.1848,
+      "step": 5229
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017534169269361671,
+      "loss": 1.4119,
+      "step": 5230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000175331448138693,
+      "loss": 1.2621,
+      "step": 5231
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017532120175551527,
+      "loss": 1.381,
+      "step": 5232
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001753109535443322,
+      "loss": 1.4758,
+      "step": 5233
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017530070350539254,
+      "loss": 1.2286,
+      "step": 5234
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000175290451638945,
+      "loss": 1.0692,
+      "step": 5235
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017528019794523842,
+      "loss": 1.2321,
+      "step": 5236
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017526994242452165,
+      "loss": 1.3016,
+      "step": 5237
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017525968507704355,
+      "loss": 1.0732,
+      "step": 5238
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017524942590305313,
+      "loss": 1.3669,
+      "step": 5239
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001752391649027993,
+      "loss": 1.4422,
+      "step": 5240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017522890207653116,
+      "loss": 0.9535,
+      "step": 5241
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017521863742449772,
+      "loss": 1.1277,
+      "step": 5242
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017520837094694812,
+      "loss": 1.2435,
+      "step": 5243
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017519810264413153,
+      "loss": 1.1671,
+      "step": 5244
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017518783251629718,
+      "loss": 1.4556,
+      "step": 5245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017517756056369428,
+      "loss": 2.5142,
+      "step": 5246
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001751672867865721,
+      "loss": 1.2231,
+      "step": 5247
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017515701118518004,
+      "loss": 2.5234,
+      "step": 5248
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00017514673375976746,
+      "loss": 1.1101,
+      "step": 5249
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001751364545105838,
+      "loss": 1.2549,
+      "step": 5250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001751261734378785,
+      "loss": 2.3723,
+      "step": 5251
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017511589054190112,
+      "loss": 1.3762,
+      "step": 5252
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001751056058229012,
+      "loss": 1.2953,
+      "step": 5253
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001750953192811283,
+      "loss": 1.209,
+      "step": 5254
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017508503091683215,
+      "loss": 1.1942,
+      "step": 5255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001750747407302624,
+      "loss": 1.2429,
+      "step": 5256
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001750644487216688,
+      "loss": 1.1967,
+      "step": 5257
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017505415489130107,
+      "loss": 1.0613,
+      "step": 5258
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001750438592394091,
+      "loss": 1.3844,
+      "step": 5259
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001750335617662428,
+      "loss": 1.3502,
+      "step": 5260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000175023262472052,
+      "loss": 1.2053,
+      "step": 5261
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017501296135708668,
+      "loss": 1.2505,
+      "step": 5262
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017500265842159687,
+      "loss": 1.299,
+      "step": 5263
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017499235366583258,
+      "loss": 1.2754,
+      "step": 5264
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017498204709004393,
+      "loss": 1.0926,
+      "step": 5265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017497173869448105,
+      "loss": 1.2744,
+      "step": 5266
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017496142847939412,
+      "loss": 1.316,
+      "step": 5267
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017495111644503334,
+      "loss": 1.1947,
+      "step": 5268
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000174940802591649,
+      "loss": 1.1178,
+      "step": 5269
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017493048691949142,
+      "loss": 1.3052,
+      "step": 5270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017492016942881092,
+      "loss": 1.2346,
+      "step": 5271
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017490985011985795,
+      "loss": 1.3471,
+      "step": 5272
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001748995289928829,
+      "loss": 1.2383,
+      "step": 5273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017488920604813633,
+      "loss": 1.0516,
+      "step": 5274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001748788812858687,
+      "loss": 1.0534,
+      "step": 5275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001748685547063306,
+      "loss": 1.3375,
+      "step": 5276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017485822630977266,
+      "loss": 1.0579,
+      "step": 5277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017484789609644556,
+      "loss": 1.2015,
+      "step": 5278
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017483756406660002,
+      "loss": 1.3225,
+      "step": 5279
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017482723022048676,
+      "loss": 1.2555,
+      "step": 5280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001748168945583566,
+      "loss": 1.1495,
+      "step": 5281
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017480655708046033,
+      "loss": 2.3921,
+      "step": 5282
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017479621778704888,
+      "loss": 1.29,
+      "step": 5283
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001747858766783732,
+      "loss": 1.1498,
+      "step": 5284
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017477553375468423,
+      "loss": 1.3059,
+      "step": 5285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000174765189016233,
+      "loss": 1.1014,
+      "step": 5286
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017475484246327056,
+      "loss": 1.3912,
+      "step": 5287
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017474449409604802,
+      "loss": 1.1797,
+      "step": 5288
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017473414391481652,
+      "loss": 1.2206,
+      "step": 5289
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001747237919198273,
+      "loss": 1.2987,
+      "step": 5290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001747134381113315,
+      "loss": 1.2218,
+      "step": 5291
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001747030824895805,
+      "loss": 2.5657,
+      "step": 5292
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017469272505482556,
+      "loss": 1.3002,
+      "step": 5293
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001746823658073181,
+      "loss": 1.1224,
+      "step": 5294
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017467200474730952,
+      "loss": 1.1898,
+      "step": 5295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017466164187505124,
+      "loss": 1.1461,
+      "step": 5296
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001746512771907948,
+      "loss": 1.264,
+      "step": 5297
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017464091069479176,
+      "loss": 1.0997,
+      "step": 5298
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017463054238729363,
+      "loss": 1.086,
+      "step": 5299
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017462017226855213,
+      "loss": 2.2666,
+      "step": 5300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001746098003388189,
+      "loss": 1.3552,
+      "step": 5301
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017459942659834568,
+      "loss": 1.2827,
+      "step": 5302
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001745890510473842,
+      "loss": 1.2468,
+      "step": 5303
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001745786736861863,
+      "loss": 1.4524,
+      "step": 5304
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017456829451500383,
+      "loss": 1.2624,
+      "step": 5305
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017455791353408867,
+      "loss": 1.2655,
+      "step": 5306
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017454753074369275,
+      "loss": 1.1163,
+      "step": 5307
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001745371461440681,
+      "loss": 1.191,
+      "step": 5308
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001745267597354667,
+      "loss": 1.1855,
+      "step": 5309
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017451637151814068,
+      "loss": 2.5537,
+      "step": 5310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001745059814923421,
+      "loss": 1.0242,
+      "step": 5311
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017449558965832313,
+      "loss": 1.1977,
+      "step": 5312
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017448519601633602,
+      "loss": 1.1319,
+      "step": 5313
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017447480056663296,
+      "loss": 1.0938,
+      "step": 5314
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001744644033094663,
+      "loss": 1.1143,
+      "step": 5315
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001744540042450883,
+      "loss": 1.2213,
+      "step": 5316
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017444360337375142,
+      "loss": 1.1207,
+      "step": 5317
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017443320069570803,
+      "loss": 1.2456,
+      "step": 5318
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017442279621121063,
+      "loss": 1.2089,
+      "step": 5319
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001744123899205117,
+      "loss": 1.1655,
+      "step": 5320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001744019818238638,
+      "loss": 1.2482,
+      "step": 5321
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017439157192151956,
+      "loss": 1.2599,
+      "step": 5322
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017438116021373161,
+      "loss": 1.1393,
+      "step": 5323
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001743707467007526,
+      "loss": 1.1483,
+      "step": 5324
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001743603313828353,
+      "loss": 1.2362,
+      "step": 5325
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017434991426023248,
+      "loss": 1.1442,
+      "step": 5326
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017433949533319693,
+      "loss": 1.3278,
+      "step": 5327
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017432907460198155,
+      "loss": 1.1153,
+      "step": 5328
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017431865206683924,
+      "loss": 1.1389,
+      "step": 5329
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017430822772802293,
+      "loss": 1.2259,
+      "step": 5330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017429780158578564,
+      "loss": 1.0565,
+      "step": 5331
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017428737364038035,
+      "loss": 1.2478,
+      "step": 5332
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017427694389206023,
+      "loss": 1.4247,
+      "step": 5333
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001742665123410783,
+      "loss": 1.1948,
+      "step": 5334
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017425607898768778,
+      "loss": 1.4145,
+      "step": 5335
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001742456438321419,
+      "loss": 1.2482,
+      "step": 5336
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017423520687469393,
+      "loss": 1.1084,
+      "step": 5337
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017422476811559712,
+      "loss": 1.0502,
+      "step": 5338
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017421432755510482,
+      "loss": 0.9462,
+      "step": 5339
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017420388519347042,
+      "loss": 1.1794,
+      "step": 5340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017419344103094736,
+      "loss": 1.1313,
+      "step": 5341
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017418299506778915,
+      "loss": 1.1103,
+      "step": 5342
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017417254730424926,
+      "loss": 1.1345,
+      "step": 5343
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017416209774058123,
+      "loss": 1.2073,
+      "step": 5344
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017415164637703875,
+      "loss": 1.2393,
+      "step": 5345
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017414119321387537,
+      "loss": 2.3191,
+      "step": 5346
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017413073825134484,
+      "loss": 1.2882,
+      "step": 5347
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017412028148970087,
+      "loss": 1.37,
+      "step": 5348
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001741098229291973,
+      "loss": 1.2123,
+      "step": 5349
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017409936257008787,
+      "loss": 1.2799,
+      "step": 5350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001740889004126265,
+      "loss": 1.3898,
+      "step": 5351
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017407843645706708,
+      "loss": 1.3471,
+      "step": 5352
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001740679707036636,
+      "loss": 1.2681,
+      "step": 5353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017405750315267003,
+      "loss": 2.4238,
+      "step": 5354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017404703380434039,
+      "loss": 1.1377,
+      "step": 5355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001740365626589288,
+      "loss": 0.9976,
+      "step": 5356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017402608971668938,
+      "loss": 1.1479,
+      "step": 5357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017401561497787631,
+      "loss": 1.2591,
+      "step": 5358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017400513844274379,
+      "loss": 1.2635,
+      "step": 5359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017399466011154603,
+      "loss": 0.9938,
+      "step": 5360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017398417998453747,
+      "loss": 1.3062,
+      "step": 5361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017397369806197236,
+      "loss": 1.431,
+      "step": 5362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017396321434410506,
+      "loss": 1.2531,
+      "step": 5363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001739527288311901,
+      "loss": 1.118,
+      "step": 5364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001739422415234819,
+      "loss": 1.2953,
+      "step": 5365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000173931752421235,
+      "loss": 1.3486,
+      "step": 5366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017392126152470392,
+      "loss": 1.2197,
+      "step": 5367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017391076883414332,
+      "loss": 1.1778,
+      "step": 5368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017390027434980784,
+      "loss": 1.2204,
+      "step": 5369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017388977807195218,
+      "loss": 1.2909,
+      "step": 5370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017387928000083108,
+      "loss": 1.0848,
+      "step": 5371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017386878013669928,
+      "loss": 1.2302,
+      "step": 5372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017385827847981165,
+      "loss": 1.4387,
+      "step": 5373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001738477750304231,
+      "loss": 1.3856,
+      "step": 5374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017383726978878843,
+      "loss": 1.1193,
+      "step": 5375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001738267627551627,
+      "loss": 1.1444,
+      "step": 5376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017381625392980083,
+      "loss": 1.2898,
+      "step": 5377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017380574331295792,
+      "loss": 1.2783,
+      "step": 5378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017379523090488904,
+      "loss": 1.3107,
+      "step": 5379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017378471670584933,
+      "loss": 1.0708,
+      "step": 5380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017377420071609394,
+      "loss": 1.2294,
+      "step": 5381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001737636829358781,
+      "loss": 1.2783,
+      "step": 5382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017375316336545708,
+      "loss": 1.0645,
+      "step": 5383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001737426420050862,
+      "loss": 1.3773,
+      "step": 5384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017373211885502075,
+      "loss": 1.1674,
+      "step": 5385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017372159391551616,
+      "loss": 1.1635,
+      "step": 5386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017371106718682788,
+      "loss": 1.0566,
+      "step": 5387
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017370053866921134,
+      "loss": 1.0118,
+      "step": 5388
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017369000836292214,
+      "loss": 1.2433,
+      "step": 5389
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017367947626821574,
+      "loss": 1.2445,
+      "step": 5390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017366894238534782,
+      "loss": 1.222,
+      "step": 5391
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017365840671457406,
+      "loss": 1.3102,
+      "step": 5392
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017364786925615005,
+      "loss": 1.4504,
+      "step": 5393
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001736373300103316,
+      "loss": 1.2059,
+      "step": 5394
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017362678897737453,
+      "loss": 1.232,
+      "step": 5395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017361624615753457,
+      "loss": 1.2577,
+      "step": 5396
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001736057015510676,
+      "loss": 0.9839,
+      "step": 5397
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017359515515822963,
+      "loss": 1.1353,
+      "step": 5398
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017358460697927653,
+      "loss": 1.1553,
+      "step": 5399
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017357405701446435,
+      "loss": 1.0069,
+      "step": 5400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017356350526404906,
+      "loss": 1.397,
+      "step": 5401
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017355295172828678,
+      "loss": 1.1579,
+      "step": 5402
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017354239640743368,
+      "loss": 1.2803,
+      "step": 5403
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001735318393017459,
+      "loss": 2.4141,
+      "step": 5404
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017352128041147965,
+      "loss": 1.3854,
+      "step": 5405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017351071973689116,
+      "loss": 1.1876,
+      "step": 5406
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001735001572782368,
+      "loss": 1.0125,
+      "step": 5407
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017348959303577287,
+      "loss": 2.1641,
+      "step": 5408
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017347902700975574,
+      "loss": 1.1684,
+      "step": 5409
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001734684592004419,
+      "loss": 1.1804,
+      "step": 5410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001734578896080878,
+      "loss": 2.3877,
+      "step": 5411
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017344731823294994,
+      "loss": 1.1749,
+      "step": 5412
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001734367450752849,
+      "loss": 1.2654,
+      "step": 5413
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001734261701353493,
+      "loss": 2.2151,
+      "step": 5414
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017341559341339972,
+      "loss": 1.1925,
+      "step": 5415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017340501490969293,
+      "loss": 1.2312,
+      "step": 5416
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017339443462448563,
+      "loss": 1.264,
+      "step": 5417
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017338385255803463,
+      "loss": 1.0847,
+      "step": 5418
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017337326871059668,
+      "loss": 1.1078,
+      "step": 5419
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017336268308242869,
+      "loss": 1.3168,
+      "step": 5420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001733520956737876,
+      "loss": 1.2061,
+      "step": 5421
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001733415064849303,
+      "loss": 1.1584,
+      "step": 5422
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001733309155161138,
+      "loss": 1.2607,
+      "step": 5423
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017332032276759515,
+      "loss": 1.3163,
+      "step": 5424
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017330972823963148,
+      "loss": 1.3476,
+      "step": 5425
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001732991319324798,
+      "loss": 1.3163,
+      "step": 5426
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017328853384639733,
+      "loss": 1.0358,
+      "step": 5427
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017327793398164132,
+      "loss": 1.3297,
+      "step": 5428
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017326733233846898,
+      "loss": 1.1896,
+      "step": 5429
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001732567289171376,
+      "loss": 1.2468,
+      "step": 5430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001732461237179046,
+      "loss": 1.0773,
+      "step": 5431
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017323551674102724,
+      "loss": 1.2346,
+      "step": 5432
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017322490798676304,
+      "loss": 1.2128,
+      "step": 5433
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001732142974553694,
+      "loss": 1.2306,
+      "step": 5434
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017320368514710385,
+      "loss": 1.0887,
+      "step": 5435
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000173193071062224,
+      "loss": 1.0359,
+      "step": 5436
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001731824552009874,
+      "loss": 1.231,
+      "step": 5437
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001731718375636517,
+      "loss": 1.4084,
+      "step": 5438
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017316121815047456,
+      "loss": 1.2637,
+      "step": 5439
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017315059696171377,
+      "loss": 1.1817,
+      "step": 5440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017313997399762701,
+      "loss": 1.1866,
+      "step": 5441
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001731293492584722,
+      "loss": 1.2971,
+      "step": 5442
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001731187227445071,
+      "loss": 1.2343,
+      "step": 5443
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017310809445598968,
+      "loss": 1.2396,
+      "step": 5444
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017309746439317784,
+      "loss": 2.3179,
+      "step": 5445
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001730868325563296,
+      "loss": 1.2505,
+      "step": 5446
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017307619894570294,
+      "loss": 1.2649,
+      "step": 5447
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017306556356155598,
+      "loss": 1.1763,
+      "step": 5448
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017305492640414682,
+      "loss": 1.2139,
+      "step": 5449
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001730442874737336,
+      "loss": 1.1868,
+      "step": 5450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017303364677057455,
+      "loss": 1.1931,
+      "step": 5451
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001730230042949279,
+      "loss": 1.3549,
+      "step": 5452
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017301236004705194,
+      "loss": 1.115,
+      "step": 5453
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017300171402720498,
+      "loss": 1.2478,
+      "step": 5454
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017299106623564544,
+      "loss": 1.1393,
+      "step": 5455
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017298041667263166,
+      "loss": 1.188,
+      "step": 5456
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001729697653384222,
+      "loss": 1.1622,
+      "step": 5457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017295911223327547,
+      "loss": 1.1596,
+      "step": 5458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017294845735745006,
+      "loss": 1.1924,
+      "step": 5459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017293780071120458,
+      "loss": 1.2096,
+      "step": 5460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001729271422947976,
+      "loss": 1.1659,
+      "step": 5461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017291648210848785,
+      "loss": 1.2138,
+      "step": 5462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017290582015253402,
+      "loss": 1.1649,
+      "step": 5463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017289515642719488,
+      "loss": 1.1996,
+      "step": 5464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001728844909327292,
+      "loss": 0.9867,
+      "step": 5465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017287382366939585,
+      "loss": 1.1179,
+      "step": 5466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017286315463745376,
+      "loss": 1.2618,
+      "step": 5467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017285248383716183,
+      "loss": 1.2563,
+      "step": 5468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017284181126877898,
+      "loss": 1.3058,
+      "step": 5469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017283113693256433,
+      "loss": 1.2897,
+      "step": 5470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017282046082877682,
+      "loss": 1.291,
+      "step": 5471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017280978295767567,
+      "loss": 1.176,
+      "step": 5472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017279910331951997,
+      "loss": 2.5659,
+      "step": 5473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001727884219145689,
+      "loss": 1.2986,
+      "step": 5474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001727777387430817,
+      "loss": 1.2261,
+      "step": 5475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017276705380531764,
+      "loss": 1.2143,
+      "step": 5476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001727563671015361,
+      "loss": 1.2122,
+      "step": 5477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017274567863199636,
+      "loss": 1.2645,
+      "step": 5478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017273498839695788,
+      "loss": 1.0641,
+      "step": 5479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017272429639668004,
+      "loss": 1.1673,
+      "step": 5480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017271360263142237,
+      "loss": 1.3346,
+      "step": 5481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017270290710144444,
+      "loss": 1.416,
+      "step": 5482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017269220980700576,
+      "loss": 1.1262,
+      "step": 5483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017268151074836596,
+      "loss": 1.2135,
+      "step": 5484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017267080992578474,
+      "loss": 1.1641,
+      "step": 5485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017266010733952176,
+      "loss": 1.318,
+      "step": 5486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001726494029898368,
+      "loss": 2.4968,
+      "step": 5487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017263869687698962,
+      "loss": 1.2078,
+      "step": 5488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017262798900124007,
+      "loss": 1.2393,
+      "step": 5489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000172617279362848,
+      "loss": 1.1497,
+      "step": 5490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017260656796207335,
+      "loss": 1.3568,
+      "step": 5491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017259585479917607,
+      "loss": 2.4724,
+      "step": 5492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017258513987441617,
+      "loss": 1.2087,
+      "step": 5493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017257442318805374,
+      "loss": 1.2477,
+      "step": 5494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017256370474034877,
+      "loss": 0.9953,
+      "step": 5495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017255298453156144,
+      "loss": 1.2678,
+      "step": 5496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017254226256195195,
+      "loss": 1.1594,
+      "step": 5497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017253153883178048,
+      "loss": 2.467,
+      "step": 5498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017252081334130732,
+      "loss": 1.3302,
+      "step": 5499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017251008609079277,
+      "loss": 1.2256,
+      "step": 5500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001724993570804971,
+      "loss": 1.1601,
+      "step": 5501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001724886263106808,
+      "loss": 1.0725,
+      "step": 5502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017247789378160424,
+      "loss": 1.335,
+      "step": 5503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001724671594935279,
+      "loss": 1.0708,
+      "step": 5504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017245642344671232,
+      "loss": 1.1416,
+      "step": 5505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017244568564141804,
+      "loss": 2.2322,
+      "step": 5506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017243494607790564,
+      "loss": 1.1849,
+      "step": 5507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001724242047564358,
+      "loss": 1.0557,
+      "step": 5508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001724134616772692,
+      "loss": 0.8833,
+      "step": 5509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017240271684066657,
+      "loss": 1.3566,
+      "step": 5510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017239197024688867,
+      "loss": 1.2248,
+      "step": 5511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001723812218961963,
+      "loss": 1.1866,
+      "step": 5512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017237047178885034,
+      "loss": 1.3046,
+      "step": 5513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001723597199251117,
+      "loss": 1.2167,
+      "step": 5514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001723489663052413,
+      "loss": 1.4368,
+      "step": 5515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017233821092950012,
+      "loss": 1.2546,
+      "step": 5516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017232745379814918,
+      "loss": 1.1418,
+      "step": 5517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017231669491144959,
+      "loss": 1.2562,
+      "step": 5518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017230593426966246,
+      "loss": 1.405,
+      "step": 5519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017229517187304888,
+      "loss": 1.2171,
+      "step": 5520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001722844077218701,
+      "loss": 1.298,
+      "step": 5521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017227364181638736,
+      "loss": 1.0184,
+      "step": 5522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017226287415686198,
+      "loss": 1.2621,
+      "step": 5523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017225210474355516,
+      "loss": 1.3475,
+      "step": 5524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001722413335767284,
+      "loss": 2.4812,
+      "step": 5525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017223056065664304,
+      "loss": 1.1498,
+      "step": 5526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017221978598356056,
+      "loss": 1.1728,
+      "step": 5527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017220900955774245,
+      "loss": 1.3199,
+      "step": 5528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017219823137945025,
+      "loss": 1.192,
+      "step": 5529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017218745144894552,
+      "loss": 1.3297,
+      "step": 5530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001721766697664899,
+      "loss": 1.1842,
+      "step": 5531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017216588633234506,
+      "loss": 1.1796,
+      "step": 5532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001721551011467727,
+      "loss": 1.1299,
+      "step": 5533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017214431421003456,
+      "loss": 1.2247,
+      "step": 5534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017213352552239246,
+      "loss": 1.2971,
+      "step": 5535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017212273508410822,
+      "loss": 1.2832,
+      "step": 5536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017211194289544374,
+      "loss": 1.1061,
+      "step": 5537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001721011489566609,
+      "loss": 1.041,
+      "step": 5538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017209035326802167,
+      "loss": 1.1678,
+      "step": 5539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017207955582978812,
+      "loss": 1.1834,
+      "step": 5540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001720687566422222,
+      "loss": 1.1587,
+      "step": 5541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017205795570558608,
+      "loss": 1.0467,
+      "step": 5542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017204715302014182,
+      "loss": 1.1276,
+      "step": 5543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017203634858615167,
+      "loss": 1.1913,
+      "step": 5544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017202554240387783,
+      "loss": 1.1819,
+      "step": 5545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017201473447358252,
+      "loss": 1.1994,
+      "step": 5546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017200392479552808,
+      "loss": 1.2119,
+      "step": 5547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017199311336997682,
+      "loss": 1.1843,
+      "step": 5548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017198230019719118,
+      "loss": 1.1725,
+      "step": 5549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017197148527743354,
+      "loss": 1.2169,
+      "step": 5550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001719606686109664,
+      "loss": 1.0442,
+      "step": 5551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017194985019805225,
+      "loss": 1.1552,
+      "step": 5552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001719390300389537,
+      "loss": 1.3141,
+      "step": 5553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017192820813393328,
+      "loss": 2.3367,
+      "step": 5554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017191738448325367,
+      "loss": 1.2444,
+      "step": 5555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017190655908717756,
+      "loss": 1.2686,
+      "step": 5556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017189573194596766,
+      "loss": 1.3022,
+      "step": 5557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017188490305988677,
+      "loss": 1.308,
+      "step": 5558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017187407242919765,
+      "loss": 1.0668,
+      "step": 5559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001718632400541632,
+      "loss": 1.2267,
+      "step": 5560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017185240593504627,
+      "loss": 1.1796,
+      "step": 5561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017184157007210985,
+      "loss": 1.4653,
+      "step": 5562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001718307324656169,
+      "loss": 1.192,
+      "step": 5563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001718198931158304,
+      "loss": 1.2986,
+      "step": 5564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017180905202301348,
+      "loss": 1.1883,
+      "step": 5565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001717982091874292,
+      "loss": 1.2134,
+      "step": 5566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017178736460934078,
+      "loss": 1.2043,
+      "step": 5567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017177651828901132,
+      "loss": 1.2927,
+      "step": 5568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001717656702267041,
+      "loss": 1.1287,
+      "step": 5569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001717548204226824,
+      "loss": 1.1991,
+      "step": 5570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017174396887720955,
+      "loss": 1.0583,
+      "step": 5571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017173311559054888,
+      "loss": 1.1338,
+      "step": 5572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001717222605629638,
+      "loss": 1.3499,
+      "step": 5573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017171140379471778,
+      "loss": 2.3115,
+      "step": 5574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017170054528607427,
+      "loss": 1.1712,
+      "step": 5575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017168968503729684,
+      "loss": 1.338,
+      "step": 5576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017167882304864904,
+      "loss": 1.1443,
+      "step": 5577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017166795932039447,
+      "loss": 1.0833,
+      "step": 5578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017165709385279682,
+      "loss": 1.0784,
+      "step": 5579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001716462266461198,
+      "loss": 1.2051,
+      "step": 5580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001716353577006271,
+      "loss": 2.6755,
+      "step": 5581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017162448701658252,
+      "loss": 1.3716,
+      "step": 5582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001716136145942499,
+      "loss": 1.1472,
+      "step": 5583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017160274043389313,
+      "loss": 1.1884,
+      "step": 5584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017159186453577605,
+      "loss": 1.2071,
+      "step": 5585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001715809869001627,
+      "loss": 1.3563,
+      "step": 5586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000171570107527317,
+      "loss": 1.1628,
+      "step": 5587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017155922641750302,
+      "loss": 1.3672,
+      "step": 5588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017154834357098483,
+      "loss": 1.0411,
+      "step": 5589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017153745898802658,
+      "loss": 1.3306,
+      "step": 5590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001715265726688924,
+      "loss": 1.1623,
+      "step": 5591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017151568461384648,
+      "loss": 1.3103,
+      "step": 5592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001715047948231531,
+      "loss": 1.2005,
+      "step": 5593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017149390329707656,
+      "loss": 1.1858,
+      "step": 5594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017148301003588117,
+      "loss": 1.0792,
+      "step": 5595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001714721150398313,
+      "loss": 1.1471,
+      "step": 5596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017146121830919134,
+      "loss": 1.3057,
+      "step": 5597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017145031984422583,
+      "loss": 1.3274,
+      "step": 5598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017143941964519922,
+      "loss": 1.2047,
+      "step": 5599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000171428517712376,
+      "loss": 1.4583,
+      "step": 5600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001714176140460208,
+      "loss": 1.2228,
+      "step": 5601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017140670864639832,
+      "loss": 1.3172,
+      "step": 5602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017139580151377313,
+      "loss": 1.2224,
+      "step": 5603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017138489264840995,
+      "loss": 1.1976,
+      "step": 5604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017137398205057357,
+      "loss": 1.1887,
+      "step": 5605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017136306972052875,
+      "loss": 1.1178,
+      "step": 5606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017135215565854036,
+      "loss": 1.2655,
+      "step": 5607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017134123986487328,
+      "loss": 1.2698,
+      "step": 5608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001713303223397924,
+      "loss": 1.3349,
+      "step": 5609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017131940308356267,
+      "loss": 1.0988,
+      "step": 5610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017130848209644914,
+      "loss": 1.1457,
+      "step": 5611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017129755937871682,
+      "loss": 1.0781,
+      "step": 5612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017128663493063084,
+      "loss": 1.275,
+      "step": 5613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017127570875245628,
+      "loss": 1.1221,
+      "step": 5614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017126478084445837,
+      "loss": 1.0955,
+      "step": 5615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017125385120690228,
+      "loss": 1.1814,
+      "step": 5616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017124291984005322,
+      "loss": 1.3641,
+      "step": 5617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017123198674417662,
+      "loss": 1.2238,
+      "step": 5618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017122105191953773,
+      "loss": 1.364,
+      "step": 5619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017121011536640193,
+      "loss": 1.2424,
+      "step": 5620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017119917708503465,
+      "loss": 1.2284,
+      "step": 5621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017118823707570142,
+      "loss": 1.1742,
+      "step": 5622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017117729533866767,
+      "loss": 1.2376,
+      "step": 5623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000171166351874199,
+      "loss": 1.2234,
+      "step": 5624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017115540668256093,
+      "loss": 1.1793,
+      "step": 5625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017114445976401917,
+      "loss": 1.3059,
+      "step": 5626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001711335111188394,
+      "loss": 1.1067,
+      "step": 5627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001711225607472873,
+      "loss": 0.9545,
+      "step": 5628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017111160864962858,
+      "loss": 1.1975,
+      "step": 5629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017110065482612915,
+      "loss": 1.2517,
+      "step": 5630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001710896992770548,
+      "loss": 1.1431,
+      "step": 5631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017107874200267143,
+      "loss": 1.3216,
+      "step": 5632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001710677830032449,
+      "loss": 1.2612,
+      "step": 5633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017105682227904127,
+      "loss": 1.2457,
+      "step": 5634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017104585983032653,
+      "loss": 1.2178,
+      "step": 5635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001710348956573667,
+      "loss": 0.9617,
+      "step": 5636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001710239297604279,
+      "loss": 1.4252,
+      "step": 5637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017101296213977625,
+      "loss": 1.1539,
+      "step": 5638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017100199279567794,
+      "loss": 1.1982,
+      "step": 5639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001709910217283992,
+      "loss": 1.1441,
+      "step": 5640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017098004893820626,
+      "loss": 1.2013,
+      "step": 5641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017096907442536546,
+      "loss": 1.1964,
+      "step": 5642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001709580981901431,
+      "loss": 1.1816,
+      "step": 5643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001709471202328056,
+      "loss": 1.2634,
+      "step": 5644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017093614055361942,
+      "loss": 1.3004,
+      "step": 5645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000170925159152851,
+      "loss": 1.3336,
+      "step": 5646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017091417603076677,
+      "loss": 1.1562,
+      "step": 5647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017090319118763346,
+      "loss": 1.1575,
+      "step": 5648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017089220462371754,
+      "loss": 1.1673,
+      "step": 5649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017088121633928563,
+      "loss": 1.2846,
+      "step": 5650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001708702263346045,
+      "loss": 1.3701,
+      "step": 5651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017085923460994083,
+      "loss": 1.4522,
+      "step": 5652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017084824116556137,
+      "loss": 1.1512,
+      "step": 5653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017083724600173295,
+      "loss": 1.1535,
+      "step": 5654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017082624911872238,
+      "loss": 1.0747,
+      "step": 5655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001708152505167966,
+      "loss": 1.0602,
+      "step": 5656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001708042501962225,
+      "loss": 1.1339,
+      "step": 5657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017079324815726706,
+      "loss": 1.2242,
+      "step": 5658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001707822444001973,
+      "loss": 1.2923,
+      "step": 5659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017077123892528029,
+      "loss": 1.1588,
+      "step": 5660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017076023173278308,
+      "loss": 1.2915,
+      "step": 5661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017074922282297287,
+      "loss": 1.1741,
+      "step": 5662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017073821219611677,
+      "loss": 1.1058,
+      "step": 5663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001707271998524821,
+      "loss": 1.0092,
+      "step": 5664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000170716185792336,
+      "loss": 1.1803,
+      "step": 5665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017070517001594587,
+      "loss": 2.604,
+      "step": 5666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017069415252357904,
+      "loss": 1.3611,
+      "step": 5667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001706831333155029,
+      "loss": 1.2738,
+      "step": 5668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017067211239198485,
+      "loss": 1.304,
+      "step": 5669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017066108975329235,
+      "loss": 1.2279,
+      "step": 5670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017065006539969296,
+      "loss": 1.2089,
+      "step": 5671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017063903933145426,
+      "loss": 1.1927,
+      "step": 5672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017062801154884376,
+      "loss": 1.0803,
+      "step": 5673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017061698205212916,
+      "loss": 1.2323,
+      "step": 5674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017060595084157808,
+      "loss": 1.08,
+      "step": 5675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001705949179174583,
+      "loss": 1.1215,
+      "step": 5676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017058388328003763,
+      "loss": 1.1955,
+      "step": 5677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017057284692958375,
+      "loss": 1.3702,
+      "step": 5678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017056180886636458,
+      "loss": 1.2048,
+      "step": 5679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000170550769090648,
+      "loss": 1.2413,
+      "step": 5680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017053972760270193,
+      "loss": 1.2696,
+      "step": 5681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017052868440279434,
+      "loss": 1.1167,
+      "step": 5682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017051763949119328,
+      "loss": 1.189,
+      "step": 5683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017050659286816676,
+      "loss": 2.3037,
+      "step": 5684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017049554453398287,
+      "loss": 1.0556,
+      "step": 5685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017048449448890977,
+      "loss": 1.1472,
+      "step": 5686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017047344273321565,
+      "loss": 1.154,
+      "step": 5687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001704623892671687,
+      "loss": 1.2444,
+      "step": 5688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001704513340910372,
+      "loss": 1.1256,
+      "step": 5689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017044027720508948,
+      "loss": 1.1501,
+      "step": 5690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017042921860959382,
+      "loss": 1.1137,
+      "step": 5691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017041815830481867,
+      "loss": 1.239,
+      "step": 5692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001704070962910324,
+      "loss": 1.2493,
+      "step": 5693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017039603256850352,
+      "loss": 1.2264,
+      "step": 5694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017038496713750055,
+      "loss": 1.1516,
+      "step": 5695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017037389999829196,
+      "loss": 1.1802,
+      "step": 5696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017036283115114644,
+      "loss": 1.1353,
+      "step": 5697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001703517605963326,
+      "loss": 1.2914,
+      "step": 5698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001703406883341191,
+      "loss": 1.119,
+      "step": 5699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017032961436477467,
+      "loss": 1.2983,
+      "step": 5700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017031853868856804,
+      "loss": 2.573,
+      "step": 5701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017030746130576805,
+      "loss": 1.248,
+      "step": 5702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001702963822166435,
+      "loss": 0.964,
+      "step": 5703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017028530142146334,
+      "loss": 1.2318,
+      "step": 5704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001702742189204964,
+      "loss": 1.109,
+      "step": 5705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001702631347140118,
+      "loss": 1.1005,
+      "step": 5706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017025204880227835,
+      "loss": 1.2512,
+      "step": 5707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017024096118556522,
+      "loss": 1.3965,
+      "step": 5708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001702298718641415,
+      "loss": 1.0822,
+      "step": 5709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017021878083827628,
+      "loss": 1.2847,
+      "step": 5710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017020768810823876,
+      "loss": 1.3369,
+      "step": 5711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017019659367429815,
+      "loss": 1.1958,
+      "step": 5712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001701854975367237,
+      "loss": 1.1205,
+      "step": 5713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017017439969578473,
+      "loss": 1.3698,
+      "step": 5714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017016330015175054,
+      "loss": 1.29,
+      "step": 5715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017015219890489052,
+      "loss": 1.2252,
+      "step": 5716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017014109595547413,
+      "loss": 1.246,
+      "step": 5717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001701299913037708,
+      "loss": 1.0837,
+      "step": 5718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017011888495005,
+      "loss": 1.1334,
+      "step": 5719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017010777689458134,
+      "loss": 1.1859,
+      "step": 5720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017009666713763442,
+      "loss": 1.1808,
+      "step": 5721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017008555567947877,
+      "loss": 1.2181,
+      "step": 5722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017007444252038414,
+      "loss": 1.0768,
+      "step": 5723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017006332766062018,
+      "loss": 1.1329,
+      "step": 5724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001700522111004567,
+      "loss": 1.1144,
+      "step": 5725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001700410928401635,
+      "loss": 1.1967,
+      "step": 5726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017002997288001037,
+      "loss": 2.4719,
+      "step": 5727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001700188512202672,
+      "loss": 1.3211,
+      "step": 5728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001700077278612039,
+      "loss": 1.1252,
+      "step": 5729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016999660280309046,
+      "loss": 1.0697,
+      "step": 5730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016998547604619684,
+      "loss": 1.1404,
+      "step": 5731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001699743475907931,
+      "loss": 2.4392,
+      "step": 5732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001699632174371493,
+      "loss": 1.2746,
+      "step": 5733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016995208558553558,
+      "loss": 1.2145,
+      "step": 5734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016994095203622214,
+      "loss": 2.4119,
+      "step": 5735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016992981678947913,
+      "loss": 1.2118,
+      "step": 5736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001699186798455768,
+      "loss": 1.2889,
+      "step": 5737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016990754120478548,
+      "loss": 1.2842,
+      "step": 5738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016989640086737546,
+      "loss": 1.1684,
+      "step": 5739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016988525883361713,
+      "loss": 1.2954,
+      "step": 5740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016987411510378087,
+      "loss": 1.2435,
+      "step": 5741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016986296967813716,
+      "loss": 1.1081,
+      "step": 5742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016985182255695652,
+      "loss": 1.3427,
+      "step": 5743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016984067374050944,
+      "loss": 1.3158,
+      "step": 5744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016982952322906652,
+      "loss": 1.1824,
+      "step": 5745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016981837102289835,
+      "loss": 1.3694,
+      "step": 5746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016980721712227563,
+      "loss": 1.0557,
+      "step": 5747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016979606152746902,
+      "loss": 1.1044,
+      "step": 5748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016978490423874925,
+      "loss": 1.0547,
+      "step": 5749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016977374525638714,
+      "loss": 1.1403,
+      "step": 5750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016976258458065353,
+      "loss": 1.3695,
+      "step": 5751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016975142221181924,
+      "loss": 1.1793,
+      "step": 5752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016974025815015518,
+      "loss": 1.2484,
+      "step": 5753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001697290923959323,
+      "loss": 1.0399,
+      "step": 5754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016971792494942158,
+      "loss": 1.2632,
+      "step": 5755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016970675581089406,
+      "loss": 1.2877,
+      "step": 5756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001696955849806208,
+      "loss": 1.3019,
+      "step": 5757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016968441245887296,
+      "loss": 1.2148,
+      "step": 5758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016967323824592162,
+      "loss": 1.2712,
+      "step": 5759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000169662062342038,
+      "loss": 1.0659,
+      "step": 5760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016965088474749334,
+      "loss": 1.1412,
+      "step": 5761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016963970546255891,
+      "loss": 1.1928,
+      "step": 5762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016962852448750603,
+      "loss": 1.2057,
+      "step": 5763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016961734182260603,
+      "loss": 1.2153,
+      "step": 5764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016960615746813032,
+      "loss": 1.353,
+      "step": 5765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016959497142435037,
+      "loss": 2.4421,
+      "step": 5766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016958378369153761,
+      "loss": 1.1802,
+      "step": 5767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001695725942699636,
+      "loss": 1.2979,
+      "step": 5768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001695614031598999,
+      "loss": 1.1653,
+      "step": 5769
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001695502103616181,
+      "loss": 1.2697,
+      "step": 5770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016953901587538983,
+      "loss": 2.365,
+      "step": 5771
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016952781970148677,
+      "loss": 1.1773,
+      "step": 5772
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001695166218401807,
+      "loss": 1.0561,
+      "step": 5773
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001695054222917433,
+      "loss": 1.436,
+      "step": 5774
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016949422105644644,
+      "loss": 1.0748,
+      "step": 5775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016948301813456194,
+      "loss": 1.2623,
+      "step": 5776
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016947181352636173,
+      "loss": 1.153,
+      "step": 5777
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001694606072321177,
+      "loss": 1.3046,
+      "step": 5778
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016944939925210185,
+      "loss": 1.2521,
+      "step": 5779
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016943818958658615,
+      "loss": 1.3454,
+      "step": 5780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016942697823584267,
+      "loss": 1.2047,
+      "step": 5781
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016941576520014353,
+      "loss": 1.2406,
+      "step": 5782
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016940455047976082,
+      "loss": 1.187,
+      "step": 5783
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016939333407496677,
+      "loss": 1.1407,
+      "step": 5784
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016938211598603353,
+      "loss": 1.1136,
+      "step": 5785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016937089621323339,
+      "loss": 1.1721,
+      "step": 5786
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001693596747568387,
+      "loss": 1.1914,
+      "step": 5787
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016934845161712173,
+      "loss": 1.2341,
+      "step": 5788
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016933722679435484,
+      "loss": 1.1108,
+      "step": 5789
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001693260002888105,
+      "loss": 1.3824,
+      "step": 5790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001693147721007612,
+      "loss": 1.3298,
+      "step": 5791
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016930354223047936,
+      "loss": 1.1182,
+      "step": 5792
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001692923106782376,
+      "loss": 1.2224,
+      "step": 5793
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016928107744430845,
+      "loss": 1.2994,
+      "step": 5794
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016926984252896456,
+      "loss": 1.2331,
+      "step": 5795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001692586059324786,
+      "loss": 1.1602,
+      "step": 5796
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016924736765512322,
+      "loss": 1.1519,
+      "step": 5797
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016923612769717127,
+      "loss": 1.0589,
+      "step": 5798
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016922488605889546,
+      "loss": 1.1857,
+      "step": 5799
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016921364274056862,
+      "loss": 1.258,
+      "step": 5800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016920239774246364,
+      "loss": 1.2285,
+      "step": 5801
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016919115106485346,
+      "loss": 1.4152,
+      "step": 5802
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016917990270801098,
+      "loss": 1.1627,
+      "step": 5803
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001691686526722092,
+      "loss": 1.1645,
+      "step": 5804
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016915740095772117,
+      "loss": 2.7566,
+      "step": 5805
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016914614756481994,
+      "loss": 1.2726,
+      "step": 5806
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016913489249377868,
+      "loss": 1.2708,
+      "step": 5807
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016912363574487046,
+      "loss": 1.1991,
+      "step": 5808
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016911237731836857,
+      "loss": 1.3486,
+      "step": 5809
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016910111721454615,
+      "loss": 1.1862,
+      "step": 5810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016908985543367654,
+      "loss": 1.132,
+      "step": 5811
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016907859197603305,
+      "loss": 1.2843,
+      "step": 5812
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016906732684188902,
+      "loss": 1.2803,
+      "step": 5813
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016905606003151784,
+      "loss": 1.2355,
+      "step": 5814
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016904479154519297,
+      "loss": 1.2369,
+      "step": 5815
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016903352138318792,
+      "loss": 1.0804,
+      "step": 5816
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016902224954577614,
+      "loss": 1.3048,
+      "step": 5817
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016901097603323124,
+      "loss": 1.2103,
+      "step": 5818
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016899970084582682,
+      "loss": 1.2613,
+      "step": 5819
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001689884239838365,
+      "loss": 1.2546,
+      "step": 5820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000168977145447534,
+      "loss": 1.0967,
+      "step": 5821
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000168965865237193,
+      "loss": 1.1195,
+      "step": 5822
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016895458335308727,
+      "loss": 1.2144,
+      "step": 5823
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001689432997954907,
+      "loss": 1.0834,
+      "step": 5824
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016893201456467697,
+      "loss": 1.058,
+      "step": 5825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016892072766092013,
+      "loss": 1.2102,
+      "step": 5826
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016890943908449402,
+      "loss": 1.0851,
+      "step": 5827
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016889814883567264,
+      "loss": 1.2676,
+      "step": 5828
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016888685691472997,
+      "loss": 1.1269,
+      "step": 5829
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001688755633219401,
+      "loss": 1.2708,
+      "step": 5830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016886426805757707,
+      "loss": 1.1068,
+      "step": 5831
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016885297112191502,
+      "loss": 1.1928,
+      "step": 5832
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001688416725152282,
+      "loss": 1.3129,
+      "step": 5833
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001688303722377907,
+      "loss": 1.1443,
+      "step": 5834
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016881907028987682,
+      "loss": 1.2635,
+      "step": 5835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001688077666717609,
+      "loss": 1.2878,
+      "step": 5836
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001687964613837172,
+      "loss": 1.309,
+      "step": 5837
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016878515442602017,
+      "loss": 1.2827,
+      "step": 5838
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016877384579894414,
+      "loss": 1.2524,
+      "step": 5839
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016876253550276361,
+      "loss": 1.2121,
+      "step": 5840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016875122353775308,
+      "loss": 1.1027,
+      "step": 5841
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001687399099041871,
+      "loss": 1.1963,
+      "step": 5842
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016872859460234017,
+      "loss": 1.0689,
+      "step": 5843
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016871727763248696,
+      "loss": 1.3231,
+      "step": 5844
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016870595899490218,
+      "loss": 1.174,
+      "step": 5845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016869463868986043,
+      "loss": 1.1538,
+      "step": 5846
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016868331671763648,
+      "loss": 1.2075,
+      "step": 5847
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016867199307850513,
+      "loss": 1.262,
+      "step": 5848
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016866066777274124,
+      "loss": 1.2568,
+      "step": 5849
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016864934080061954,
+      "loss": 1.2261,
+      "step": 5850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016863801216241502,
+      "loss": 1.324,
+      "step": 5851
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016862668185840262,
+      "loss": 1.1628,
+      "step": 5852
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016861534988885733,
+      "loss": 1.3086,
+      "step": 5853
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016860401625405408,
+      "loss": 1.1009,
+      "step": 5854
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016859268095426803,
+      "loss": 1.1462,
+      "step": 5855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016858134398977428,
+      "loss": 1.194,
+      "step": 5856
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001685700053608479,
+      "loss": 2.4011,
+      "step": 5857
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016855866506776412,
+      "loss": 1.1213,
+      "step": 5858
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016854732311079817,
+      "loss": 1.3165,
+      "step": 5859
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001685359794902253,
+      "loss": 1.2176,
+      "step": 5860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001685246342063208,
+      "loss": 1.1193,
+      "step": 5861
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016851328725936003,
+      "loss": 1.0659,
+      "step": 5862
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016850193864961836,
+      "loss": 1.2024,
+      "step": 5863
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016849058837737127,
+      "loss": 1.2683,
+      "step": 5864
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016847923644289412,
+      "loss": 1.1838,
+      "step": 5865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016846788284646252,
+      "loss": 1.097,
+      "step": 5866
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016845652758835195,
+      "loss": 0.9967,
+      "step": 5867
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016844517066883806,
+      "loss": 0.9919,
+      "step": 5868
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001684338120881964,
+      "loss": 1.1755,
+      "step": 5869
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016842245184670265,
+      "loss": 1.3577,
+      "step": 5870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016841108994463256,
+      "loss": 0.9577,
+      "step": 5871
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016839972638226186,
+      "loss": 1.2257,
+      "step": 5872
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016838836115986636,
+      "loss": 1.3295,
+      "step": 5873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001683769942777218,
+      "loss": 2.469,
+      "step": 5874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016836562573610417,
+      "loss": 1.3802,
+      "step": 5875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016835425553528928,
+      "loss": 1.3387,
+      "step": 5876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016834288367555312,
+      "loss": 1.2017,
+      "step": 5877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016833151015717169,
+      "loss": 2.4531,
+      "step": 5878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000168320134980421,
+      "loss": 1.1435,
+      "step": 5879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016830875814557712,
+      "loss": 1.2202,
+      "step": 5880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016829737965291615,
+      "loss": 1.173,
+      "step": 5881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016828599950271427,
+      "loss": 1.1704,
+      "step": 5882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016827461769524765,
+      "loss": 1.2211,
+      "step": 5883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016826323423079254,
+      "loss": 1.0812,
+      "step": 5884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016825184910962518,
+      "loss": 1.299,
+      "step": 5885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001682404623320219,
+      "loss": 1.1356,
+      "step": 5886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016822907389825903,
+      "loss": 2.5649,
+      "step": 5887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000168217683808613,
+      "loss": 1.127,
+      "step": 5888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001682062920633602,
+      "loss": 1.2023,
+      "step": 5889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016819489866277715,
+      "loss": 1.3737,
+      "step": 5890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016818350360714034,
+      "loss": 1.2629,
+      "step": 5891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016817210689672627,
+      "loss": 1.2497,
+      "step": 5892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016816070853181158,
+      "loss": 1.4548,
+      "step": 5893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016814930851267293,
+      "loss": 1.3768,
+      "step": 5894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016813790683958694,
+      "loss": 1.1687,
+      "step": 5895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016812650351283036,
+      "loss": 0.9906,
+      "step": 5896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016811509853267994,
+      "loss": 1.2649,
+      "step": 5897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016810369189941242,
+      "loss": 1.201,
+      "step": 5898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001680922836133047,
+      "loss": 1.2656,
+      "step": 5899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001680808736746336,
+      "loss": 1.144,
+      "step": 5900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016806946208367606,
+      "loss": 1.189,
+      "step": 5901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016805804884070905,
+      "loss": 1.2094,
+      "step": 5902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016804663394600952,
+      "loss": 1.2131,
+      "step": 5903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016803521739985455,
+      "loss": 1.1112,
+      "step": 5904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001680237992025212,
+      "loss": 1.2476,
+      "step": 5905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016801237935428658,
+      "loss": 1.2621,
+      "step": 5906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001680009578554278,
+      "loss": 1.1004,
+      "step": 5907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016798953470622212,
+      "loss": 1.3115,
+      "step": 5908
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016797810990694674,
+      "loss": 1.0809,
+      "step": 5909
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016796668345787894,
+      "loss": 1.1003,
+      "step": 5910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000167955255359296,
+      "loss": 1.1082,
+      "step": 5911
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016794382561147536,
+      "loss": 1.2157,
+      "step": 5912
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016793239421469437,
+      "loss": 1.4201,
+      "step": 5913
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016792096116923042,
+      "loss": 1.1386,
+      "step": 5914
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016790952647536102,
+      "loss": 1.4025,
+      "step": 5915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016789809013336367,
+      "loss": 1.0791,
+      "step": 5916
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016788665214351599,
+      "loss": 1.2954,
+      "step": 5917
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016787521250609549,
+      "loss": 2.3738,
+      "step": 5918
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001678637712213798,
+      "loss": 1.2794,
+      "step": 5919
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001678523282896467,
+      "loss": 1.3849,
+      "step": 5920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016784088371117378,
+      "loss": 1.1769,
+      "step": 5921
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016782943748623887,
+      "loss": 1.0811,
+      "step": 5922
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016781798961511973,
+      "loss": 1.2505,
+      "step": 5923
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001678065400980942,
+      "loss": 2.4482,
+      "step": 5924
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016779508893544016,
+      "loss": 1.4265,
+      "step": 5925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016778363612743552,
+      "loss": 1.2434,
+      "step": 5926
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016777218167435826,
+      "loss": 1.2855,
+      "step": 5927
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016776072557648632,
+      "loss": 1.2047,
+      "step": 5928
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016774926783409778,
+      "loss": 1.1662,
+      "step": 5929
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001677378084474707,
+      "loss": 1.21,
+      "step": 5930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016772634741688314,
+      "loss": 1.2702,
+      "step": 5931
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016771488474261333,
+      "loss": 1.36,
+      "step": 5932
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016770342042493944,
+      "loss": 1.0757,
+      "step": 5933
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016769195446413973,
+      "loss": 1.2938,
+      "step": 5934
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016768048686049237,
+      "loss": 1.2441,
+      "step": 5935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001676690176142758,
+      "loss": 1.2133,
+      "step": 5936
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016765754672576827,
+      "loss": 2.4993,
+      "step": 5937
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016764607419524825,
+      "loss": 2.5193,
+      "step": 5938
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016763460002299413,
+      "loss": 1.1576,
+      "step": 5939
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016762312420928437,
+      "loss": 1.1412,
+      "step": 5940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016761164675439753,
+      "loss": 1.3024,
+      "step": 5941
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016760016765861213,
+      "loss": 1.2795,
+      "step": 5942
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016758868692220676,
+      "loss": 1.0296,
+      "step": 5943
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016757720454546008,
+      "loss": 1.307,
+      "step": 5944
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001675657205286507,
+      "loss": 1.2888,
+      "step": 5945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001675542348720574,
+      "loss": 1.1356,
+      "step": 5946
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001675427475759589,
+      "loss": 1.2264,
+      "step": 5947
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016753125864063398,
+      "loss": 1.2036,
+      "step": 5948
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001675197680663615,
+      "loss": 1.0599,
+      "step": 5949
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001675082758534203,
+      "loss": 1.1528,
+      "step": 5950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016749678200208927,
+      "loss": 1.1655,
+      "step": 5951
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016748528651264742,
+      "loss": 1.2513,
+      "step": 5952
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016747378938537373,
+      "loss": 1.241,
+      "step": 5953
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016746229062054716,
+      "loss": 1.2371,
+      "step": 5954
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016745079021844686,
+      "loss": 1.2265,
+      "step": 5955
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016743928817935193,
+      "loss": 1.3192,
+      "step": 5956
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016742778450354147,
+      "loss": 1.2057,
+      "step": 5957
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001674162791912947,
+      "loss": 1.149,
+      "step": 5958
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016740477224289082,
+      "loss": 1.124,
+      "step": 5959
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016739326365860913,
+      "loss": 1.2091,
+      "step": 5960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016738175343872892,
+      "loss": 1.2023,
+      "step": 5961
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016737024158352957,
+      "loss": 1.1124,
+      "step": 5962
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001673587280932904,
+      "loss": 1.39,
+      "step": 5963
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016734721296829093,
+      "loss": 1.1793,
+      "step": 5964
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016733569620881053,
+      "loss": 1.2033,
+      "step": 5965
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016732417781512875,
+      "loss": 1.1343,
+      "step": 5966
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016731265778752513,
+      "loss": 1.0048,
+      "step": 5967
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001673011361262793,
+      "loss": 1.1744,
+      "step": 5968
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016728961283167076,
+      "loss": 1.1191,
+      "step": 5969
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016727808790397932,
+      "loss": 1.1327,
+      "step": 5970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016726656134348462,
+      "loss": 1.0766,
+      "step": 5971
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001672550331504664,
+      "loss": 1.2591,
+      "step": 5972
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016724350332520443,
+      "loss": 1.1805,
+      "step": 5973
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016723197186797854,
+      "loss": 1.4002,
+      "step": 5974
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016722043877906862,
+      "loss": 1.2379,
+      "step": 5975
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016720890405875455,
+      "loss": 2.6682,
+      "step": 5976
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001671973677073163,
+      "loss": 1.3516,
+      "step": 5977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001671858297250338,
+      "loss": 1.1499,
+      "step": 5978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001671742901121871,
+      "loss": 1.2058,
+      "step": 5979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016716274886905623,
+      "loss": 1.1987,
+      "step": 5980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016715120599592134,
+      "loss": 1.3429,
+      "step": 5981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016713966149306257,
+      "loss": 1.0945,
+      "step": 5982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016712811536076006,
+      "loss": 1.17,
+      "step": 5983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000167116567599294,
+      "loss": 1.1676,
+      "step": 5984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016710501820894473,
+      "loss": 1.0407,
+      "step": 5985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016709346718999255,
+      "loss": 1.0923,
+      "step": 5986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001670819145427177,
+      "loss": 1.1192,
+      "step": 5987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001670703602674006,
+      "loss": 1.2125,
+      "step": 5988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016705880436432172,
+      "loss": 1.2706,
+      "step": 5989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016704724683376147,
+      "loss": 1.3583,
+      "step": 5990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016703568767600032,
+      "loss": 2.4167,
+      "step": 5991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016702412689131884,
+      "loss": 1.1985,
+      "step": 5992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016701256447999765,
+      "loss": 1.3006,
+      "step": 5993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016700100044231728,
+      "loss": 1.2359,
+      "step": 5994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001669894347785584,
+      "loss": 1.088,
+      "step": 5995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016697786748900173,
+      "loss": 0.9608,
+      "step": 5996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016696629857392798,
+      "loss": 1.3351,
+      "step": 5997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016695472803361793,
+      "loss": 1.0688,
+      "step": 5998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001669431558683524,
+      "loss": 1.174,
+      "step": 5999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016693158207841224,
+      "loss": 1.2292,
+      "step": 6000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016692000666407831,
+      "loss": 1.2051,
+      "step": 6001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016690842962563157,
+      "loss": 1.1815,
+      "step": 6002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000166896850963353,
+      "loss": 1.2223,
+      "step": 6003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016688527067752355,
+      "loss": 1.3723,
+      "step": 6004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016687368876842432,
+      "loss": 2.5198,
+      "step": 6005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001668621052363364,
+      "loss": 1.428,
+      "step": 6006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016685052008154087,
+      "loss": 1.201,
+      "step": 6007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016683893330431894,
+      "loss": 1.3866,
+      "step": 6008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016682734490495177,
+      "loss": 1.0079,
+      "step": 6009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016681575488372067,
+      "loss": 1.1633,
+      "step": 6010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016680416324090687,
+      "loss": 0.9297,
+      "step": 6011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001667925699767917,
+      "loss": 1.0395,
+      "step": 6012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016678097509165656,
+      "loss": 1.0763,
+      "step": 6013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016676937858578277,
+      "loss": 1.1415,
+      "step": 6014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016675778045945188,
+      "loss": 1.0898,
+      "step": 6015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016674618071294526,
+      "loss": 1.0377,
+      "step": 6016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016673457934654454,
+      "loss": 1.2067,
+      "step": 6017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016672297636053117,
+      "loss": 1.181,
+      "step": 6018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016671137175518683,
+      "loss": 1.064,
+      "step": 6019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016669976553079313,
+      "loss": 1.1834,
+      "step": 6020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016668815768763174,
+      "loss": 1.061,
+      "step": 6021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001666765482259844,
+      "loss": 1.2424,
+      "step": 6022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001666649371461328,
+      "loss": 1.1519,
+      "step": 6023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001666533244483588,
+      "loss": 2.541,
+      "step": 6024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016664171013294423,
+      "loss": 1.1463,
+      "step": 6025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016663009420017096,
+      "loss": 1.3027,
+      "step": 6026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016661847665032085,
+      "loss": 1.1208,
+      "step": 6027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016660685748367594,
+      "loss": 1.1555,
+      "step": 6028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016659523670051814,
+      "loss": 1.1521,
+      "step": 6029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016658361430112954,
+      "loss": 1.23,
+      "step": 6030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016657199028579216,
+      "loss": 1.2801,
+      "step": 6031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016656036465478814,
+      "loss": 1.2469,
+      "step": 6032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001665487374083996,
+      "loss": 1.248,
+      "step": 6033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001665371085469088,
+      "loss": 1.2726,
+      "step": 6034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016652547807059786,
+      "loss": 1.2368,
+      "step": 6035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001665138459797491,
+      "loss": 1.1749,
+      "step": 6036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016650221227464487,
+      "loss": 2.2957,
+      "step": 6037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016649057695556746,
+      "loss": 1.2661,
+      "step": 6038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001664789400227992,
+      "loss": 1.2678,
+      "step": 6039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016646730147662261,
+      "loss": 1.2868,
+      "step": 6040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016645566131732014,
+      "loss": 2.446,
+      "step": 6041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016644401954517423,
+      "loss": 1.1934,
+      "step": 6042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016643237616046746,
+      "loss": 1.1863,
+      "step": 6043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016642073116348243,
+      "loss": 1.181,
+      "step": 6044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001664090845545017,
+      "loss": 1.1539,
+      "step": 6045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016639743633380797,
+      "loss": 1.1179,
+      "step": 6046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016638578650168392,
+      "loss": 1.2755,
+      "step": 6047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016637413505841227,
+      "loss": 1.0035,
+      "step": 6048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016636248200427582,
+      "loss": 1.1872,
+      "step": 6049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001663508273395574,
+      "loss": 1.2912,
+      "step": 6050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001663391710645398,
+      "loss": 1.1748,
+      "step": 6051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000166327513179506,
+      "loss": 1.1817,
+      "step": 6052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016631585368473886,
+      "loss": 1.3214,
+      "step": 6053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016630419258052134,
+      "loss": 1.2841,
+      "step": 6054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001662925298671365,
+      "loss": 2.2725,
+      "step": 6055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016628086554486737,
+      "loss": 1.1759,
+      "step": 6056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016626919961399705,
+      "loss": 1.1687,
+      "step": 6057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016625753207480864,
+      "loss": 1.0793,
+      "step": 6058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016624586292758532,
+      "loss": 1.1463,
+      "step": 6059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016623419217261026,
+      "loss": 1.3304,
+      "step": 6060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016622251981016678,
+      "loss": 1.1714,
+      "step": 6061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016621084584053807,
+      "loss": 1.0367,
+      "step": 6062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001661991702640075,
+      "loss": 1.2521,
+      "step": 6063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016618749308085846,
+      "loss": 1.3194,
+      "step": 6064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016617581429137428,
+      "loss": 1.3329,
+      "step": 6065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016616413389583843,
+      "loss": 1.1826,
+      "step": 6066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001661524518945344,
+      "loss": 1.3,
+      "step": 6067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001661407682877457,
+      "loss": 1.2338,
+      "step": 6068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001661290830757559,
+      "loss": 1.1785,
+      "step": 6069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001661173962588485,
+      "loss": 1.2972,
+      "step": 6070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001661057078373073,
+      "loss": 1.3585,
+      "step": 6071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016609401781141582,
+      "loss": 1.055,
+      "step": 6072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016608232618145786,
+      "loss": 1.2266,
+      "step": 6073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016607063294771712,
+      "loss": 1.2609,
+      "step": 6074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001660589381104774,
+      "loss": 1.1535,
+      "step": 6075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016604724167002254,
+      "loss": 1.1009,
+      "step": 6076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016603554362663641,
+      "loss": 1.2537,
+      "step": 6077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016602384398060289,
+      "loss": 1.1829,
+      "step": 6078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016601214273220593,
+      "loss": 1.2302,
+      "step": 6079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016600043988172956,
+      "loss": 1.1717,
+      "step": 6080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016598873542945775,
+      "loss": 1.167,
+      "step": 6081
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016597702937567457,
+      "loss": 1.219,
+      "step": 6082
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016596532172066412,
+      "loss": 1.0849,
+      "step": 6083
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016595361246471058,
+      "loss": 1.1095,
+      "step": 6084
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016594190160809804,
+      "loss": 1.1802,
+      "step": 6085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001659301891511108,
+      "loss": 1.17,
+      "step": 6086
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016591847509403308,
+      "loss": 1.0867,
+      "step": 6087
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016590675943714916,
+      "loss": 1.0377,
+      "step": 6088
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016589504218074343,
+      "loss": 0.95,
+      "step": 6089
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001658833233251002,
+      "loss": 1.1143,
+      "step": 6090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001658716028705039,
+      "loss": 1.4088,
+      "step": 6091
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016585988081723894,
+      "loss": 1.2664,
+      "step": 6092
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001658481571655899,
+      "loss": 1.1883,
+      "step": 6093
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016583643191584124,
+      "loss": 1.1996,
+      "step": 6094
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016582470506827754,
+      "loss": 1.2154,
+      "step": 6095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016581297662318344,
+      "loss": 1.3107,
+      "step": 6096
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016580124658084348,
+      "loss": 1.0927,
+      "step": 6097
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016578951494154246,
+      "loss": 1.2911,
+      "step": 6098
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000165777781705565,
+      "loss": 1.2695,
+      "step": 6099
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016576604687319597,
+      "loss": 1.2398,
+      "step": 6100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016575431044472006,
+      "loss": 1.4178,
+      "step": 6101
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001657425724204222,
+      "loss": 1.1525,
+      "step": 6102
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001657308328005872,
+      "loss": 1.3783,
+      "step": 6103
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001657190915855,
+      "loss": 1.2,
+      "step": 6104
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016570734877544555,
+      "loss": 1.3922,
+      "step": 6105
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016569560437070883,
+      "loss": 1.2473,
+      "step": 6106
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016568385837157487,
+      "loss": 1.2325,
+      "step": 6107
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001656721107783288,
+      "loss": 1.0728,
+      "step": 6108
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016566036159125562,
+      "loss": 1.2545,
+      "step": 6109
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016564861081064055,
+      "loss": 2.5117,
+      "step": 6110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016563685843676878,
+      "loss": 1.4005,
+      "step": 6111
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016562510446992547,
+      "loss": 1.1627,
+      "step": 6112
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016561334891039597,
+      "loss": 1.1668,
+      "step": 6113
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001656015917584655,
+      "loss": 1.2363,
+      "step": 6114
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016558983301441946,
+      "loss": 1.2289,
+      "step": 6115
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001655780726785432,
+      "loss": 2.2529,
+      "step": 6116
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016556631075112212,
+      "loss": 1.0516,
+      "step": 6117
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016555454723244176,
+      "loss": 1.1278,
+      "step": 6118
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016554278212278746,
+      "loss": 1.0022,
+      "step": 6119
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001655310154224449,
+      "loss": 1.5059,
+      "step": 6120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016551924713169958,
+      "loss": 1.1072,
+      "step": 6121
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016550747725083714,
+      "loss": 1.1366,
+      "step": 6122
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016549570578014322,
+      "loss": 1.2758,
+      "step": 6123
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001654839327199035,
+      "loss": 1.1198,
+      "step": 6124
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001654721580704037,
+      "loss": 1.2713,
+      "step": 6125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016546038183192958,
+      "loss": 1.136,
+      "step": 6126
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016544860400476697,
+      "loss": 1.2507,
+      "step": 6127
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001654368245892017,
+      "loss": 1.0219,
+      "step": 6128
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016542504358551966,
+      "loss": 1.1442,
+      "step": 6129
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001654132609940067,
+      "loss": 1.2737,
+      "step": 6130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016540147681494888,
+      "loss": 1.1269,
+      "step": 6131
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016538969104863216,
+      "loss": 1.1776,
+      "step": 6132
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016537790369534254,
+      "loss": 1.1095,
+      "step": 6133
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016536611475536611,
+      "loss": 1.2493,
+      "step": 6134
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000165354324228989,
+      "loss": 1.2603,
+      "step": 6135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016534253211649734,
+      "loss": 1.0299,
+      "step": 6136
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016533073841817734,
+      "loss": 1.2259,
+      "step": 6137
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001653189431343152,
+      "loss": 1.1562,
+      "step": 6138
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001653071462651972,
+      "loss": 1.281,
+      "step": 6139
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016529534781110967,
+      "loss": 1.1409,
+      "step": 6140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016528354777233886,
+      "loss": 1.1031,
+      "step": 6141
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016527174614917124,
+      "loss": 1.3169,
+      "step": 6142
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016525994294189325,
+      "loss": 1.2239,
+      "step": 6143
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016524813815079124,
+      "loss": 1.1185,
+      "step": 6144
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016523633177615177,
+      "loss": 0.9985,
+      "step": 6145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001652245238182614,
+      "loss": 1.4562,
+      "step": 6146
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016521271427740665,
+      "loss": 1.2819,
+      "step": 6147
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001652009031538742,
+      "loss": 1.2714,
+      "step": 6148
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001651890904479506,
+      "loss": 2.5803,
+      "step": 6149
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016517727615992261,
+      "loss": 1.4033,
+      "step": 6150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016516546029007696,
+      "loss": 1.0815,
+      "step": 6151
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016515364283870037,
+      "loss": 1.147,
+      "step": 6152
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016514182380607966,
+      "loss": 1.2798,
+      "step": 6153
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001651300031925017,
+      "loss": 2.4314,
+      "step": 6154
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016511818099825335,
+      "loss": 1.2818,
+      "step": 6155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016510635722362152,
+      "loss": 1.1802,
+      "step": 6156
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016509453186889315,
+      "loss": 2.4849,
+      "step": 6157
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001650827049343553,
+      "loss": 1.2485,
+      "step": 6158
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016507087642029492,
+      "loss": 1.1775,
+      "step": 6159
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016505904632699916,
+      "loss": 1.23,
+      "step": 6160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016504721465475508,
+      "loss": 2.2256,
+      "step": 6161
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016503538140384984,
+      "loss": 1.3336,
+      "step": 6162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016502354657457063,
+      "loss": 1.1451,
+      "step": 6163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001650117101672047,
+      "loss": 1.4786,
+      "step": 6164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016499987218203925,
+      "loss": 1.1835,
+      "step": 6165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001649880326193616,
+      "loss": 1.2725,
+      "step": 6166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016497619147945916,
+      "loss": 1.303,
+      "step": 6167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016496434876261926,
+      "loss": 1.3679,
+      "step": 6168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001649525044691293,
+      "loss": 1.2204,
+      "step": 6169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001649406585992767,
+      "loss": 1.1257,
+      "step": 6170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016492881115334906,
+      "loss": 1.3159,
+      "step": 6171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016491696213163384,
+      "loss": 1.0864,
+      "step": 6172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016490511153441858,
+      "loss": 1.2191,
+      "step": 6173
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016489325936199097,
+      "loss": 1.1926,
+      "step": 6174
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016488140561463863,
+      "loss": 2.5259,
+      "step": 6175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001648695502926492,
+      "loss": 1.3631,
+      "step": 6176
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016485769339631045,
+      "loss": 1.1813,
+      "step": 6177
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001648458349259101,
+      "loss": 1.1505,
+      "step": 6178
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016483397488173596,
+      "loss": 1.053,
+      "step": 6179
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016482211326407592,
+      "loss": 1.2023,
+      "step": 6180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001648102500732178,
+      "loss": 1.1834,
+      "step": 6181
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001647983853094495,
+      "loss": 1.3276,
+      "step": 6182
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016478651897305904,
+      "loss": 1.0521,
+      "step": 6183
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016477465106433433,
+      "loss": 1.3747,
+      "step": 6184
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00016476278158356346,
+      "loss": 1.1448,
+      "step": 6185
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016475091053103443,
+      "loss": 1.0535,
+      "step": 6186
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016473903790703543,
+      "loss": 1.1928,
+      "step": 6187
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016472716371185455,
+      "loss": 1.1546,
+      "step": 6188
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016471528794577997,
+      "loss": 1.0919,
+      "step": 6189
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001647034106090999,
+      "loss": 1.2103,
+      "step": 6190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016469153170210266,
+      "loss": 1.1782,
+      "step": 6191
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016467965122507648,
+      "loss": 1.2754,
+      "step": 6192
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016466776917830966,
+      "loss": 1.2109,
+      "step": 6193
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001646558855620907,
+      "loss": 1.1703,
+      "step": 6194
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016464400037670787,
+      "loss": 1.1113,
+      "step": 6195
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016463211362244967,
+      "loss": 1.3749,
+      "step": 6196
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001646202252996046,
+      "loss": 1.1838,
+      "step": 6197
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016460833540846124,
+      "loss": 1.3876,
+      "step": 6198
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016459644394930804,
+      "loss": 1.1165,
+      "step": 6199
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016458455092243366,
+      "loss": 1.2553,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016457265632812668,
+      "loss": 1.1509,
+      "step": 6201
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016456076016667587,
+      "loss": 1.2513,
+      "step": 6202
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016454886243836986,
+      "loss": 1.1207,
+      "step": 6203
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016453696314349748,
+      "loss": 1.1549,
+      "step": 6204
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001645250622823474,
+      "loss": 2.7212,
+      "step": 6205
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016451315985520858,
+      "loss": 1.1757,
+      "step": 6206
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016450125586236981,
+      "loss": 1.0357,
+      "step": 6207
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016448935030412002,
+      "loss": 2.3752,
+      "step": 6208
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016447744318074813,
+      "loss": 1.1878,
+      "step": 6209
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016446553449254317,
+      "loss": 0.99,
+      "step": 6210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016445362423979406,
+      "loss": 1.27,
+      "step": 6211
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016444171242278993,
+      "loss": 1.2369,
+      "step": 6212
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001644297990418199,
+      "loss": 1.3808,
+      "step": 6213
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016441788409717303,
+      "loss": 0.9972,
+      "step": 6214
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016440596758913851,
+      "loss": 1.0505,
+      "step": 6215
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016439404951800555,
+      "loss": 1.1299,
+      "step": 6216
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001643821298840634,
+      "loss": 1.1821,
+      "step": 6217
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016437020868760138,
+      "loss": 1.1447,
+      "step": 6218
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016435828592890875,
+      "loss": 1.2588,
+      "step": 6219
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001643463616082749,
+      "loss": 1.1277,
+      "step": 6220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016433443572598923,
+      "loss": 1.1344,
+      "step": 6221
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016432250828234116,
+      "loss": 1.0903,
+      "step": 6222
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016431057927762015,
+      "loss": 1.3359,
+      "step": 6223
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016429864871211577,
+      "loss": 1.1829,
+      "step": 6224
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001642867165861175,
+      "loss": 1.1916,
+      "step": 6225
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016427478289991493,
+      "loss": 1.1016,
+      "step": 6226
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016426284765379775,
+      "loss": 1.2318,
+      "step": 6227
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016425091084805557,
+      "loss": 1.0797,
+      "step": 6228
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001642389724829781,
+      "loss": 1.3705,
+      "step": 6229
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016422703255885508,
+      "loss": 1.1432,
+      "step": 6230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016421509107597627,
+      "loss": 1.2061,
+      "step": 6231
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001642031480346315,
+      "loss": 1.2469,
+      "step": 6232
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016419120343511064,
+      "loss": 1.2598,
+      "step": 6233
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016417925727770354,
+      "loss": 1.1014,
+      "step": 6234
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016416730956270013,
+      "loss": 1.1849,
+      "step": 6235
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016415536029039043,
+      "loss": 1.1497,
+      "step": 6236
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016414340946106437,
+      "loss": 1.0856,
+      "step": 6237
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000164131457075012,
+      "loss": 0.976,
+      "step": 6238
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016411950313252347,
+      "loss": 1.1791,
+      "step": 6239
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016410754763388882,
+      "loss": 1.2957,
+      "step": 6240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016409559057939822,
+      "loss": 1.0276,
+      "step": 6241
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016408363196934188,
+      "loss": 1.21,
+      "step": 6242
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016407167180401005,
+      "loss": 1.2872,
+      "step": 6243
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016405971008369292,
+      "loss": 1.3535,
+      "step": 6244
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016404774680868088,
+      "loss": 1.0642,
+      "step": 6245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001640357819792642,
+      "loss": 1.2997,
+      "step": 6246
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001640238155957333,
+      "loss": 1.265,
+      "step": 6247
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016401184765837861,
+      "loss": 1.2775,
+      "step": 6248
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016399987816749053,
+      "loss": 1.2421,
+      "step": 6249
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016398790712335962,
+      "loss": 1.1111,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001639759345262764,
+      "loss": 2.3486,
+      "step": 6251
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016396396037653135,
+      "loss": 1.1718,
+      "step": 6252
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001639519846744152,
+      "loss": 1.2363,
+      "step": 6253
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016394000742021853,
+      "loss": 1.1626,
+      "step": 6254
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016392802861423203,
+      "loss": 1.0479,
+      "step": 6255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016391604825674643,
+      "loss": 2.533,
+      "step": 6256
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016390406634805244,
+      "loss": 1.1258,
+      "step": 6257
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016389208288844094,
+      "loss": 1.2488,
+      "step": 6258
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016388009787820268,
+      "loss": 1.0193,
+      "step": 6259
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016386811131762857,
+      "loss": 1.3691,
+      "step": 6260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001638561232070095,
+      "loss": 1.3625,
+      "step": 6261
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016384413354663644,
+      "loss": 1.2576,
+      "step": 6262
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016383214233680035,
+      "loss": 1.2786,
+      "step": 6263
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016382014957779228,
+      "loss": 1.2994,
+      "step": 6264
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016380815526990326,
+      "loss": 1.1369,
+      "step": 6265
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016379615941342437,
+      "loss": 1.3528,
+      "step": 6266
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001637841620086468,
+      "loss": 1.2128,
+      "step": 6267
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016377216305586166,
+      "loss": 1.1583,
+      "step": 6268
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016376016255536014,
+      "loss": 1.2632,
+      "step": 6269
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001637481605074336,
+      "loss": 1.3673,
+      "step": 6270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001637361569123732,
+      "loss": 1.1122,
+      "step": 6271
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016372415177047037,
+      "loss": 1.0742,
+      "step": 6272
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016371214508201638,
+      "loss": 1.2686,
+      "step": 6273
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016370013684730266,
+      "loss": 1.1399,
+      "step": 6274
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016368812706662065,
+      "loss": 2.5295,
+      "step": 6275
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001636761157402618,
+      "loss": 1.2709,
+      "step": 6276
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001636641028685176,
+      "loss": 1.2047,
+      "step": 6277
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016365208845167968,
+      "loss": 1.1262,
+      "step": 6278
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016364007249003958,
+      "loss": 1.0952,
+      "step": 6279
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016362805498388886,
+      "loss": 2.2178,
+      "step": 6280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016361603593351926,
+      "loss": 1.2491,
+      "step": 6281
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016360401533922243,
+      "loss": 1.1826,
+      "step": 6282
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016359199320129016,
+      "loss": 1.3373,
+      "step": 6283
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016357996952001414,
+      "loss": 1.2087,
+      "step": 6284
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001635679442956862,
+      "loss": 1.2775,
+      "step": 6285
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016355591752859825,
+      "loss": 1.3028,
+      "step": 6286
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016354388921904208,
+      "loss": 1.11,
+      "step": 6287
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001635318593673097,
+      "loss": 1.1042,
+      "step": 6288
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00016351982797369302,
+      "loss": 1.2991,
+      "step": 6289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000163507795038484,
+      "loss": 1.2227,
+      "step": 6290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016349576056197477,
+      "loss": 1.1654,
+      "step": 6291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016348372454445733,
+      "loss": 1.3587,
+      "step": 6292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016347168698622376,
+      "loss": 1.2872,
+      "step": 6293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001634596478875663,
+      "loss": 1.0714,
+      "step": 6294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016344760724877706,
+      "loss": 1.1772,
+      "step": 6295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016343556507014828,
+      "loss": 1.1904,
+      "step": 6296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001634235213519722,
+      "loss": 1.5604,
+      "step": 6297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016341147609454115,
+      "loss": 1.1384,
+      "step": 6298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016339942929814743,
+      "loss": 1.2708,
+      "step": 6299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016338738096308343,
+      "loss": 1.3218,
+      "step": 6300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016337533108964157,
+      "loss": 1.3009,
+      "step": 6301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016336327967811427,
+      "loss": 1.3777,
+      "step": 6302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000163351226728794,
+      "loss": 1.1839,
+      "step": 6303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001633391722419733,
+      "loss": 2.3716,
+      "step": 6304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001633271162179447,
+      "loss": 1.3676,
+      "step": 6305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016331505865700084,
+      "loss": 1.2752,
+      "step": 6306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016330299955943432,
+      "loss": 1.1567,
+      "step": 6307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016329093892553779,
+      "loss": 1.2773,
+      "step": 6308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016327887675560398,
+      "loss": 1.2213,
+      "step": 6309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016326681304992566,
+      "loss": 1.0306,
+      "step": 6310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016325474780879554,
+      "loss": 1.3707,
+      "step": 6311
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016324268103250648,
+      "loss": 1.2771,
+      "step": 6312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016323061272135135,
+      "loss": 1.0192,
+      "step": 6313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000163218542875623,
+      "loss": 1.1417,
+      "step": 6314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016320647149561438,
+      "loss": 1.1675,
+      "step": 6315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001631943985816185,
+      "loss": 1.1859,
+      "step": 6316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016318232413392827,
+      "loss": 1.1522,
+      "step": 6317
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016317024815283682,
+      "loss": 1.1182,
+      "step": 6318
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016315817063863714,
+      "loss": 1.1502,
+      "step": 6319
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001631460915916224,
+      "loss": 2.3015,
+      "step": 6320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016313401101208577,
+      "loss": 1.1417,
+      "step": 6321
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001631219289003204,
+      "loss": 1.2186,
+      "step": 6322
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001631098452566195,
+      "loss": 0.9792,
+      "step": 6323
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016309776008127642,
+      "loss": 1.2228,
+      "step": 6324
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016308567337458438,
+      "loss": 1.0925,
+      "step": 6325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001630735851368367,
+      "loss": 1.1989,
+      "step": 6326
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016306149536832683,
+      "loss": 1.1062,
+      "step": 6327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016304940406934818,
+      "loss": 1.1517,
+      "step": 6328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016303731124019412,
+      "loss": 1.0654,
+      "step": 6329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001630252168811582,
+      "loss": 1.2585,
+      "step": 6330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016301312099253396,
+      "loss": 1.0341,
+      "step": 6331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001630010235746149,
+      "loss": 1.016,
+      "step": 6332
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016298892462769465,
+      "loss": 1.2346,
+      "step": 6333
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016297682415206683,
+      "loss": 1.2734,
+      "step": 6334
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016296472214802514,
+      "loss": 1.1427,
+      "step": 6335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001629526186158633,
+      "loss": 1.1133,
+      "step": 6336
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016294051355587497,
+      "loss": 1.2385,
+      "step": 6337
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016292840696835402,
+      "loss": 1.1946,
+      "step": 6338
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016291629885359425,
+      "loss": 1.1791,
+      "step": 6339
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016290418921188952,
+      "loss": 1.0856,
+      "step": 6340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016289207804353367,
+      "loss": 1.1839,
+      "step": 6341
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001628799653488207,
+      "loss": 1.319,
+      "step": 6342
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001628678511280446,
+      "loss": 1.2057,
+      "step": 6343
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016285573538149926,
+      "loss": 1.2507,
+      "step": 6344
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016284361810947883,
+      "loss": 1.1593,
+      "step": 6345
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016283149931227737,
+      "loss": 2.3408,
+      "step": 6346
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016281937899018898,
+      "loss": 1.1638,
+      "step": 6347
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016280725714350777,
+      "loss": 1.4573,
+      "step": 6348
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016279513377252802,
+      "loss": 1.2686,
+      "step": 6349
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016278300887754389,
+      "loss": 1.0718,
+      "step": 6350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016277088245884963,
+      "loss": 1.0634,
+      "step": 6351
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016275875451673963,
+      "loss": 2.6201,
+      "step": 6352
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016274662505150818,
+      "loss": 0.9708,
+      "step": 6353
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016273449406344965,
+      "loss": 1.2866,
+      "step": 6354
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016272236155285844,
+      "loss": 1.2047,
+      "step": 6355
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016271022752002901,
+      "loss": 1.4232,
+      "step": 6356
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016269809196525588,
+      "loss": 1.2368,
+      "step": 6357
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001626859548888335,
+      "loss": 1.154,
+      "step": 6358
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001626738162910565,
+      "loss": 1.0379,
+      "step": 6359
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001626616761722195,
+      "loss": 1.2359,
+      "step": 6360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016264953453261702,
+      "loss": 1.184,
+      "step": 6361
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001626373913725438,
+      "loss": 1.1373,
+      "step": 6362
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016262524669229459,
+      "loss": 2.3408,
+      "step": 6363
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016261310049216408,
+      "loss": 1.1589,
+      "step": 6364
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016260095277244703,
+      "loss": 1.1391,
+      "step": 6365
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016258880353343832,
+      "loss": 1.2333,
+      "step": 6366
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001625766527754328,
+      "loss": 1.3212,
+      "step": 6367
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016256450049872528,
+      "loss": 1.1318,
+      "step": 6368
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016255234670361078,
+      "loss": 1.3316,
+      "step": 6369
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016254019139038425,
+      "loss": 1.0646,
+      "step": 6370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016252803455934066,
+      "loss": 1.1353,
+      "step": 6371
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016251587621077506,
+      "loss": 1.1088,
+      "step": 6372
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016250371634498256,
+      "loss": 1.2003,
+      "step": 6373
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016249155496225824,
+      "loss": 1.0529,
+      "step": 6374
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016247939206289727,
+      "loss": 1.1793,
+      "step": 6375
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016246722764719478,
+      "loss": 1.1309,
+      "step": 6376
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016245506171544606,
+      "loss": 1.2738,
+      "step": 6377
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016244289426794634,
+      "loss": 1.0848,
+      "step": 6378
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016243072530499095,
+      "loss": 1.2612,
+      "step": 6379
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001624185548268752,
+      "loss": 1.2967,
+      "step": 6380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016240638283389447,
+      "loss": 1.1542,
+      "step": 6381
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016239420932634413,
+      "loss": 1.203,
+      "step": 6382
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001623820343045197,
+      "loss": 0.8912,
+      "step": 6383
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016236985776871656,
+      "loss": 2.5278,
+      "step": 6384
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016235767971923033,
+      "loss": 1.317,
+      "step": 6385
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016234550015635648,
+      "loss": 1.3162,
+      "step": 6386
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016233331908039066,
+      "loss": 1.2062,
+      "step": 6387
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016232113649162849,
+      "loss": 1.2307,
+      "step": 6388
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016230895239036558,
+      "loss": 1.3099,
+      "step": 6389
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016229676677689773,
+      "loss": 0.9694,
+      "step": 6390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001622845796515206,
+      "loss": 1.353,
+      "step": 6391
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016227239101452998,
+      "loss": 1.2913,
+      "step": 6392
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00016226020086622172,
+      "loss": 1.1922,
+      "step": 6393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001622480092068916,
+      "loss": 1.386,
+      "step": 6394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016223581603683557,
+      "loss": 1.1631,
+      "step": 6395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001622236213563495,
+      "loss": 1.1777,
+      "step": 6396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016221142516572941,
+      "loss": 0.9554,
+      "step": 6397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016219922746527122,
+      "loss": 1.2357,
+      "step": 6398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000162187028255271,
+      "loss": 1.2749,
+      "step": 6399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016217482753602488,
+      "loss": 1.0735,
+      "step": 6400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016216262530782883,
+      "loss": 1.154,
+      "step": 6401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016215042157097912,
+      "loss": 1.1588,
+      "step": 6402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016213821632577182,
+      "loss": 1.2715,
+      "step": 6403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016212600957250323,
+      "loss": 1.1162,
+      "step": 6404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016211380131146957,
+      "loss": 1.0898,
+      "step": 6405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016210159154296714,
+      "loss": 1.1144,
+      "step": 6406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001620893802672922,
+      "loss": 1.1655,
+      "step": 6407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016207716748474122,
+      "loss": 1.2462,
+      "step": 6408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016206495319561054,
+      "loss": 2.5652,
+      "step": 6409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016205273740019655,
+      "loss": 1.1719,
+      "step": 6410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001620405200987958,
+      "loss": 1.217,
+      "step": 6411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016202830129170476,
+      "loss": 1.0804,
+      "step": 6412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016201608097921998,
+      "loss": 1.2012,
+      "step": 6413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000162003859161638,
+      "loss": 1.0742,
+      "step": 6414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016199163583925552,
+      "loss": 1.3803,
+      "step": 6415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016197941101236913,
+      "loss": 1.3693,
+      "step": 6416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016196718468127554,
+      "loss": 1.2031,
+      "step": 6417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016195495684627147,
+      "loss": 1.2093,
+      "step": 6418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016194272750765372,
+      "loss": 1.0832,
+      "step": 6419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000161930496665719,
+      "loss": 1.1436,
+      "step": 6420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016191826432076424,
+      "loss": 1.0952,
+      "step": 6421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016190603047308626,
+      "loss": 2.4321,
+      "step": 6422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000161893795122982,
+      "loss": 1.2814,
+      "step": 6423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016188155827074837,
+      "loss": 1.1588,
+      "step": 6424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001618693199166824,
+      "loss": 1.094,
+      "step": 6425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016185708006108106,
+      "loss": 1.1835,
+      "step": 6426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016184483870424146,
+      "loss": 2.3152,
+      "step": 6427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001618325958464606,
+      "loss": 1.3028,
+      "step": 6428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001618203514880357,
+      "loss": 1.0403,
+      "step": 6429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016180810562926387,
+      "loss": 1.1732,
+      "step": 6430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016179585827044232,
+      "loss": 1.2397,
+      "step": 6431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016178360941186833,
+      "loss": 1.0992,
+      "step": 6432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016177135905383913,
+      "loss": 1.2522,
+      "step": 6433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016175910719665203,
+      "loss": 1.3703,
+      "step": 6434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001617468538406044,
+      "loss": 1.087,
+      "step": 6435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016173459898599356,
+      "loss": 1.1048,
+      "step": 6436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000161722342633117,
+      "loss": 1.1651,
+      "step": 6437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016171008478227215,
+      "loss": 1.2717,
+      "step": 6438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016169782543375654,
+      "loss": 1.3046,
+      "step": 6439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016168556458786762,
+      "loss": 1.1893,
+      "step": 6440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016167330224490297,
+      "loss": 1.4294,
+      "step": 6441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016166103840516027,
+      "loss": 1.1995,
+      "step": 6442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016164877306893707,
+      "loss": 1.2036,
+      "step": 6443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016163650623653107,
+      "loss": 1.1924,
+      "step": 6444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016162423790824,
+      "loss": 1.2583,
+      "step": 6445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001616119680843616,
+      "loss": 1.2347,
+      "step": 6446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016159969676519362,
+      "loss": 1.0985,
+      "step": 6447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001615874239510339,
+      "loss": 1.2617,
+      "step": 6448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016157514964218033,
+      "loss": 1.104,
+      "step": 6449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001615628738389307,
+      "loss": 1.0543,
+      "step": 6450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016155059654158305,
+      "loss": 1.185,
+      "step": 6451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001615383177504353,
+      "loss": 1.1881,
+      "step": 6452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016152603746578544,
+      "loss": 1.3083,
+      "step": 6453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001615137556879315,
+      "loss": 0.9763,
+      "step": 6454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016150147241717156,
+      "loss": 1.1016,
+      "step": 6455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016148918765380378,
+      "loss": 1.2695,
+      "step": 6456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001614769013981262,
+      "loss": 1.1031,
+      "step": 6457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016146461365043705,
+      "loss": 1.2207,
+      "step": 6458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001614523244110346,
+      "loss": 1.2714,
+      "step": 6459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016144003368021706,
+      "loss": 1.1107,
+      "step": 6460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016142774145828267,
+      "loss": 1.2844,
+      "step": 6461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016141544774552985,
+      "loss": 1.2509,
+      "step": 6462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016140315254225692,
+      "loss": 1.0428,
+      "step": 6463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016139085584876225,
+      "loss": 1.2457,
+      "step": 6464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001613785576653443,
+      "loss": 1.2637,
+      "step": 6465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001613662579923015,
+      "loss": 1.1799,
+      "step": 6466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016135395682993244,
+      "loss": 1.1445,
+      "step": 6467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016134165417853563,
+      "loss": 1.2057,
+      "step": 6468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016132935003840962,
+      "loss": 1.0459,
+      "step": 6469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016131704440985303,
+      "loss": 1.0889,
+      "step": 6470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016130473729316454,
+      "loss": 1.1939,
+      "step": 6471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001612924286886428,
+      "loss": 1.2448,
+      "step": 6472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016128011859658657,
+      "loss": 1.1069,
+      "step": 6473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016126780701729456,
+      "loss": 1.2457,
+      "step": 6474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016125549395106564,
+      "loss": 1.2516,
+      "step": 6475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016124317939819854,
+      "loss": 1.0969,
+      "step": 6476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016123086335899223,
+      "loss": 1.0904,
+      "step": 6477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016121854583374558,
+      "loss": 1.1954,
+      "step": 6478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001612062268227575,
+      "loss": 1.1294,
+      "step": 6479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000161193906326327,
+      "loss": 0.9001,
+      "step": 6480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016118158434475305,
+      "loss": 2.4897,
+      "step": 6481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016116926087833474,
+      "loss": 1.1465,
+      "step": 6482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016115693592737117,
+      "loss": 1.1089,
+      "step": 6483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016114460949216136,
+      "loss": 1.1103,
+      "step": 6484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001611322815730046,
+      "loss": 1.1369,
+      "step": 6485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016111995217019999,
+      "loss": 1.1466,
+      "step": 6486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016110762128404677,
+      "loss": 1.2195,
+      "step": 6487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016109528891484424,
+      "loss": 1.1575,
+      "step": 6488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016108295506289167,
+      "loss": 1.2323,
+      "step": 6489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016107061972848842,
+      "loss": 1.1772,
+      "step": 6490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016105828291193384,
+      "loss": 1.2202,
+      "step": 6491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016104594461352733,
+      "loss": 1.2,
+      "step": 6492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016103360483356836,
+      "loss": 1.1755,
+      "step": 6493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016102126357235642,
+      "loss": 1.2184,
+      "step": 6494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000161008920830191,
+      "loss": 1.2011,
+      "step": 6495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016099657660737163,
+      "loss": 1.1857,
+      "step": 6496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00016098423090419795,
+      "loss": 1.3436,
+      "step": 6497
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016097188372096954,
+      "loss": 0.9945,
+      "step": 6498
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016095953505798608,
+      "loss": 1.2463,
+      "step": 6499
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016094718491554727,
+      "loss": 1.1795,
+      "step": 6500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016093483329395283,
+      "loss": 1.132,
+      "step": 6501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016092248019350253,
+      "loss": 1.2113,
+      "step": 6502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016091012561449616,
+      "loss": 1.2147,
+      "step": 6503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016089776955723363,
+      "loss": 1.0877,
+      "step": 6504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001608854120220147,
+      "loss": 1.2093,
+      "step": 6505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016087305300913936,
+      "loss": 1.1756,
+      "step": 6506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001608606925189075,
+      "loss": 1.0406,
+      "step": 6507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001608483305516192,
+      "loss": 1.115,
+      "step": 6508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001608359671075744,
+      "loss": 1.3275,
+      "step": 6509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016082360218707314,
+      "loss": 1.202,
+      "step": 6510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016081123579041558,
+      "loss": 1.2725,
+      "step": 6511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016079886791790177,
+      "loss": 1.286,
+      "step": 6512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016078649856983193,
+      "loss": 1.2192,
+      "step": 6513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016077412774650625,
+      "loss": 1.1547,
+      "step": 6514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016076175544822493,
+      "loss": 1.3667,
+      "step": 6515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016074938167528826,
+      "loss": 1.1535,
+      "step": 6516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016073700642799655,
+      "loss": 1.0674,
+      "step": 6517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016072462970665013,
+      "loss": 1.085,
+      "step": 6518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016071225151154936,
+      "loss": 1.0777,
+      "step": 6519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001606998718429947,
+      "loss": 1.1592,
+      "step": 6520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016068749070128656,
+      "loss": 1.1301,
+      "step": 6521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016067510808672543,
+      "loss": 1.0543,
+      "step": 6522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016066272399961186,
+      "loss": 1.199,
+      "step": 6523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016065033844024634,
+      "loss": 1.1359,
+      "step": 6524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016063795140892952,
+      "loss": 1.0793,
+      "step": 6525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000160625562905962,
+      "loss": 1.0478,
+      "step": 6526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016061317293164447,
+      "loss": 1.1373,
+      "step": 6527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016060078148627757,
+      "loss": 1.2422,
+      "step": 6528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001605883885701621,
+      "loss": 1.2601,
+      "step": 6529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001605759941835988,
+      "loss": 1.3709,
+      "step": 6530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016056359832688846,
+      "loss": 1.1955,
+      "step": 6531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016055120100033195,
+      "loss": 1.297,
+      "step": 6532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001605388022042301,
+      "loss": 1.2229,
+      "step": 6533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016052640193888387,
+      "loss": 1.3433,
+      "step": 6534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001605140002045942,
+      "loss": 1.1552,
+      "step": 6535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016050159700166207,
+      "loss": 1.2689,
+      "step": 6536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001604891923303885,
+      "loss": 1.1378,
+      "step": 6537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016047678619107454,
+      "loss": 1.1884,
+      "step": 6538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016046437858402128,
+      "loss": 1.2335,
+      "step": 6539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001604519695095298,
+      "loss": 1.1944,
+      "step": 6540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016043955896790138,
+      "loss": 1.1676,
+      "step": 6541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001604271469594371,
+      "loss": 1.1745,
+      "step": 6542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016041473348443823,
+      "loss": 0.9025,
+      "step": 6543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016040231854320608,
+      "loss": 1.0981,
+      "step": 6544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001603899021360419,
+      "loss": 1.2441,
+      "step": 6545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016037748426324706,
+      "loss": 1.4108,
+      "step": 6546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001603650649251229,
+      "loss": 1.1146,
+      "step": 6547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016035264412197087,
+      "loss": 2.3779,
+      "step": 6548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016034022185409242,
+      "loss": 1.0401,
+      "step": 6549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000160327798121789,
+      "loss": 1.3221,
+      "step": 6550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016031537292536213,
+      "loss": 1.3252,
+      "step": 6551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016030294626511338,
+      "loss": 1.2445,
+      "step": 6552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016029051814134434,
+      "loss": 1.143,
+      "step": 6553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001602780885543566,
+      "loss": 1.2606,
+      "step": 6554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016026565750445188,
+      "loss": 1.1866,
+      "step": 6555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016025322499193186,
+      "loss": 1.2123,
+      "step": 6556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001602407910170982,
+      "loss": 0.9934,
+      "step": 6557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016022835558025274,
+      "loss": 1.1087,
+      "step": 6558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016021591868169724,
+      "loss": 1.2287,
+      "step": 6559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001602034803217336,
+      "loss": 1.1332,
+      "step": 6560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001601910405006636,
+      "loss": 1.1159,
+      "step": 6561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016017859921878922,
+      "loss": 1.1534,
+      "step": 6562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016016615647641236,
+      "loss": 1.2684,
+      "step": 6563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016015371227383503,
+      "loss": 1.1733,
+      "step": 6564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016014126661135922,
+      "loss": 1.2606,
+      "step": 6565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016012881948928703,
+      "loss": 1.0253,
+      "step": 6566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016011637090792048,
+      "loss": 1.188,
+      "step": 6567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016010392086756168,
+      "loss": 1.3052,
+      "step": 6568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016009146936851287,
+      "loss": 1.2114,
+      "step": 6569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016007901641107618,
+      "loss": 1.2116,
+      "step": 6570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016006656199555382,
+      "loss": 1.2888,
+      "step": 6571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016005410612224814,
+      "loss": 2.5581,
+      "step": 6572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00016004164879146132,
+      "loss": 1.2024,
+      "step": 6573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001600291900034958,
+      "loss": 1.2189,
+      "step": 6574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001600167297586539,
+      "loss": 1.1918,
+      "step": 6575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000160004268057238,
+      "loss": 1.071,
+      "step": 6576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015999180489955055,
+      "loss": 1.1275,
+      "step": 6577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001599793402858941,
+      "loss": 1.0842,
+      "step": 6578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015996687421657103,
+      "loss": 1.3051,
+      "step": 6579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015995440669188397,
+      "loss": 1.1456,
+      "step": 6580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001599419377121355,
+      "loss": 1.1347,
+      "step": 6581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015992946727762822,
+      "loss": 1.1127,
+      "step": 6582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015991699538866477,
+      "loss": 1.1121,
+      "step": 6583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015990452204554781,
+      "loss": 1.0475,
+      "step": 6584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015989204724858012,
+      "loss": 1.1097,
+      "step": 6585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015987957099806443,
+      "loss": 1.1696,
+      "step": 6586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015986709329430355,
+      "loss": 1.122,
+      "step": 6587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015985461413760028,
+      "loss": 1.3092,
+      "step": 6588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015984213352825748,
+      "loss": 0.9938,
+      "step": 6589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015982965146657806,
+      "loss": 1.2082,
+      "step": 6590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000159817167952865,
+      "loss": 1.0197,
+      "step": 6591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015980468298742117,
+      "loss": 1.2099,
+      "step": 6592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015979219657054965,
+      "loss": 1.069,
+      "step": 6593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015977970870255343,
+      "loss": 1.3456,
+      "step": 6594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015976721938373566,
+      "loss": 1.1616,
+      "step": 6595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015975472861439935,
+      "loss": 1.2967,
+      "step": 6596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015974223639484772,
+      "loss": 1.22,
+      "step": 6597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001597297427253839,
+      "loss": 1.247,
+      "step": 6598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015971724760631114,
+      "loss": 1.0927,
+      "step": 6599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015970475103793265,
+      "loss": 1.1083,
+      "step": 6600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00015969225302055176,
+      "loss": 1.2908,
+      "step": 6601
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015967975355447178,
+      "loss": 2.4893,
+      "step": 6602
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015966725263999606,
+      "loss": 1.2391,
+      "step": 6603
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015965475027742798,
+      "loss": 1.1915,
+      "step": 6604
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015964224646707094,
+      "loss": 1.1297,
+      "step": 6605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015962974120922846,
+      "loss": 1.1276,
+      "step": 6606
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015961723450420403,
+      "loss": 1.173,
+      "step": 6607
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015960472635230115,
+      "loss": 1.2711,
+      "step": 6608
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001595922167538234,
+      "loss": 1.3449,
+      "step": 6609
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015957970570907435,
+      "loss": 1.2421,
+      "step": 6610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015956719321835768,
+      "loss": 1.3264,
+      "step": 6611
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015955467928197706,
+      "loss": 1.2686,
+      "step": 6612
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015954216390023614,
+      "loss": 1.2035,
+      "step": 6613
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015952964707343875,
+      "loss": 1.3647,
+      "step": 6614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015951712880188859,
+      "loss": 1.286,
+      "step": 6615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001595046090858895,
+      "loss": 1.2856,
+      "step": 6616
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015949208792574536,
+      "loss": 1.149,
+      "step": 6617
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015947956532176,
+      "loss": 1.1432,
+      "step": 6618
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015946704127423736,
+      "loss": 1.1485,
+      "step": 6619
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015945451578348142,
+      "loss": 1.2444,
+      "step": 6620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001594419888497961,
+      "loss": 1.2773,
+      "step": 6621
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015942946047348548,
+      "loss": 1.1707,
+      "step": 6622
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015941693065485354,
+      "loss": 1.0938,
+      "step": 6623
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015940439939420448,
+      "loss": 1.2451,
+      "step": 6624
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015939186669184242,
+      "loss": 1.4839,
+      "step": 6625
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015937933254807138,
+      "loss": 1.0957,
+      "step": 6626
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001593667969631957,
+      "loss": 1.2743,
+      "step": 6627
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015935425993751957,
+      "loss": 1.1065,
+      "step": 6628
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015934172147134726,
+      "loss": 1.2303,
+      "step": 6629
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015932918156498304,
+      "loss": 1.132,
+      "step": 6630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015931664021873128,
+      "loss": 1.2043,
+      "step": 6631
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015930409743289636,
+      "loss": 1.0407,
+      "step": 6632
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015929155320778263,
+      "loss": 0.9893,
+      "step": 6633
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001592790075436946,
+      "loss": 1.1201,
+      "step": 6634
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001592664604409367,
+      "loss": 1.1703,
+      "step": 6635
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001592539118998135,
+      "loss": 2.519,
+      "step": 6636
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001592413619206295,
+      "loss": 1.2656,
+      "step": 6637
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015922881050368926,
+      "loss": 1.2476,
+      "step": 6638
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001592162576492974,
+      "loss": 1.2786,
+      "step": 6639
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015920370335775866,
+      "loss": 1.3118,
+      "step": 6640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015919114762937762,
+      "loss": 1.0374,
+      "step": 6641
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015917859046445904,
+      "loss": 2.4082,
+      "step": 6642
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001591660318633077,
+      "loss": 1.3538,
+      "step": 6643
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015915347182622833,
+      "loss": 1.0594,
+      "step": 6644
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015914091035352582,
+      "loss": 1.1869,
+      "step": 6645
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015912834744550498,
+      "loss": 1.1133,
+      "step": 6646
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015911578310247077,
+      "loss": 1.1211,
+      "step": 6647
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015910321732472805,
+      "loss": 1.1533,
+      "step": 6648
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015909065011258182,
+      "loss": 1.213,
+      "step": 6649
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015907808146633706,
+      "loss": 1.2555,
+      "step": 6650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015906551138629885,
+      "loss": 1.2706,
+      "step": 6651
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001590529398727722,
+      "loss": 1.1798,
+      "step": 6652
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015904036692606223,
+      "loss": 1.155,
+      "step": 6653
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015902779254647412,
+      "loss": 1.3831,
+      "step": 6654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015901521673431299,
+      "loss": 1.0131,
+      "step": 6655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015900263948988408,
+      "loss": 1.1466,
+      "step": 6656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015899006081349258,
+      "loss": 1.2485,
+      "step": 6657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015897748070544388,
+      "loss": 1.1418,
+      "step": 6658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015896489916604315,
+      "loss": 1.1772,
+      "step": 6659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015895231619559588,
+      "loss": 1.0818,
+      "step": 6660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015893973179440733,
+      "loss": 1.2241,
+      "step": 6661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015892714596278302,
+      "loss": 1.3236,
+      "step": 6662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001589145587010283,
+      "loss": 1.1962,
+      "step": 6663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015890197000944874,
+      "loss": 2.4041,
+      "step": 6664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001588893798883498,
+      "loss": 1.1319,
+      "step": 6665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001588767883380371,
+      "loss": 1.1543,
+      "step": 6666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001588641953588162,
+      "loss": 1.2061,
+      "step": 6667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015885160095099266,
+      "loss": 1.3038,
+      "step": 6668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015883900511487223,
+      "loss": 1.1702,
+      "step": 6669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015882640785076058,
+      "loss": 1.2952,
+      "step": 6670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015881380915896344,
+      "loss": 1.1482,
+      "step": 6671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015880120903978654,
+      "loss": 1.0811,
+      "step": 6672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001587886074935357,
+      "loss": 1.1407,
+      "step": 6673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015877600452051676,
+      "loss": 1.2606,
+      "step": 6674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015876340012103559,
+      "loss": 1.1165,
+      "step": 6675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015875079429539807,
+      "loss": 1.1367,
+      "step": 6676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015873818704391018,
+      "loss": 1.3364,
+      "step": 6677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015872557836687788,
+      "loss": 1.2112,
+      "step": 6678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015871296826460712,
+      "loss": 1.1545,
+      "step": 6679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015870035673740395,
+      "loss": 1.0111,
+      "step": 6680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015868774378557455,
+      "loss": 1.2435,
+      "step": 6681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015867512940942488,
+      "loss": 1.293,
+      "step": 6682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015866251360926122,
+      "loss": 1.0821,
+      "step": 6683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015864989638538965,
+      "loss": 1.4182,
+      "step": 6684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015863727773811643,
+      "loss": 1.2094,
+      "step": 6685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001586246576677478,
+      "loss": 1.0861,
+      "step": 6686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015861203617459002,
+      "loss": 1.1513,
+      "step": 6687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015859941325894946,
+      "loss": 1.098,
+      "step": 6688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015858678892113242,
+      "loss": 1.0703,
+      "step": 6689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001585741631614453,
+      "loss": 1.2994,
+      "step": 6690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015856153598019451,
+      "loss": 1.0711,
+      "step": 6691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015854890737768655,
+      "loss": 1.1625,
+      "step": 6692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015853627735422786,
+      "loss": 1.2287,
+      "step": 6693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015852364591012501,
+      "loss": 1.0979,
+      "step": 6694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015851101304568453,
+      "loss": 1.1489,
+      "step": 6695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015849837876121296,
+      "loss": 1.1498,
+      "step": 6696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015848574305701702,
+      "loss": 1.1786,
+      "step": 6697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015847310593340334,
+      "loss": 2.3611,
+      "step": 6698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015846046739067862,
+      "loss": 0.9598,
+      "step": 6699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015844782742914957,
+      "loss": 1.265,
+      "step": 6700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015843518604912292,
+      "loss": 1.2788,
+      "step": 6701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015842254325090557,
+      "loss": 1.0931,
+      "step": 6702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015840989903480427,
+      "loss": 1.1712,
+      "step": 6703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015839725340112594,
+      "loss": 1.2651,
+      "step": 6704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00015838460635017745,
+      "loss": 1.2,
+      "step": 6705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015837195788226578,
+      "loss": 1.3133,
+      "step": 6706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015835930799769783,
+      "loss": 1.3112,
+      "step": 6707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015834665669678063,
+      "loss": 1.0299,
+      "step": 6708
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015833400397982127,
+      "loss": 0.9875,
+      "step": 6709
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015832134984712678,
+      "loss": 2.5764,
+      "step": 6710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001583086942990043,
+      "loss": 1.1802,
+      "step": 6711
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015829603733576088,
+      "loss": 1.1961,
+      "step": 6712
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015828337895770384,
+      "loss": 1.0707,
+      "step": 6713
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001582707191651403,
+      "loss": 1.0275,
+      "step": 6714
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015825805795837753,
+      "loss": 1.1951,
+      "step": 6715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001582453953377228,
+      "loss": 1.2023,
+      "step": 6716
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015823273130348344,
+      "loss": 1.2477,
+      "step": 6717
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001582200658559668,
+      "loss": 1.3846,
+      "step": 6718
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015820739899548026,
+      "loss": 1.1657,
+      "step": 6719
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015819473072233123,
+      "loss": 1.3185,
+      "step": 6720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001581820610368272,
+      "loss": 1.1746,
+      "step": 6721
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015816938993927555,
+      "loss": 1.131,
+      "step": 6722
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015815671742998392,
+      "loss": 1.1824,
+      "step": 6723
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015814404350925982,
+      "loss": 1.1732,
+      "step": 6724
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015813136817741083,
+      "loss": 0.8935,
+      "step": 6725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001581186914347446,
+      "loss": 1.0835,
+      "step": 6726
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015810601328156872,
+      "loss": 1.2515,
+      "step": 6727
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015809333371819099,
+      "loss": 1.1063,
+      "step": 6728
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015808065274491905,
+      "loss": 1.2293,
+      "step": 6729
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001580679703620607,
+      "loss": 1.3073,
+      "step": 6730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015805528656992373,
+      "loss": 1.1863,
+      "step": 6731
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015804260136881594,
+      "loss": 1.2966,
+      "step": 6732
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015802991475904522,
+      "loss": 1.2191,
+      "step": 6733
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015801722674091949,
+      "loss": 1.2299,
+      "step": 6734
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015800453731474665,
+      "loss": 1.0443,
+      "step": 6735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015799184648083463,
+      "loss": 1.0395,
+      "step": 6736
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001579791542394915,
+      "loss": 1.0151,
+      "step": 6737
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015796646059102525,
+      "loss": 1.0106,
+      "step": 6738
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015795376553574396,
+      "loss": 1.1845,
+      "step": 6739
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015794106907395577,
+      "loss": 1.0795,
+      "step": 6740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015792837120596874,
+      "loss": 1.2024,
+      "step": 6741
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015791567193209108,
+      "loss": 1.0712,
+      "step": 6742
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015790297125263102,
+      "loss": 1.3019,
+      "step": 6743
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001578902691678968,
+      "loss": 1.1107,
+      "step": 6744
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001578775656781966,
+      "loss": 1.1445,
+      "step": 6745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015786486078383884,
+      "loss": 1.2789,
+      "step": 6746
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015785215448513176,
+      "loss": 1.1409,
+      "step": 6747
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015783944678238385,
+      "loss": 1.1041,
+      "step": 6748
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015782673767590342,
+      "loss": 1.0626,
+      "step": 6749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015781402716599901,
+      "loss": 0.9619,
+      "step": 6750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000157801315252979,
+      "loss": 1.2018,
+      "step": 6751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015778860193715194,
+      "loss": 1.2606,
+      "step": 6752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015777588721882637,
+      "loss": 1.0681,
+      "step": 6753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015776317109831087,
+      "loss": 1.1451,
+      "step": 6754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015775045357591406,
+      "loss": 2.3049,
+      "step": 6755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015773773465194458,
+      "loss": 1.0486,
+      "step": 6756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015772501432671118,
+      "loss": 1.2403,
+      "step": 6757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015771229260052246,
+      "loss": 1.2606,
+      "step": 6758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015769956947368722,
+      "loss": 1.2249,
+      "step": 6759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015768684494651427,
+      "loss": 1.3074,
+      "step": 6760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015767411901931237,
+      "loss": 1.1304,
+      "step": 6761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015766139169239046,
+      "loss": 1.2428,
+      "step": 6762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001576486629660573,
+      "loss": 1.0433,
+      "step": 6763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015763593284062193,
+      "loss": 1.2897,
+      "step": 6764
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015762320131639326,
+      "loss": 1.294,
+      "step": 6765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015761046839368024,
+      "loss": 2.6724,
+      "step": 6766
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015759773407279193,
+      "loss": 1.2115,
+      "step": 6767
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015758499835403738,
+      "loss": 1.1821,
+      "step": 6768
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015757226123772568,
+      "loss": 1.2045,
+      "step": 6769
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015755952272416597,
+      "loss": 1.2468,
+      "step": 6770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015754678281366732,
+      "loss": 1.1871,
+      "step": 6771
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015753404150653906,
+      "loss": 0.9786,
+      "step": 6772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001575212988030903,
+      "loss": 1.1623,
+      "step": 6773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015750855470363038,
+      "loss": 1.1604,
+      "step": 6774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015749580920846853,
+      "loss": 1.201,
+      "step": 6775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015748306231791408,
+      "loss": 1.1053,
+      "step": 6776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015747031403227641,
+      "loss": 1.2111,
+      "step": 6777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015745756435186494,
+      "loss": 1.1058,
+      "step": 6778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015744481327698907,
+      "loss": 1.1889,
+      "step": 6779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015743206080795824,
+      "loss": 1.2668,
+      "step": 6780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015741930694508201,
+      "loss": 1.1102,
+      "step": 6781
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015740655168866984,
+      "loss": 1.2445,
+      "step": 6782
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001573937950390313,
+      "loss": 1.2738,
+      "step": 6783
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000157381036996476,
+      "loss": 1.2188,
+      "step": 6784
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015736827756131363,
+      "loss": 1.1936,
+      "step": 6785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001573555167338538,
+      "loss": 1.2577,
+      "step": 6786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001573427545144062,
+      "loss": 1.1439,
+      "step": 6787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015732999090328054,
+      "loss": 1.2584,
+      "step": 6788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015731722590078668,
+      "loss": 1.1047,
+      "step": 6789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015730445950723433,
+      "loss": 0.9828,
+      "step": 6790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015729169172293338,
+      "loss": 1.2434,
+      "step": 6791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015727892254819363,
+      "loss": 1.1622,
+      "step": 6792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015726615198332505,
+      "loss": 1.1852,
+      "step": 6793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015725338002863754,
+      "loss": 1.3051,
+      "step": 6794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001572406066844411,
+      "loss": 1.2419,
+      "step": 6795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015722783195104567,
+      "loss": 1.2135,
+      "step": 6796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015721505582876134,
+      "loss": 1.2041,
+      "step": 6797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015720227831789816,
+      "loss": 1.2029,
+      "step": 6798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015718949941876623,
+      "loss": 1.2554,
+      "step": 6799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001571767191316757,
+      "loss": 1.2276,
+      "step": 6800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015716393745693677,
+      "loss": 1.2083,
+      "step": 6801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015715115439485957,
+      "loss": 1.1242,
+      "step": 6802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015713836994575437,
+      "loss": 1.1153,
+      "step": 6803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015712558410993147,
+      "loss": 1.2618,
+      "step": 6804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015711279688770114,
+      "loss": 1.2505,
+      "step": 6805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015710000827937374,
+      "loss": 1.1967,
+      "step": 6806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015708721828525964,
+      "loss": 1.1779,
+      "step": 6807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00015707442690566926,
+      "loss": 1.0724,
+      "step": 6808
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015706163414091302,
+      "loss": 1.1929,
+      "step": 6809
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015704883999130136,
+      "loss": 1.161,
+      "step": 6810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015703604445714485,
+      "loss": 1.1852,
+      "step": 6811
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015702324753875398,
+      "loss": 1.3079,
+      "step": 6812
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001570104492364394,
+      "loss": 1.2516,
+      "step": 6813
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001569976495505116,
+      "loss": 1.0976,
+      "step": 6814
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015698484848128134,
+      "loss": 1.1469,
+      "step": 6815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015697204602905924,
+      "loss": 1.2476,
+      "step": 6816
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000156959242194156,
+      "loss": 1.2312,
+      "step": 6817
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015694643697688237,
+      "loss": 1.3271,
+      "step": 6818
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015693363037754915,
+      "loss": 1.0839,
+      "step": 6819
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001569208223964671,
+      "loss": 1.0392,
+      "step": 6820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001569080130339471,
+      "loss": 1.0099,
+      "step": 6821
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001568952022903,
+      "loss": 1.1819,
+      "step": 6822
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015688239016583677,
+      "loss": 1.0889,
+      "step": 6823
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001568695766608683,
+      "loss": 1.2151,
+      "step": 6824
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015685676177570556,
+      "loss": 1.1796,
+      "step": 6825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015684394551065958,
+      "loss": 1.2155,
+      "step": 6826
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015683112786604145,
+      "loss": 1.0731,
+      "step": 6827
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015681830884216213,
+      "loss": 1.1595,
+      "step": 6828
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015680548843933286,
+      "loss": 1.1477,
+      "step": 6829
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001567926666578647,
+      "loss": 1.2269,
+      "step": 6830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015677984349806885,
+      "loss": 1.1998,
+      "step": 6831
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015676701896025653,
+      "loss": 1.1344,
+      "step": 6832
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015675419304473896,
+      "loss": 1.1356,
+      "step": 6833
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015674136575182746,
+      "loss": 1.0809,
+      "step": 6834
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001567285370818333,
+      "loss": 1.1756,
+      "step": 6835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015671570703506787,
+      "loss": 1.0757,
+      "step": 6836
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015670287561184252,
+      "loss": 0.9814,
+      "step": 6837
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015669004281246865,
+      "loss": 1.2361,
+      "step": 6838
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015667720863725776,
+      "loss": 1.0443,
+      "step": 6839
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015666437308652122,
+      "loss": 1.2362,
+      "step": 6840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015665153616057064,
+      "loss": 1.2892,
+      "step": 6841
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015663869785971754,
+      "loss": 2.5728,
+      "step": 6842
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015662585818427348,
+      "loss": 1.0707,
+      "step": 6843
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015661301713455013,
+      "loss": 1.26,
+      "step": 6844
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015660017471085905,
+      "loss": 1.1577,
+      "step": 6845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015658733091351198,
+      "loss": 1.1278,
+      "step": 6846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001565744857428206,
+      "loss": 1.124,
+      "step": 6847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015656163919909664,
+      "loss": 1.149,
+      "step": 6848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015654879128265193,
+      "loss": 1.2242,
+      "step": 6849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015653594199379825,
+      "loss": 2.3938,
+      "step": 6850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015652309133284747,
+      "loss": 1.2317,
+      "step": 6851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001565102393001115,
+      "loss": 1.0862,
+      "step": 6852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015649738589590214,
+      "loss": 1.313,
+      "step": 6853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001564845311205314,
+      "loss": 1.2664,
+      "step": 6854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015647167497431127,
+      "loss": 2.4277,
+      "step": 6855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015645881745755377,
+      "loss": 2.3853,
+      "step": 6856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001564459585705709,
+      "loss": 1.1295,
+      "step": 6857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015643309831367476,
+      "loss": 1.2857,
+      "step": 6858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001564202366871775,
+      "loss": 1.1124,
+      "step": 6859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015640737369139123,
+      "loss": 1.3644,
+      "step": 6860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001563945093266281,
+      "loss": 1.105,
+      "step": 6861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015638164359320033,
+      "loss": 2.6548,
+      "step": 6862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001563687764914202,
+      "loss": 1.2282,
+      "step": 6863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015635590802159997,
+      "loss": 1.2847,
+      "step": 6864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015634303818405198,
+      "loss": 1.2569,
+      "step": 6865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001563301669790885,
+      "loss": 1.1417,
+      "step": 6866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000156317294407022,
+      "loss": 1.262,
+      "step": 6867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015630442046816486,
+      "loss": 1.1958,
+      "step": 6868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015629154516282946,
+      "loss": 1.131,
+      "step": 6869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015627866849132836,
+      "loss": 1.133,
+      "step": 6870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015626579045397397,
+      "loss": 1.3343,
+      "step": 6871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015625291105107896,
+      "loss": 1.1615,
+      "step": 6872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001562400302829558,
+      "loss": 1.0829,
+      "step": 6873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015622714814991716,
+      "loss": 1.2144,
+      "step": 6874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015621426465227569,
+      "loss": 1.2899,
+      "step": 6875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015620137979034405,
+      "loss": 1.2154,
+      "step": 6876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001561884935644349,
+      "loss": 2.3953,
+      "step": 6877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015617560597486102,
+      "loss": 1.1975,
+      "step": 6878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001561627170219352,
+      "loss": 1.1445,
+      "step": 6879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001561498267059702,
+      "loss": 1.2251,
+      "step": 6880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015613693502727896,
+      "loss": 0.974,
+      "step": 6881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001561240419861743,
+      "loss": 1.1558,
+      "step": 6882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015611114758296908,
+      "loss": 1.3279,
+      "step": 6883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015609825181797627,
+      "loss": 1.1212,
+      "step": 6884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015608535469150884,
+      "loss": 1.3126,
+      "step": 6885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015607245620387981,
+      "loss": 1.2352,
+      "step": 6886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015605955635540224,
+      "loss": 1.1627,
+      "step": 6887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001560466551463892,
+      "loss": 1.196,
+      "step": 6888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001560337525771537,
+      "loss": 1.0722,
+      "step": 6889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015602084864800903,
+      "loss": 1.244,
+      "step": 6890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015600794335926825,
+      "loss": 1.0258,
+      "step": 6891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001559950367112446,
+      "loss": 1.298,
+      "step": 6892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015598212870425132,
+      "loss": 1.2456,
+      "step": 6893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015596921933860167,
+      "loss": 1.0516,
+      "step": 6894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015595630861460897,
+      "loss": 1.1098,
+      "step": 6895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015594339653258652,
+      "loss": 1.1221,
+      "step": 6896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015593048309284774,
+      "loss": 1.0243,
+      "step": 6897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000155917568295706,
+      "loss": 1.0064,
+      "step": 6898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015590465214147475,
+      "loss": 1.2021,
+      "step": 6899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015589173463046745,
+      "loss": 1.2306,
+      "step": 6900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001558788157629976,
+      "loss": 1.1696,
+      "step": 6901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015586589553937872,
+      "loss": 1.179,
+      "step": 6902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015585297395992443,
+      "loss": 1.1063,
+      "step": 6903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015584005102494826,
+      "loss": 1.2099,
+      "step": 6904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015582712673476392,
+      "loss": 1.0548,
+      "step": 6905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000155814201089685,
+      "loss": 1.3226,
+      "step": 6906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001558012740900252,
+      "loss": 1.1617,
+      "step": 6907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015578834573609833,
+      "loss": 1.2401,
+      "step": 6908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001557754160282181,
+      "loss": 1.277,
+      "step": 6909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015576248496669827,
+      "loss": 1.1976,
+      "step": 6910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015574955255185275,
+      "loss": 1.1555,
+      "step": 6911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00015573661878399538,
+      "loss": 2.7163,
+      "step": 6912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015572368366344001,
+      "loss": 1.3894,
+      "step": 6913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015571074719050062,
+      "loss": 1.2579,
+      "step": 6914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015569780936549116,
+      "loss": 1.3831,
+      "step": 6915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015568487018872558,
+      "loss": 1.1563,
+      "step": 6916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015567192966051797,
+      "loss": 1.2869,
+      "step": 6917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015565898778118235,
+      "loss": 1.2869,
+      "step": 6918
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015564604455103282,
+      "loss": 1.1591,
+      "step": 6919
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015563309997038353,
+      "loss": 1.111,
+      "step": 6920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001556201540395486,
+      "loss": 0.9948,
+      "step": 6921
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015560720675884224,
+      "loss": 1.2017,
+      "step": 6922
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015559425812857865,
+      "loss": 1.2028,
+      "step": 6923
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015558130814907214,
+      "loss": 1.249,
+      "step": 6924
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015556835682063697,
+      "loss": 1.1631,
+      "step": 6925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015555540414358745,
+      "loss": 1.0414,
+      "step": 6926
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015554245011823793,
+      "loss": 1.1594,
+      "step": 6927
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015552949474490285,
+      "loss": 1.1182,
+      "step": 6928
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015551653802389653,
+      "loss": 1.2061,
+      "step": 6929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015550357995553353,
+      "loss": 1.1917,
+      "step": 6930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015549062054012825,
+      "loss": 1.0732,
+      "step": 6931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015547765977799532,
+      "loss": 1.3424,
+      "step": 6932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015546469766944916,
+      "loss": 1.1274,
+      "step": 6933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015545173421480443,
+      "loss": 1.2107,
+      "step": 6934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015543876941437573,
+      "loss": 1.3066,
+      "step": 6935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001554258032684777,
+      "loss": 1.1464,
+      "step": 6936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000155412835777425,
+      "loss": 1.1502,
+      "step": 6937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015539986694153242,
+      "loss": 1.1157,
+      "step": 6938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015538689676111462,
+      "loss": 1.2471,
+      "step": 6939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015537392523648643,
+      "loss": 1.0912,
+      "step": 6940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015536095236796265,
+      "loss": 1.3367,
+      "step": 6941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015534797815585814,
+      "loss": 1.1525,
+      "step": 6942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015533500260048774,
+      "loss": 1.0264,
+      "step": 6943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015532202570216642,
+      "loss": 1.1165,
+      "step": 6944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015530904746120902,
+      "loss": 1.1644,
+      "step": 6945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015529606787793062,
+      "loss": 1.2442,
+      "step": 6946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015528308695264618,
+      "loss": 1.1971,
+      "step": 6947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015527010468567077,
+      "loss": 2.3442,
+      "step": 6948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015525712107731943,
+      "loss": 1.0862,
+      "step": 6949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015524413612790725,
+      "loss": 2.4167,
+      "step": 6950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001552311498377494,
+      "loss": 1.1199,
+      "step": 6951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015521816220716105,
+      "loss": 1.2692,
+      "step": 6952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001552051732364574,
+      "loss": 1.1121,
+      "step": 6953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015519218292595367,
+      "loss": 1.2349,
+      "step": 6954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015517919127596515,
+      "loss": 1.3504,
+      "step": 6955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015516619828680712,
+      "loss": 1.0017,
+      "step": 6956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015515320395879495,
+      "loss": 1.2182,
+      "step": 6957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015514020829224392,
+      "loss": 1.3563,
+      "step": 6958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015512721128746953,
+      "loss": 1.1149,
+      "step": 6959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015511421294478714,
+      "loss": 1.1934,
+      "step": 6960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015510121326451227,
+      "loss": 1.1814,
+      "step": 6961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015508821224696035,
+      "loss": 1.1947,
+      "step": 6962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015507520989244697,
+      "loss": 1.1565,
+      "step": 6963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015506220620128765,
+      "loss": 1.2127,
+      "step": 6964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015504920117379798,
+      "loss": 1.1229,
+      "step": 6965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001550361948102936,
+      "loss": 1.063,
+      "step": 6966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001550231871110902,
+      "loss": 1.0636,
+      "step": 6967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001550101780765034,
+      "loss": 1.061,
+      "step": 6968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015499716770684898,
+      "loss": 1.2081,
+      "step": 6969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015498415600244267,
+      "loss": 1.4154,
+      "step": 6970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015497114296360027,
+      "loss": 1.0428,
+      "step": 6971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015495812859063758,
+      "loss": 1.1684,
+      "step": 6972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015494511288387047,
+      "loss": 1.3309,
+      "step": 6973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015493209584361482,
+      "loss": 1.2145,
+      "step": 6974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015491907747018651,
+      "loss": 1.1775,
+      "step": 6975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015490605776390156,
+      "loss": 1.157,
+      "step": 6976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015489303672507586,
+      "loss": 1.1124,
+      "step": 6977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015488001435402557,
+      "loss": 1.1526,
+      "step": 6978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015486699065106655,
+      "loss": 1.3598,
+      "step": 6979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015485396561651501,
+      "loss": 1.1451,
+      "step": 6980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015484093925068705,
+      "loss": 0.8984,
+      "step": 6981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015482791155389876,
+      "loss": 1.0111,
+      "step": 6982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015481488252646632,
+      "loss": 1.1468,
+      "step": 6983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015480185216870597,
+      "loss": 1.2036,
+      "step": 6984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015478882048093398,
+      "loss": 1.0789,
+      "step": 6985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015477578746346654,
+      "loss": 1.2489,
+      "step": 6986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015476275311661998,
+      "loss": 1.2898,
+      "step": 6987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001547497174407107,
+      "loss": 1.1837,
+      "step": 6988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015473668043605498,
+      "loss": 1.1082,
+      "step": 6989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015472364210296927,
+      "loss": 1.1337,
+      "step": 6990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015471060244177,
+      "loss": 1.1966,
+      "step": 6991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015469756145277362,
+      "loss": 1.1444,
+      "step": 6992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015468451913629668,
+      "loss": 1.1775,
+      "step": 6993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001546714754926556,
+      "loss": 1.1382,
+      "step": 6994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015465843052216705,
+      "loss": 1.1294,
+      "step": 6995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015464538422514757,
+      "loss": 2.5977,
+      "step": 6996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015463233660191377,
+      "loss": 1.3453,
+      "step": 6997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015461928765278242,
+      "loss": 1.2169,
+      "step": 6998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015460623737807008,
+      "loss": 1.0645,
+      "step": 6999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015459318577809355,
+      "loss": 1.2552,
+      "step": 7000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015458013285316952,
+      "loss": 1.1697,
+      "step": 7001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015456707860361487,
+      "loss": 1.1694,
+      "step": 7002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015455402302974632,
+      "loss": 1.1136,
+      "step": 7003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015454096613188076,
+      "loss": 1.2673,
+      "step": 7004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015452790791033514,
+      "loss": 1.1061,
+      "step": 7005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015451484836542627,
+      "loss": 2.6863,
+      "step": 7006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015450178749747118,
+      "loss": 1.1222,
+      "step": 7007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001544887253067868,
+      "loss": 1.1534,
+      "step": 7008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015447566179369016,
+      "loss": 1.2307,
+      "step": 7009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001544625969584983,
+      "loss": 1.1315,
+      "step": 7010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001544495308015283,
+      "loss": 1.1793,
+      "step": 7011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015443646332309724,
+      "loss": 1.1255,
+      "step": 7012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015442339452352235,
+      "loss": 1.3375,
+      "step": 7013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015441032440312072,
+      "loss": 1.1622,
+      "step": 7014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015439725296220958,
+      "loss": 1.0781,
+      "step": 7015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00015438418020110613,
+      "loss": 1.142,
+      "step": 7016
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001543711061201277,
+      "loss": 1.0308,
+      "step": 7017
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015435803071959159,
+      "loss": 1.2104,
+      "step": 7018
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015434495399981504,
+      "loss": 1.169,
+      "step": 7019
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015433187596111556,
+      "loss": 1.2487,
+      "step": 7020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015431879660381044,
+      "loss": 2.9463,
+      "step": 7021
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001543057159282171,
+      "loss": 1.093,
+      "step": 7022
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001542926339346531,
+      "loss": 1.136,
+      "step": 7023
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015427955062343585,
+      "loss": 1.1501,
+      "step": 7024
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001542664659948829,
+      "loss": 0.9585,
+      "step": 7025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001542533800493118,
+      "loss": 1.329,
+      "step": 7026
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001542402927870401,
+      "loss": 1.1988,
+      "step": 7027
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015422720420838552,
+      "loss": 2.804,
+      "step": 7028
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015421411431366564,
+      "loss": 1.2383,
+      "step": 7029
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015420102310319817,
+      "loss": 1.1901,
+      "step": 7030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001541879305773008,
+      "loss": 1.0482,
+      "step": 7031
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015417483673629132,
+      "loss": 1.1073,
+      "step": 7032
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015416174158048746,
+      "loss": 1.33,
+      "step": 7033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001541486451102071,
+      "loss": 1.1366,
+      "step": 7034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000154135547325768,
+      "loss": 1.1145,
+      "step": 7035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015412244822748808,
+      "loss": 1.1344,
+      "step": 7036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015410934781568528,
+      "loss": 1.1014,
+      "step": 7037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001540962460906775,
+      "loss": 1.198,
+      "step": 7038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001540831430527827,
+      "loss": 1.3295,
+      "step": 7039
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015407003870231893,
+      "loss": 1.1055,
+      "step": 7040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015405693303960417,
+      "loss": 1.0901,
+      "step": 7041
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015404382606495657,
+      "loss": 1.2758,
+      "step": 7042
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015403071777869413,
+      "loss": 1.1417,
+      "step": 7043
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015401760818113504,
+      "loss": 1.0163,
+      "step": 7044
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015400449727259746,
+      "loss": 1.133,
+      "step": 7045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015399138505339956,
+      "loss": 1.0357,
+      "step": 7046
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015397827152385958,
+      "loss": 1.0174,
+      "step": 7047
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015396515668429582,
+      "loss": 1.1353,
+      "step": 7048
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015395204053502649,
+      "loss": 1.2059,
+      "step": 7049
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015393892307636995,
+      "loss": 1.3131,
+      "step": 7050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015392580430864458,
+      "loss": 1.0804,
+      "step": 7051
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015391268423216873,
+      "loss": 1.1486,
+      "step": 7052
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015389956284726084,
+      "loss": 2.5967,
+      "step": 7053
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001538864401542393,
+      "loss": 1.2227,
+      "step": 7054
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015387331615342268,
+      "loss": 2.6169,
+      "step": 7055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015386019084512946,
+      "loss": 1.1351,
+      "step": 7056
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015384706422967814,
+      "loss": 1.3327,
+      "step": 7057
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001538339363073874,
+      "loss": 1.1421,
+      "step": 7058
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015382080707857572,
+      "loss": 1.3593,
+      "step": 7059
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001538076765435618,
+      "loss": 2.3093,
+      "step": 7060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015379454470266433,
+      "loss": 0.9653,
+      "step": 7061
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015378141155620196,
+      "loss": 1.3262,
+      "step": 7062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015376827710449347,
+      "loss": 1.153,
+      "step": 7063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015375514134785762,
+      "loss": 1.1389,
+      "step": 7064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001537420042866132,
+      "loss": 1.1106,
+      "step": 7065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000153728865921079,
+      "loss": 1.2691,
+      "step": 7066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000153715726251574,
+      "loss": 1.0442,
+      "step": 7067
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015370258527841696,
+      "loss": 1.0511,
+      "step": 7068
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001536894430019269,
+      "loss": 1.091,
+      "step": 7069
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001536762994224227,
+      "loss": 1.1481,
+      "step": 7070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015366315454022335,
+      "loss": 1.1811,
+      "step": 7071
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015365000835564798,
+      "loss": 1.1915,
+      "step": 7072
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015363686086901552,
+      "loss": 1.2451,
+      "step": 7073
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015362371208064508,
+      "loss": 1.29,
+      "step": 7074
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001536105619908558,
+      "loss": 1.1141,
+      "step": 7075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015359741059996681,
+      "loss": 1.2669,
+      "step": 7076
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015358425790829732,
+      "loss": 1.1701,
+      "step": 7077
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015357110391616647,
+      "loss": 1.3171,
+      "step": 7078
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015355794862389357,
+      "loss": 1.1405,
+      "step": 7079
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015354479203179785,
+      "loss": 1.1388,
+      "step": 7080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015353163414019862,
+      "loss": 1.1971,
+      "step": 7081
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015351847494941522,
+      "loss": 1.0411,
+      "step": 7082
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015350531445976702,
+      "loss": 2.6746,
+      "step": 7083
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015349215267157344,
+      "loss": 1.2221,
+      "step": 7084
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001534789895851539,
+      "loss": 1.2105,
+      "step": 7085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015346582520082776,
+      "loss": 0.9764,
+      "step": 7086
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015345265951891465,
+      "loss": 1.2249,
+      "step": 7087
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015343949253973407,
+      "loss": 1.0938,
+      "step": 7088
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015342632426360551,
+      "loss": 1.3088,
+      "step": 7089
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001534131546908486,
+      "loss": 1.0881,
+      "step": 7090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015339998382178296,
+      "loss": 1.0847,
+      "step": 7091
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015338681165672825,
+      "loss": 1.199,
+      "step": 7092
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015337363819600412,
+      "loss": 1.094,
+      "step": 7093
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001533604634399303,
+      "loss": 1.2445,
+      "step": 7094
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015334728738882654,
+      "loss": 1.2546,
+      "step": 7095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001533341100430126,
+      "loss": 1.2313,
+      "step": 7096
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001533209314028083,
+      "loss": 1.0361,
+      "step": 7097
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015330775146853344,
+      "loss": 1.2453,
+      "step": 7098
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015329457024050797,
+      "loss": 0.9486,
+      "step": 7099
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015328138771905173,
+      "loss": 1.3,
+      "step": 7100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015326820390448463,
+      "loss": 1.2723,
+      "step": 7101
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001532550187971267,
+      "loss": 1.1742,
+      "step": 7102
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015324183239729796,
+      "loss": 1.1688,
+      "step": 7103
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015322864470531834,
+      "loss": 1.1017,
+      "step": 7104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015321545572150792,
+      "loss": 1.1732,
+      "step": 7105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015320226544618683,
+      "loss": 1.1174,
+      "step": 7106
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001531890738796752,
+      "loss": 1.1325,
+      "step": 7107
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015317588102229312,
+      "loss": 1.1083,
+      "step": 7108
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001531626868743608,
+      "loss": 1.2261,
+      "step": 7109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015314949143619853,
+      "loss": 1.1648,
+      "step": 7110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015313629470812642,
+      "loss": 1.1877,
+      "step": 7111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015312309669046484,
+      "loss": 1.3284,
+      "step": 7112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001531098973835341,
+      "loss": 0.9098,
+      "step": 7113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015309669678765448,
+      "loss": 1.1222,
+      "step": 7114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015308349490314642,
+      "loss": 2.8755,
+      "step": 7115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015307029173033024,
+      "loss": 1.17,
+      "step": 7116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001530570872695265,
+      "loss": 1.2368,
+      "step": 7117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015304388152105556,
+      "loss": 1.2518,
+      "step": 7118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015303067448523794,
+      "loss": 1.05,
+      "step": 7119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015301746616239414,
+      "loss": 1.0617,
+      "step": 7120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001530042565528448,
+      "loss": 1.1547,
+      "step": 7121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015299104565691043,
+      "loss": 1.1784,
+      "step": 7122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015297783347491172,
+      "loss": 1.364,
+      "step": 7123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015296462000716926,
+      "loss": 1.1894,
+      "step": 7124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015295140525400377,
+      "loss": 1.1311,
+      "step": 7125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015293818921573597,
+      "loss": 1.3394,
+      "step": 7126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015292497189268658,
+      "loss": 1.1675,
+      "step": 7127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015291175328517638,
+      "loss": 1.3377,
+      "step": 7128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001528985333935262,
+      "loss": 1.3599,
+      "step": 7129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015288531221805687,
+      "loss": 1.2373,
+      "step": 7130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015287208975908926,
+      "loss": 1.2178,
+      "step": 7131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001528588660169443,
+      "loss": 1.2267,
+      "step": 7132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001528456409919428,
+      "loss": 1.1382,
+      "step": 7133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001528324146844059,
+      "loss": 1.1056,
+      "step": 7134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015281918709465447,
+      "loss": 1.2496,
+      "step": 7135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001528059582230096,
+      "loss": 1.2523,
+      "step": 7136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015279272806979232,
+      "loss": 1.2131,
+      "step": 7137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015277949663532374,
+      "loss": 1.1498,
+      "step": 7138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015276626391992496,
+      "loss": 1.1858,
+      "step": 7139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015275302992391715,
+      "loss": 1.1731,
+      "step": 7140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015273979464762146,
+      "loss": 1.0798,
+      "step": 7141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015272655809135915,
+      "loss": 1.1667,
+      "step": 7142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015271332025545143,
+      "loss": 1.2409,
+      "step": 7143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015270008114021954,
+      "loss": 1.2901,
+      "step": 7144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015268684074598484,
+      "loss": 1.2458,
+      "step": 7145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001526735990730687,
+      "loss": 1.2446,
+      "step": 7146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015266035612179242,
+      "loss": 1.2155,
+      "step": 7147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001526471118924774,
+      "loss": 1.1443,
+      "step": 7148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015263386638544513,
+      "loss": 1.1615,
+      "step": 7149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015262061960101702,
+      "loss": 1.2986,
+      "step": 7150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015260737153951457,
+      "loss": 1.0792,
+      "step": 7151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001525941222012593,
+      "loss": 1.0914,
+      "step": 7152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001525808715865728,
+      "loss": 1.2035,
+      "step": 7153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015256761969577665,
+      "loss": 1.2484,
+      "step": 7154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015255436652919236,
+      "loss": 1.0018,
+      "step": 7155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015254111208714175,
+      "loss": 1.1823,
+      "step": 7156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015252785636994638,
+      "loss": 1.2343,
+      "step": 7157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015251459937792798,
+      "loss": 1.1767,
+      "step": 7158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015250134111140833,
+      "loss": 1.2036,
+      "step": 7159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015248808157070914,
+      "loss": 1.1355,
+      "step": 7160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001524748207561523,
+      "loss": 1.0681,
+      "step": 7161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015246155866805954,
+      "loss": 1.238,
+      "step": 7162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015244829530675275,
+      "loss": 1.1181,
+      "step": 7163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015243503067255388,
+      "loss": 1.2471,
+      "step": 7164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015242176476578485,
+      "loss": 1.2916,
+      "step": 7165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015240849758676753,
+      "loss": 1.0652,
+      "step": 7166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000152395229135824,
+      "loss": 1.277,
+      "step": 7167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015238195941327622,
+      "loss": 1.262,
+      "step": 7168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001523686884194463,
+      "loss": 1.2834,
+      "step": 7169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015235541615465627,
+      "loss": 1.3674,
+      "step": 7170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015234214261922822,
+      "loss": 1.1295,
+      "step": 7171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015232886781348434,
+      "loss": 1.1964,
+      "step": 7172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015231559173774678,
+      "loss": 1.0858,
+      "step": 7173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015230231439233779,
+      "loss": 1.267,
+      "step": 7174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015228903577757953,
+      "loss": 1.2091,
+      "step": 7175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015227575589379434,
+      "loss": 1.131,
+      "step": 7176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015226247474130441,
+      "loss": 2.3992,
+      "step": 7177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001522491923204322,
+      "loss": 1.1809,
+      "step": 7178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015223590863149995,
+      "loss": 1.2792,
+      "step": 7179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001522226236748301,
+      "loss": 1.0935,
+      "step": 7180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015220933745074507,
+      "loss": 0.8077,
+      "step": 7181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001521960499595673,
+      "loss": 1.0852,
+      "step": 7182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015218276120161932,
+      "loss": 1.1641,
+      "step": 7183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015216947117722355,
+      "loss": 1.1101,
+      "step": 7184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015215617988670262,
+      "loss": 1.1174,
+      "step": 7185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015214288733037904,
+      "loss": 1.1027,
+      "step": 7186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001521295935085754,
+      "loss": 0.9684,
+      "step": 7187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015211629842161442,
+      "loss": 1.2136,
+      "step": 7188
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015210300206981866,
+      "loss": 1.1222,
+      "step": 7189
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001520897044535109,
+      "loss": 2.5088,
+      "step": 7190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015207640557301385,
+      "loss": 1.2077,
+      "step": 7191
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015206310542865024,
+      "loss": 1.2651,
+      "step": 7192
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015204980402074285,
+      "loss": 1.106,
+      "step": 7193
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015203650134961453,
+      "loss": 1.1663,
+      "step": 7194
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001520231974155881,
+      "loss": 1.2814,
+      "step": 7195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001520098922189865,
+      "loss": 1.2837,
+      "step": 7196
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015199658576013254,
+      "loss": 1.1012,
+      "step": 7197
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015198327803934923,
+      "loss": 2.6101,
+      "step": 7198
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015196996905695955,
+      "loss": 1.4079,
+      "step": 7199
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015195665881328644,
+      "loss": 1.1349,
+      "step": 7200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000151943347308653,
+      "loss": 1.084,
+      "step": 7201
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015193003454338228,
+      "loss": 1.0509,
+      "step": 7202
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015191672051779732,
+      "loss": 1.2867,
+      "step": 7203
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001519034052322213,
+      "loss": 1.0422,
+      "step": 7204
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015189008868697737,
+      "loss": 1.2411,
+      "step": 7205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015187677088238867,
+      "loss": 1.1764,
+      "step": 7206
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015186345181877848,
+      "loss": 1.2467,
+      "step": 7207
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015185013149647,
+      "loss": 1.0133,
+      "step": 7208
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001518368099157865,
+      "loss": 1.2581,
+      "step": 7209
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015182348707705134,
+      "loss": 1.2458,
+      "step": 7210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015181016298058782,
+      "loss": 1.1805,
+      "step": 7211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015179683762671934,
+      "loss": 2.4043,
+      "step": 7212
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015178351101576923,
+      "loss": 1.062,
+      "step": 7213
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000151770183148061,
+      "loss": 1.0918,
+      "step": 7214
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015175685402391806,
+      "loss": 1.2345,
+      "step": 7215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001517435236436639,
+      "loss": 1.291,
+      "step": 7216
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015173019200762208,
+      "loss": 1.1574,
+      "step": 7217
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015171685911611612,
+      "loss": 1.2983,
+      "step": 7218
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001517035249694696,
+      "loss": 1.1942,
+      "step": 7219
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015169018956800618,
+      "loss": 1.1985,
+      "step": 7220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015167685291204945,
+      "loss": 1.2612,
+      "step": 7221
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001516635150019231,
+      "loss": 1.0617,
+      "step": 7222
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00015165017583795083,
+      "loss": 1.1298,
+      "step": 7223
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001516368354204564,
+      "loss": 1.1647,
+      "step": 7224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001516234937497635,
+      "loss": 1.2666,
+      "step": 7225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015161015082619606,
+      "loss": 1.1169,
+      "step": 7226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015159680665007775,
+      "loss": 1.1399,
+      "step": 7227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015158346122173252,
+      "loss": 1.0364,
+      "step": 7228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015157011454148428,
+      "loss": 1.2885,
+      "step": 7229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015155676660965684,
+      "loss": 1.154,
+      "step": 7230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015154341742657426,
+      "loss": 1.2357,
+      "step": 7231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015153006699256046,
+      "loss": 2.4343,
+      "step": 7232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015151671530793943,
+      "loss": 1.1176,
+      "step": 7233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015150336237303527,
+      "loss": 0.9979,
+      "step": 7234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000151490008188172,
+      "loss": 1.1243,
+      "step": 7235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015147665275367373,
+      "loss": 1.1167,
+      "step": 7236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015146329606986457,
+      "loss": 1.174,
+      "step": 7237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015144993813706874,
+      "loss": 1.0011,
+      "step": 7238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015143657895561033,
+      "loss": 1.0825,
+      "step": 7239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015142321852581369,
+      "loss": 1.1902,
+      "step": 7240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015140985684800298,
+      "loss": 1.092,
+      "step": 7241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001513964939225025,
+      "loss": 0.9991,
+      "step": 7242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015138312974963654,
+      "loss": 1.2294,
+      "step": 7243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015136976432972952,
+      "loss": 1.2332,
+      "step": 7244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001513563976631057,
+      "loss": 2.5789,
+      "step": 7245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015134302975008956,
+      "loss": 1.3287,
+      "step": 7246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001513296605910055,
+      "loss": 1.1738,
+      "step": 7247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015131629018617804,
+      "loss": 2.3428,
+      "step": 7248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015130291853593158,
+      "loss": 1.0496,
+      "step": 7249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001512895456405907,
+      "loss": 1.2109,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015127617150047993,
+      "loss": 1.3516,
+      "step": 7251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015126279611592388,
+      "loss": 1.2402,
+      "step": 7252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015124941948724716,
+      "loss": 1.0545,
+      "step": 7253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015123604161477436,
+      "loss": 1.2364,
+      "step": 7254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015122266249883022,
+      "loss": 1.173,
+      "step": 7255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001512092821397394,
+      "loss": 0.9222,
+      "step": 7256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001511959005378267,
+      "loss": 1.2184,
+      "step": 7257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001511825176934168,
+      "loss": 1.0957,
+      "step": 7258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015116913360683457,
+      "loss": 0.9389,
+      "step": 7259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015115574827840475,
+      "loss": 1.2687,
+      "step": 7260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015114236170845228,
+      "loss": 1.1713,
+      "step": 7261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015112897389730196,
+      "loss": 1.1761,
+      "step": 7262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015111558484527877,
+      "loss": 1.2524,
+      "step": 7263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015110219455270768,
+      "loss": 1.1942,
+      "step": 7264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015108880301991357,
+      "loss": 2.4167,
+      "step": 7265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015107541024722153,
+      "loss": 1.2189,
+      "step": 7266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015106201623495656,
+      "loss": 1.062,
+      "step": 7267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001510486209834437,
+      "loss": 1.0544,
+      "step": 7268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015103522449300808,
+      "loss": 1.2684,
+      "step": 7269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015102182676397482,
+      "loss": 0.9787,
+      "step": 7270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001510084277966691,
+      "loss": 1.1918,
+      "step": 7271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001509950275914161,
+      "loss": 1.1461,
+      "step": 7272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015098162614854096,
+      "loss": 1.1454,
+      "step": 7273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000150968223468369,
+      "loss": 1.1827,
+      "step": 7274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015095481955122551,
+      "loss": 1.0266,
+      "step": 7275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015094141439743572,
+      "loss": 1.3728,
+      "step": 7276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015092800800732504,
+      "loss": 1.2281,
+      "step": 7277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001509146003812188,
+      "loss": 1.2364,
+      "step": 7278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001509011915194424,
+      "loss": 1.1509,
+      "step": 7279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001508877814223213,
+      "loss": 2.3271,
+      "step": 7280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015087437009018086,
+      "loss": 1.0988,
+      "step": 7281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015086095752334667,
+      "loss": 1.1594,
+      "step": 7282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001508475437221442,
+      "loss": 1.2233,
+      "step": 7283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000150834128686899,
+      "loss": 1.1891,
+      "step": 7284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015082071241793665,
+      "loss": 0.9767,
+      "step": 7285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015080729491558277,
+      "loss": 1.1656,
+      "step": 7286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015079387618016293,
+      "loss": 0.8925,
+      "step": 7287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001507804562120029,
+      "loss": 1.0327,
+      "step": 7288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015076703501142827,
+      "loss": 2.251,
+      "step": 7289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015075361257876484,
+      "loss": 1.2881,
+      "step": 7290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015074018891433837,
+      "loss": 1.1744,
+      "step": 7291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015072676401847458,
+      "loss": 1.1323,
+      "step": 7292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015071333789149928,
+      "loss": 1.2702,
+      "step": 7293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015069991053373846,
+      "loss": 1.1398,
+      "step": 7294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001506864819455178,
+      "loss": 1.1655,
+      "step": 7295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015067305212716333,
+      "loss": 1.1077,
+      "step": 7296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015065962107900095,
+      "loss": 1.1302,
+      "step": 7297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015064618880135662,
+      "loss": 1.2265,
+      "step": 7298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015063275529455633,
+      "loss": 1.1469,
+      "step": 7299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015061932055892614,
+      "loss": 1.2277,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015060588459479206,
+      "loss": 2.3582,
+      "step": 7301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001505924474024802,
+      "loss": 1.2102,
+      "step": 7302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015057900898231665,
+      "loss": 1.1191,
+      "step": 7303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015056556933462756,
+      "loss": 1.0916,
+      "step": 7304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001505521284597391,
+      "loss": 1.2378,
+      "step": 7305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001505386863579775,
+      "loss": 1.1708,
+      "step": 7306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000150525243029669,
+      "loss": 1.2204,
+      "step": 7307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015051179847513978,
+      "loss": 1.2317,
+      "step": 7308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015049835269471623,
+      "loss": 1.1489,
+      "step": 7309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015048490568872464,
+      "loss": 1.2361,
+      "step": 7310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015047145745749135,
+      "loss": 1.1485,
+      "step": 7311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001504580080013427,
+      "loss": 1.2281,
+      "step": 7312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015044455732060518,
+      "loss": 1.1994,
+      "step": 7313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001504311054156052,
+      "loss": 0.9909,
+      "step": 7314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015041765228666922,
+      "loss": 2.4626,
+      "step": 7315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015040419793412374,
+      "loss": 1.1387,
+      "step": 7316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001503907423582953,
+      "loss": 1.1407,
+      "step": 7317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015037728555951046,
+      "loss": 1.0465,
+      "step": 7318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001503638275380958,
+      "loss": 1.2509,
+      "step": 7319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015035036829437795,
+      "loss": 1.0936,
+      "step": 7320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015033690782868358,
+      "loss": 1.1615,
+      "step": 7321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015032344614133932,
+      "loss": 1.0817,
+      "step": 7322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001503099832326719,
+      "loss": 1.1856,
+      "step": 7323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015029651910300807,
+      "loss": 1.1196,
+      "step": 7324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015028305375267456,
+      "loss": 1.1184,
+      "step": 7325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015026958718199822,
+      "loss": 1.2787,
+      "step": 7326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015025611939130582,
+      "loss": 1.1358,
+      "step": 7327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00015024265038092422,
+      "loss": 1.1072,
+      "step": 7328
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001502291801511804,
+      "loss": 1.0846,
+      "step": 7329
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015021570870240116,
+      "loss": 1.3145,
+      "step": 7330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001502022360349135,
+      "loss": 1.269,
+      "step": 7331
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015018876214904442,
+      "loss": 1.263,
+      "step": 7332
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015017528704512083,
+      "loss": 1.1915,
+      "step": 7333
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015016181072346986,
+      "loss": 1.2838,
+      "step": 7334
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001501483331844185,
+      "loss": 1.1648,
+      "step": 7335
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001501348544282939,
+      "loss": 0.931,
+      "step": 7336
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015012137445542315,
+      "loss": 1.2068,
+      "step": 7337
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015010789326613347,
+      "loss": 1.0997,
+      "step": 7338
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015009441086075192,
+      "loss": 1.0452,
+      "step": 7339
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001500809272396058,
+      "loss": 1.1571,
+      "step": 7340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001500674424030223,
+      "loss": 1.258,
+      "step": 7341
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015005395635132873,
+      "loss": 1.1931,
+      "step": 7342
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015004046908485236,
+      "loss": 1.2344,
+      "step": 7343
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015002698060392056,
+      "loss": 1.2159,
+      "step": 7344
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015001349090886066,
+      "loss": 1.1537,
+      "step": 7345
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 1.2849,
+      "step": 7346
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001499865078776661,
+      "loss": 1.2371,
+      "step": 7347
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014997301454218633,
+      "loss": 1.1987,
+      "step": 7348
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014995951999388816,
+      "loss": 1.2876,
+      "step": 7349
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014994602423309917,
+      "loss": 1.1635,
+      "step": 7350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014993252726014684,
+      "loss": 1.071,
+      "step": 7351
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001499190290753587,
+      "loss": 1.222,
+      "step": 7352
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014990552967906247,
+      "loss": 1.1448,
+      "step": 7353
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014989202907158563,
+      "loss": 1.1408,
+      "step": 7354
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014987852725325586,
+      "loss": 1.144,
+      "step": 7355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014986502422440094,
+      "loss": 1.1971,
+      "step": 7356
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014985151998534848,
+      "loss": 1.2112,
+      "step": 7357
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014983801453642627,
+      "loss": 1.3739,
+      "step": 7358
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014982450787796205,
+      "loss": 1.3072,
+      "step": 7359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014981100001028365,
+      "loss": 1.2362,
+      "step": 7360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014979749093371888,
+      "loss": 2.3296,
+      "step": 7361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001497839806485956,
+      "loss": 1.0931,
+      "step": 7362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001497704691552417,
+      "loss": 1.1997,
+      "step": 7363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001497569564539851,
+      "loss": 1.2953,
+      "step": 7364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014974344254515374,
+      "loss": 2.3887,
+      "step": 7365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001497299274290756,
+      "loss": 1.105,
+      "step": 7366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001497164111060787,
+      "loss": 1.1681,
+      "step": 7367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014970289357649102,
+      "loss": 1.1122,
+      "step": 7368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014968937484064067,
+      "loss": 1.2364,
+      "step": 7369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014967585489885575,
+      "loss": 1.1953,
+      "step": 7370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014966233375146435,
+      "loss": 1.1527,
+      "step": 7371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014964881139879462,
+      "loss": 1.1069,
+      "step": 7372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014963528784117476,
+      "loss": 1.1276,
+      "step": 7373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014962176307893298,
+      "loss": 1.1833,
+      "step": 7374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014960823711239756,
+      "loss": 0.9809,
+      "step": 7375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014959470994189666,
+      "loss": 1.2246,
+      "step": 7376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014958118156775866,
+      "loss": 1.1072,
+      "step": 7377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001495676519903119,
+      "loss": 2.3857,
+      "step": 7378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014955412120988464,
+      "loss": 1.2216,
+      "step": 7379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014954058922680535,
+      "loss": 1.1657,
+      "step": 7380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014952705604140241,
+      "loss": 1.1215,
+      "step": 7381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014951352165400427,
+      "loss": 1.0669,
+      "step": 7382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014949998606493946,
+      "loss": 2.27,
+      "step": 7383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014948644927453635,
+      "loss": 1.1874,
+      "step": 7384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014947291128312358,
+      "loss": 1.3481,
+      "step": 7385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014945937209102968,
+      "loss": 1.1533,
+      "step": 7386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014944583169858322,
+      "loss": 1.145,
+      "step": 7387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014943229010611287,
+      "loss": 1.0202,
+      "step": 7388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014941874731394723,
+      "loss": 1.0046,
+      "step": 7389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014940520332241495,
+      "loss": 2.3933,
+      "step": 7390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001493916581318448,
+      "loss": 1.1349,
+      "step": 7391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014937811174256548,
+      "loss": 1.1112,
+      "step": 7392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014936456415490576,
+      "loss": 1.1291,
+      "step": 7393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014935101536919446,
+      "loss": 1.1492,
+      "step": 7394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014933746538576035,
+      "loss": 1.2679,
+      "step": 7395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001493239142049323,
+      "loss": 1.0843,
+      "step": 7396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014931036182703924,
+      "loss": 1.2175,
+      "step": 7397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014929680825241,
+      "loss": 1.271,
+      "step": 7398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014928325348137356,
+      "loss": 1.171,
+      "step": 7399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014926969751425886,
+      "loss": 1.0634,
+      "step": 7400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014925614035139493,
+      "loss": 1.0636,
+      "step": 7401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014924258199311077,
+      "loss": 1.0911,
+      "step": 7402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014922902243973547,
+      "loss": 1.2391,
+      "step": 7403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014921546169159808,
+      "loss": 1.1415,
+      "step": 7404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014920189974902773,
+      "loss": 1.3591,
+      "step": 7405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014918833661235355,
+      "loss": 2.2617,
+      "step": 7406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001491747722819047,
+      "loss": 1.1569,
+      "step": 7407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001491612067580104,
+      "loss": 1.2014,
+      "step": 7408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014914764004099985,
+      "loss": 1.1672,
+      "step": 7409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014913407213120234,
+      "loss": 1.1349,
+      "step": 7410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014912050302894715,
+      "loss": 1.1033,
+      "step": 7411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014910693273456358,
+      "loss": 1.1979,
+      "step": 7412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000149093361248381,
+      "loss": 1.2474,
+      "step": 7413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014907978857072872,
+      "loss": 1.1405,
+      "step": 7414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014906621470193621,
+      "loss": 1.0906,
+      "step": 7415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014905263964233288,
+      "loss": 1.1007,
+      "step": 7416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001490390633922482,
+      "loss": 1.3094,
+      "step": 7417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014902548595201163,
+      "loss": 1.3047,
+      "step": 7418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001490119073219527,
+      "loss": 0.9247,
+      "step": 7419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014899832750240092,
+      "loss": 1.0895,
+      "step": 7420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014898474649368594,
+      "loss": 1.1175,
+      "step": 7421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014897116429613734,
+      "loss": 1.1975,
+      "step": 7422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014895758091008473,
+      "loss": 1.1788,
+      "step": 7423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014894399633585774,
+      "loss": 1.421,
+      "step": 7424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014893041057378612,
+      "loss": 1.2541,
+      "step": 7425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001489168236241996,
+      "loss": 1.1945,
+      "step": 7426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014890323548742785,
+      "loss": 1.1908,
+      "step": 7427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014888964616380072,
+      "loss": 1.2859,
+      "step": 7428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000148876055653648,
+      "loss": 1.2507,
+      "step": 7429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001488624639572995,
+      "loss": 1.0913,
+      "step": 7430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001488488710750851,
+      "loss": 1.0876,
+      "step": 7431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00014883527700733468,
+      "loss": 1.0349,
+      "step": 7432
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001488216817543782,
+      "loss": 1.2795,
+      "step": 7433
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014880808531654552,
+      "loss": 1.2784,
+      "step": 7434
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014879448769416676,
+      "loss": 1.2368,
+      "step": 7435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014878088888757178,
+      "loss": 1.1778,
+      "step": 7436
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014876728889709072,
+      "loss": 2.4922,
+      "step": 7437
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014875368772305358,
+      "loss": 1.2709,
+      "step": 7438
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014874008536579046,
+      "loss": 1.1052,
+      "step": 7439
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014872648182563154,
+      "loss": 1.1256,
+      "step": 7440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014871287710290694,
+      "loss": 1.2333,
+      "step": 7441
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001486992711979468,
+      "loss": 1.058,
+      "step": 7442
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014868566411108138,
+      "loss": 1.1653,
+      "step": 7443
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014867205584264084,
+      "loss": 1.2122,
+      "step": 7444
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014865844639295558,
+      "loss": 2.4766,
+      "step": 7445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014864483576235577,
+      "loss": 1.0352,
+      "step": 7446
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014863122395117177,
+      "loss": 1.204,
+      "step": 7447
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014861761095973398,
+      "loss": 1.2167,
+      "step": 7448
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001486039967883727,
+      "loss": 1.2045,
+      "step": 7449
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001485903814374184,
+      "loss": 1.0111,
+      "step": 7450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001485767649072015,
+      "loss": 0.8923,
+      "step": 7451
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014856314719805243,
+      "loss": 1.3055,
+      "step": 7452
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014854952831030175,
+      "loss": 1.2519,
+      "step": 7453
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001485359082442799,
+      "loss": 1.0811,
+      "step": 7454
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001485222870003175,
+      "loss": 1.2039,
+      "step": 7455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014850866457874514,
+      "loss": 1.2297,
+      "step": 7456
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001484950409798934,
+      "loss": 1.2198,
+      "step": 7457
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001484814162040929,
+      "loss": 1.0521,
+      "step": 7458
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001484677902516743,
+      "loss": 1.1014,
+      "step": 7459
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014845416312296832,
+      "loss": 1.1962,
+      "step": 7460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001484405348183057,
+      "loss": 2.4473,
+      "step": 7461
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014842690533801718,
+      "loss": 1.2133,
+      "step": 7462
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014841327468243353,
+      "loss": 1.1489,
+      "step": 7463
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014839964285188556,
+      "loss": 1.2111,
+      "step": 7464
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001483860098467041,
+      "loss": 1.0642,
+      "step": 7465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014837237566722,
+      "loss": 1.0319,
+      "step": 7466
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014835874031376422,
+      "loss": 1.1219,
+      "step": 7467
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001483451037866676,
+      "loss": 1.2224,
+      "step": 7468
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001483314660862612,
+      "loss": 1.1364,
+      "step": 7469
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014831782721287589,
+      "loss": 1.2112,
+      "step": 7470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014830418716684272,
+      "loss": 1.3372,
+      "step": 7471
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014829054594849272,
+      "loss": 1.0813,
+      "step": 7472
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014827690355815698,
+      "loss": 1.2368,
+      "step": 7473
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014826325999616654,
+      "loss": 1.0961,
+      "step": 7474
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014824961526285258,
+      "loss": 1.2038,
+      "step": 7475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014823596935854624,
+      "loss": 1.1133,
+      "step": 7476
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001482223222835787,
+      "loss": 2.3325,
+      "step": 7477
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014820867403828112,
+      "loss": 1.1802,
+      "step": 7478
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014819502462298478,
+      "loss": 1.0932,
+      "step": 7479
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014818137403802096,
+      "loss": 2.3806,
+      "step": 7480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014816772228372086,
+      "loss": 1.1417,
+      "step": 7481
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001481540693604159,
+      "loss": 1.2524,
+      "step": 7482
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014814041526843742,
+      "loss": 1.3379,
+      "step": 7483
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014812676000811677,
+      "loss": 1.1916,
+      "step": 7484
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014811310357978533,
+      "loss": 1.015,
+      "step": 7485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001480994459837746,
+      "loss": 1.2255,
+      "step": 7486
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014808578722041597,
+      "loss": 1.2111,
+      "step": 7487
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014807212729004102,
+      "loss": 1.1709,
+      "step": 7488
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014805846619298115,
+      "loss": 1.186,
+      "step": 7489
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000148044803929568,
+      "loss": 1.1456,
+      "step": 7490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001480311405001331,
+      "loss": 1.188,
+      "step": 7491
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014801747590500815,
+      "loss": 1.0533,
+      "step": 7492
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001480038101445246,
+      "loss": 2.5403,
+      "step": 7493
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001479901432190143,
+      "loss": 1.1969,
+      "step": 7494
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014797647512880882,
+      "loss": 2.6494,
+      "step": 7495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001479628058742399,
+      "loss": 1.1929,
+      "step": 7496
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014794913545563933,
+      "loss": 1.1182,
+      "step": 7497
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014793546387333883,
+      "loss": 1.2439,
+      "step": 7498
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014792179112767023,
+      "loss": 1.1033,
+      "step": 7499
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001479081172189654,
+      "loss": 1.2315,
+      "step": 7500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014789444214755607,
+      "loss": 1.0403,
+      "step": 7501
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014788076591377427,
+      "loss": 1.0451,
+      "step": 7502
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014786708851795182,
+      "loss": 1.1576,
+      "step": 7503
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001478534099604207,
+      "loss": 1.1214,
+      "step": 7504
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001478397302415129,
+      "loss": 1.1888,
+      "step": 7505
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001478260493615604,
+      "loss": 1.346,
+      "step": 7506
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014781236732089523,
+      "loss": 1.1608,
+      "step": 7507
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014779868411984943,
+      "loss": 1.2379,
+      "step": 7508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001477849997587551,
+      "loss": 2.5475,
+      "step": 7509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014777131423794436,
+      "loss": 1.17,
+      "step": 7510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001477576275577493,
+      "loss": 1.4045,
+      "step": 7511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014774393971850214,
+      "loss": 1.2164,
+      "step": 7512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014773025072053507,
+      "loss": 1.2781,
+      "step": 7513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001477165605641803,
+      "loss": 1.1442,
+      "step": 7514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001477028692497701,
+      "loss": 1.1981,
+      "step": 7515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014768917677763677,
+      "loss": 1.1256,
+      "step": 7516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014767548314811255,
+      "loss": 1.0909,
+      "step": 7517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001476617883615298,
+      "loss": 1.2046,
+      "step": 7518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014764809241822095,
+      "loss": 1.2145,
+      "step": 7519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014763439531851832,
+      "loss": 1.0893,
+      "step": 7520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014762069706275436,
+      "loss": 1.2945,
+      "step": 7521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014760699765126157,
+      "loss": 1.2157,
+      "step": 7522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014759329708437227,
+      "loss": 2.4028,
+      "step": 7523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014757959536241917,
+      "loss": 1.2142,
+      "step": 7524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014756589248573465,
+      "loss": 1.1542,
+      "step": 7525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014755218845465134,
+      "loss": 1.2949,
+      "step": 7526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001475384832695018,
+      "loss": 0.9829,
+      "step": 7527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014752477693061864,
+      "loss": 1.1105,
+      "step": 7528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001475110694383346,
+      "loss": 1.2567,
+      "step": 7529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014749736079298223,
+      "loss": 1.2249,
+      "step": 7530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001474836509948943,
+      "loss": 1.2902,
+      "step": 7531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001474699400444035,
+      "loss": 2.3477,
+      "step": 7532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014745622794184259,
+      "loss": 1.3102,
+      "step": 7533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014744251468754442,
+      "loss": 2.2483,
+      "step": 7534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014742880028184173,
+      "loss": 1.1585,
+      "step": 7535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00014741508472506743,
+      "loss": 1.0829,
+      "step": 7536
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001474013680175543,
+      "loss": 1.2419,
+      "step": 7537
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014738765015963532,
+      "loss": 1.2381,
+      "step": 7538
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014737393115164335,
+      "loss": 1.2571,
+      "step": 7539
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001473602109939114,
+      "loss": 1.0001,
+      "step": 7540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001473464896867724,
+      "loss": 1.137,
+      "step": 7541
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001473327672305594,
+      "loss": 1.1762,
+      "step": 7542
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014731904362560544,
+      "loss": 1.1389,
+      "step": 7543
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014730531887224352,
+      "loss": 1.1949,
+      "step": 7544
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001472915929708068,
+      "loss": 1.2726,
+      "step": 7545
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014727786592162837,
+      "loss": 1.1646,
+      "step": 7546
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001472641377250414,
+      "loss": 1.082,
+      "step": 7547
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014725040838137903,
+      "loss": 0.9974,
+      "step": 7548
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001472366778909745,
+      "loss": 2.2175,
+      "step": 7549
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014722294625416104,
+      "loss": 1.1473,
+      "step": 7550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001472092134712719,
+      "loss": 1.2405,
+      "step": 7551
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014719547954264033,
+      "loss": 1.0889,
+      "step": 7552
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001471817444685997,
+      "loss": 1.2927,
+      "step": 7553
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001471680082494833,
+      "loss": 1.2162,
+      "step": 7554
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014715427088562457,
+      "loss": 1.2409,
+      "step": 7555
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014714053237735686,
+      "loss": 1.2628,
+      "step": 7556
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001471267927250136,
+      "loss": 1.2144,
+      "step": 7557
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014711305192892827,
+      "loss": 1.1363,
+      "step": 7558
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001470993099894343,
+      "loss": 2.1364,
+      "step": 7559
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014708556690686526,
+      "loss": 1.1437,
+      "step": 7560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014707182268155464,
+      "loss": 1.1451,
+      "step": 7561
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014705807731383604,
+      "loss": 1.295,
+      "step": 7562
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014704433080404302,
+      "loss": 1.1199,
+      "step": 7563
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014703058315250925,
+      "loss": 1.122,
+      "step": 7564
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001470168343595683,
+      "loss": 1.2447,
+      "step": 7565
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014700308442555394,
+      "loss": 1.1552,
+      "step": 7566
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014698933335079983,
+      "loss": 1.1326,
+      "step": 7567
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014697558113563965,
+      "loss": 1.2065,
+      "step": 7568
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014696182778040724,
+      "loss": 1.2768,
+      "step": 7569
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014694807328543633,
+      "loss": 1.1255,
+      "step": 7570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014693431765106077,
+      "loss": 1.1588,
+      "step": 7571
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014692056087761438,
+      "loss": 1.0409,
+      "step": 7572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014690680296543108,
+      "loss": 1.3288,
+      "step": 7573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014689304391484465,
+      "loss": 1.0494,
+      "step": 7574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014687928372618915,
+      "loss": 1.1453,
+      "step": 7575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014686552239979845,
+      "loss": 1.3672,
+      "step": 7576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014685175993600655,
+      "loss": 2.4136,
+      "step": 7577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014683799633514747,
+      "loss": 1.2248,
+      "step": 7578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001468242315975552,
+      "loss": 0.9998,
+      "step": 7579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001468104657235639,
+      "loss": 1.2009,
+      "step": 7580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014679669871350755,
+      "loss": 1.1739,
+      "step": 7581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014678293056772035,
+      "loss": 1.196,
+      "step": 7582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001467691612865364,
+      "loss": 1.1614,
+      "step": 7583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014675539087028987,
+      "loss": 1.0848,
+      "step": 7584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014674161931931497,
+      "loss": 1.1872,
+      "step": 7585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014672784663394595,
+      "loss": 1.2772,
+      "step": 7586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014671407281451705,
+      "loss": 1.1821,
+      "step": 7587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014670029786136257,
+      "loss": 1.0447,
+      "step": 7588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014668652177481678,
+      "loss": 1.1201,
+      "step": 7589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014667274455521404,
+      "loss": 1.0573,
+      "step": 7590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014665896620288875,
+      "loss": 2.4612,
+      "step": 7591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014664518671817525,
+      "loss": 1.1684,
+      "step": 7592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014663140610140794,
+      "loss": 1.2713,
+      "step": 7593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014661762435292137,
+      "loss": 1.2423,
+      "step": 7594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014660384147304996,
+      "loss": 1.2215,
+      "step": 7595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014659005746212815,
+      "loss": 1.0693,
+      "step": 7596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014657627232049058,
+      "loss": 1.174,
+      "step": 7597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014656248604847174,
+      "loss": 1.1948,
+      "step": 7598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014654869864640622,
+      "loss": 1.1583,
+      "step": 7599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014653491011462866,
+      "loss": 1.272,
+      "step": 7600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014652112045347368,
+      "loss": 1.2343,
+      "step": 7601
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.000146507329663276,
+      "loss": 1.2344,
+      "step": 7602
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001464935377443702,
+      "loss": 1.3074,
+      "step": 7603
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014647974469709108,
+      "loss": 0.9549,
+      "step": 7604
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001464659505217734,
+      "loss": 1.0033,
+      "step": 7605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014645215521875193,
+      "loss": 1.3553,
+      "step": 7606
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014643835878836144,
+      "loss": 1.1487,
+      "step": 7607
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014642456123093678,
+      "loss": 1.1525,
+      "step": 7608
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014641076254681283,
+      "loss": 1.0374,
+      "step": 7609
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001463969627363245,
+      "loss": 1.0692,
+      "step": 7610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001463831617998066,
+      "loss": 1.0121,
+      "step": 7611
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014636935973759419,
+      "loss": 0.9958,
+      "step": 7612
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014635555655002216,
+      "loss": 1.1616,
+      "step": 7613
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014634175223742553,
+      "loss": 1.0501,
+      "step": 7614
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014632794680013932,
+      "loss": 1.0627,
+      "step": 7615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014631414023849858,
+      "loss": 1.2247,
+      "step": 7616
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001463003325528384,
+      "loss": 1.1449,
+      "step": 7617
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001462865237434939,
+      "loss": 1.1404,
+      "step": 7618
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014627271381080015,
+      "loss": 1.1721,
+      "step": 7619
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014625890275509238,
+      "loss": 1.146,
+      "step": 7620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014624509057670576,
+      "loss": 1.0952,
+      "step": 7621
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014623127727597546,
+      "loss": 0.9764,
+      "step": 7622
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014621746285323678,
+      "loss": 1.2996,
+      "step": 7623
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014620364730882492,
+      "loss": 1.2319,
+      "step": 7624
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014618983064307526,
+      "loss": 1.1342,
+      "step": 7625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014617601285632308,
+      "loss": 0.9552,
+      "step": 7626
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001461621939489037,
+      "loss": 1.1857,
+      "step": 7627
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014614837392115253,
+      "loss": 1.0114,
+      "step": 7628
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.000146134552773405,
+      "loss": 1.2451,
+      "step": 7629
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014612073050599648,
+      "loss": 1.1307,
+      "step": 7630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014610690711926248,
+      "loss": 1.0691,
+      "step": 7631
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014609308261353849,
+      "loss": 1.1404,
+      "step": 7632
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014607925698915998,
+      "loss": 1.1461,
+      "step": 7633
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001460654302464625,
+      "loss": 1.1254,
+      "step": 7634
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014605160238578168,
+      "loss": 2.2441,
+      "step": 7635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014603777340745303,
+      "loss": 1.1119,
+      "step": 7636
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014602394331181222,
+      "loss": 2.2068,
+      "step": 7637
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00014601011209919488,
+      "loss": 1.088,
+      "step": 7638
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001459962797699367,
+      "loss": 1.1357,
+      "step": 7639
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001459824463243734,
+      "loss": 1.1414,
+      "step": 7640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014596861176284064,
+      "loss": 1.3601,
+      "step": 7641
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014595477608567425,
+      "loss": 1.1321,
+      "step": 7642
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014594093929320997,
+      "loss": 1.3323,
+      "step": 7643
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014592710138578365,
+      "loss": 1.2312,
+      "step": 7644
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014591326236373112,
+      "loss": 1.3007,
+      "step": 7645
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001458994222273882,
+      "loss": 1.1368,
+      "step": 7646
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014588558097709088,
+      "loss": 1.095,
+      "step": 7647
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000145871738613175,
+      "loss": 1.1926,
+      "step": 7648
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014585789513597646,
+      "loss": 1.164,
+      "step": 7649
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014584405054583134,
+      "loss": 1.0339,
+      "step": 7650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014583020484307558,
+      "loss": 1.1921,
+      "step": 7651
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014581635802804525,
+      "loss": 1.2542,
+      "step": 7652
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001458025101010764,
+      "loss": 1.0728,
+      "step": 7653
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014578866106250505,
+      "loss": 1.0028,
+      "step": 7654
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014577481091266738,
+      "loss": 1.2531,
+      "step": 7655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001457609596518995,
+      "loss": 1.1535,
+      "step": 7656
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014574710728053756,
+      "loss": 1.2283,
+      "step": 7657
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014573325379891772,
+      "loss": 1.0027,
+      "step": 7658
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001457193992073763,
+      "loss": 2.449,
+      "step": 7659
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014570554350624943,
+      "loss": 1.3095,
+      "step": 7660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014569168669587346,
+      "loss": 2.2385,
+      "step": 7661
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014567782877658467,
+      "loss": 0.989,
+      "step": 7662
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014566396974871935,
+      "loss": 1.0614,
+      "step": 7663
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001456501096126139,
+      "loss": 1.1086,
+      "step": 7664
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014563624836860464,
+      "loss": 1.2434,
+      "step": 7665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000145622386017028,
+      "loss": 1.2606,
+      "step": 7666
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014560852255822044,
+      "loss": 1.1935,
+      "step": 7667
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014559465799251838,
+      "loss": 1.0656,
+      "step": 7668
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014558079232025837,
+      "loss": 1.045,
+      "step": 7669
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014556692554177684,
+      "loss": 1.1323,
+      "step": 7670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014555305765741037,
+      "loss": 1.1125,
+      "step": 7671
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014553918866749553,
+      "loss": 1.0514,
+      "step": 7672
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014552531857236887,
+      "loss": 1.0378,
+      "step": 7673
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014551144737236707,
+      "loss": 1.1108,
+      "step": 7674
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014549757506782674,
+      "loss": 1.2496,
+      "step": 7675
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014548370165908458,
+      "loss": 1.1105,
+      "step": 7676
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014546982714647725,
+      "loss": 1.2554,
+      "step": 7677
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014545595153034154,
+      "loss": 1.1277,
+      "step": 7678
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001454420748110141,
+      "loss": 1.1303,
+      "step": 7679
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014542819698883184,
+      "loss": 0.9171,
+      "step": 7680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014541431806413147,
+      "loss": 1.2926,
+      "step": 7681
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014540043803724982,
+      "loss": 1.0967,
+      "step": 7682
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014538655690852386,
+      "loss": 0.9595,
+      "step": 7683
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001453726746782904,
+      "loss": 1.296,
+      "step": 7684
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001453587913468863,
+      "loss": 1.1277,
+      "step": 7685
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014534490691464863,
+      "loss": 1.0095,
+      "step": 7686
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001453310213819142,
+      "loss": 1.3923,
+      "step": 7687
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001453171347490202,
+      "loss": 2.48,
+      "step": 7688
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014530324701630344,
+      "loss": 1.1119,
+      "step": 7689
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001452893581841011,
+      "loss": 1.1265,
+      "step": 7690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001452754682527503,
+      "loss": 1.0115,
+      "step": 7691
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000145261577222588,
+      "loss": 1.1945,
+      "step": 7692
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001452476850939514,
+      "loss": 1.1607,
+      "step": 7693
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014523379186717766,
+      "loss": 2.3457,
+      "step": 7694
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014521989754260394,
+      "loss": 1.2142,
+      "step": 7695
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001452060021205675,
+      "loss": 1.0637,
+      "step": 7696
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014519210560140548,
+      "loss": 1.1812,
+      "step": 7697
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014517820798545523,
+      "loss": 1.0895,
+      "step": 7698
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014516430927305395,
+      "loss": 0.9829,
+      "step": 7699
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014515040946453905,
+      "loss": 1.1219,
+      "step": 7700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014513650856024784,
+      "loss": 1.1724,
+      "step": 7701
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014512260656051766,
+      "loss": 1.2161,
+      "step": 7702
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001451087034656859,
+      "loss": 1.17,
+      "step": 7703
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014509479927609,
+      "loss": 1.2744,
+      "step": 7704
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014508089399206745,
+      "loss": 1.0995,
+      "step": 7705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014506698761395566,
+      "loss": 1.2318,
+      "step": 7706
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014505308014209212,
+      "loss": 1.1796,
+      "step": 7707
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014503917157681443,
+      "loss": 1.1182,
+      "step": 7708
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014502526191846005,
+      "loss": 1.2281,
+      "step": 7709
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014501135116736668,
+      "loss": 1.3962,
+      "step": 7710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014499743932387183,
+      "loss": 1.1059,
+      "step": 7711
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014498352638831312,
+      "loss": 1.2314,
+      "step": 7712
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014496961236102834,
+      "loss": 2.3345,
+      "step": 7713
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000144955697242355,
+      "loss": 0.9601,
+      "step": 7714
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014494178103263092,
+      "loss": 1.1314,
+      "step": 7715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014492786373219383,
+      "loss": 1.0043,
+      "step": 7716
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014491394534138147,
+      "loss": 1.2049,
+      "step": 7717
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014490002586053165,
+      "loss": 1.1621,
+      "step": 7718
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014488610528998219,
+      "loss": 0.9642,
+      "step": 7719
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014487218363007097,
+      "loss": 1.0887,
+      "step": 7720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014485826088113577,
+      "loss": 1.3343,
+      "step": 7721
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014484433704351453,
+      "loss": 1.1638,
+      "step": 7722
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014483041211754518,
+      "loss": 1.1933,
+      "step": 7723
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014481648610356572,
+      "loss": 1.3473,
+      "step": 7724
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000144802559001914,
+      "loss": 1.2618,
+      "step": 7725
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014478863081292818,
+      "loss": 0.986,
+      "step": 7726
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014477470153694618,
+      "loss": 1.2495,
+      "step": 7727
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001447607711743061,
+      "loss": 1.004,
+      "step": 7728
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014474683972534602,
+      "loss": 1.1934,
+      "step": 7729
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000144732907190404,
+      "loss": 1.1323,
+      "step": 7730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014471897356981822,
+      "loss": 1.1519,
+      "step": 7731
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014470503886392685,
+      "loss": 2.3342,
+      "step": 7732
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014469110307306805,
+      "loss": 1.1018,
+      "step": 7733
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001446771661975801,
+      "loss": 1.167,
+      "step": 7734
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014466322823780114,
+      "loss": 1.48,
+      "step": 7735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001446492891940695,
+      "loss": 1.3073,
+      "step": 7736
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014463534906672347,
+      "loss": 1.1128,
+      "step": 7737
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014462140785610132,
+      "loss": 1.1354,
+      "step": 7738
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014460746556254145,
+      "loss": 1.1248,
+      "step": 7739
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014459352218638226,
+      "loss": 1.2967,
+      "step": 7740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014457957772796203,
+      "loss": 1.0596,
+      "step": 7741
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001445656321876193,
+      "loss": 1.1389,
+      "step": 7742
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001445516855656925,
+      "loss": 1.1485,
+      "step": 7743
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00014453773786252006,
+      "loss": 1.0929,
+      "step": 7744
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001445237890784405,
+      "loss": 1.2292,
+      "step": 7745
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014450983921379237,
+      "loss": 1.032,
+      "step": 7746
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001444958882689142,
+      "loss": 1.132,
+      "step": 7747
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014448193624414462,
+      "loss": 1.0764,
+      "step": 7748
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001444679831398222,
+      "loss": 1.2267,
+      "step": 7749
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014445402895628555,
+      "loss": 1.1968,
+      "step": 7750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001444400736938734,
+      "loss": 1.0771,
+      "step": 7751
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014442611735292435,
+      "loss": 2.2952,
+      "step": 7752
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014441215993377723,
+      "loss": 1.1405,
+      "step": 7753
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014439820143677062,
+      "loss": 1.2415,
+      "step": 7754
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014438424186224343,
+      "loss": 1.0689,
+      "step": 7755
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001443702812105344,
+      "loss": 1.1639,
+      "step": 7756
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014435631948198232,
+      "loss": 1.1046,
+      "step": 7757
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014434235667692605,
+      "loss": 1.0576,
+      "step": 7758
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001443283927957045,
+      "loss": 1.1324,
+      "step": 7759
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014431442783865653,
+      "loss": 1.1408,
+      "step": 7760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014430046180612105,
+      "loss": 1.1296,
+      "step": 7761
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000144286494698437,
+      "loss": 1.1439,
+      "step": 7762
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001442725265159434,
+      "loss": 1.2103,
+      "step": 7763
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014425855725897922,
+      "loss": 1.1187,
+      "step": 7764
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014424458692788354,
+      "loss": 0.9368,
+      "step": 7765
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001442306155229953,
+      "loss": 1.097,
+      "step": 7766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001442166430446537,
+      "loss": 1.2422,
+      "step": 7767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014420266949319776,
+      "loss": 1.158,
+      "step": 7768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014418869486896663,
+      "loss": 1.1057,
+      "step": 7769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001441747191722995,
+      "loss": 1.0013,
+      "step": 7770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001441607424035355,
+      "loss": 1.276,
+      "step": 7771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001441467645630139,
+      "loss": 1.1672,
+      "step": 7772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001441327856510739,
+      "loss": 1.0968,
+      "step": 7773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014411880566805475,
+      "loss": 1.4738,
+      "step": 7774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014410482461429578,
+      "loss": 1.1557,
+      "step": 7775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014409084249013623,
+      "loss": 1.2687,
+      "step": 7776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001440768592959155,
+      "loss": 1.1415,
+      "step": 7777
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014406287503197294,
+      "loss": 2.4996,
+      "step": 7778
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014404888969864793,
+      "loss": 1.2932,
+      "step": 7779
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014403490329627992,
+      "loss": 1.1222,
+      "step": 7780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014402091582520828,
+      "loss": 1.3143,
+      "step": 7781
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014400692728577257,
+      "loss": 1.0701,
+      "step": 7782
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001439929376783122,
+      "loss": 1.0178,
+      "step": 7783
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014397894700316676,
+      "loss": 1.2171,
+      "step": 7784
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014396495526067574,
+      "loss": 1.2427,
+      "step": 7785
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001439509624511788,
+      "loss": 1.1916,
+      "step": 7786
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014393696857501538,
+      "loss": 1.2316,
+      "step": 7787
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014392297363252524,
+      "loss": 1.0293,
+      "step": 7788
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014390897762404803,
+      "loss": 2.2761,
+      "step": 7789
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001438949805499233,
+      "loss": 1.1504,
+      "step": 7790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014388098241049094,
+      "loss": 1.1591,
+      "step": 7791
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001438669832060905,
+      "loss": 1.0835,
+      "step": 7792
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014385298293706183,
+      "loss": 1.2107,
+      "step": 7793
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014383898160374469,
+      "loss": 1.4233,
+      "step": 7794
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001438249792064789,
+      "loss": 1.2679,
+      "step": 7795
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014381097574560424,
+      "loss": 1.116,
+      "step": 7796
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014379697122146059,
+      "loss": 1.1893,
+      "step": 7797
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014378296563438784,
+      "loss": 1.0234,
+      "step": 7798
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014376895898472592,
+      "loss": 1.0619,
+      "step": 7799
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014375495127281473,
+      "loss": 1.2631,
+      "step": 7800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014374094249899427,
+      "loss": 1.1888,
+      "step": 7801
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001437269326636045,
+      "loss": 1.0558,
+      "step": 7802
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014371292176698538,
+      "loss": 1.1503,
+      "step": 7803
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014369890980947702,
+      "loss": 1.322,
+      "step": 7804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001436848967914195,
+      "loss": 2.2964,
+      "step": 7805
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001436708827131528,
+      "loss": 1.1919,
+      "step": 7806
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014365686757501713,
+      "loss": 1.0933,
+      "step": 7807
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014364285137735263,
+      "loss": 1.1875,
+      "step": 7808
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001436288341204994,
+      "loss": 2.363,
+      "step": 7809
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014361481580479768,
+      "loss": 1.3586,
+      "step": 7810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014360079643058765,
+      "loss": 1.158,
+      "step": 7811
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001435867759982096,
+      "loss": 1.3646,
+      "step": 7812
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014357275450800378,
+      "loss": 1.1921,
+      "step": 7813
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014355873196031046,
+      "loss": 2.2808,
+      "step": 7814
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014354470835547002,
+      "loss": 1.1366,
+      "step": 7815
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014353068369382275,
+      "loss": 1.0774,
+      "step": 7816
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014351665797570902,
+      "loss": 1.0604,
+      "step": 7817
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014350263120146926,
+      "loss": 1.2357,
+      "step": 7818
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014348860337144386,
+      "loss": 1.2176,
+      "step": 7819
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014347457448597327,
+      "loss": 1.0863,
+      "step": 7820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014346054454539804,
+      "loss": 1.0719,
+      "step": 7821
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014344651355005853,
+      "loss": 1.1479,
+      "step": 7822
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001434324815002954,
+      "loss": 1.2704,
+      "step": 7823
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014341844839644912,
+      "loss": 1.2031,
+      "step": 7824
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014340441423886028,
+      "loss": 1.2858,
+      "step": 7825
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014339037902786952,
+      "loss": 1.1557,
+      "step": 7826
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001433763427638174,
+      "loss": 1.1627,
+      "step": 7827
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001433623054470446,
+      "loss": 1.0999,
+      "step": 7828
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014334826707789185,
+      "loss": 1.1209,
+      "step": 7829
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001433342276566998,
+      "loss": 1.0729,
+      "step": 7830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014332018718380915,
+      "loss": 1.3593,
+      "step": 7831
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014330614565956072,
+      "loss": 1.1122,
+      "step": 7832
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014329210308429525,
+      "loss": 1.3344,
+      "step": 7833
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001432780594583536,
+      "loss": 1.4218,
+      "step": 7834
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014326401478207654,
+      "loss": 0.9214,
+      "step": 7835
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014324996905580492,
+      "loss": 0.979,
+      "step": 7836
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014323592227987968,
+      "loss": 1.1647,
+      "step": 7837
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014322187445464172,
+      "loss": 2.3208,
+      "step": 7838
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014320782558043194,
+      "loss": 1.3103,
+      "step": 7839
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014319377565759128,
+      "loss": 1.2281,
+      "step": 7840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001431797246864608,
+      "loss": 1.0961,
+      "step": 7841
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001431656726673814,
+      "loss": 1.1049,
+      "step": 7842
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014315161960069424,
+      "loss": 1.1809,
+      "step": 7843
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001431375654867403,
+      "loss": 1.1114,
+      "step": 7844
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014312351032586066,
+      "loss": 1.3325,
+      "step": 7845
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014310945411839652,
+      "loss": 1.1388,
+      "step": 7846
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014309539686468888,
+      "loss": 1.1488,
+      "step": 7847
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014308133856507903,
+      "loss": 1.1833,
+      "step": 7848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014306727921990808,
+      "loss": 1.0575,
+      "step": 7849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014305321882951725,
+      "loss": 1.1998,
+      "step": 7850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014303915739424783,
+      "loss": 1.2391,
+      "step": 7851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014302509491444102,
+      "loss": 1.1153,
+      "step": 7852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014301103139043817,
+      "loss": 1.0099,
+      "step": 7853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014299696682258052,
+      "loss": 1.0577,
+      "step": 7854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014298290121120947,
+      "loss": 1.1081,
+      "step": 7855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014296883455666636,
+      "loss": 0.9746,
+      "step": 7856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014295476685929258,
+      "loss": 1.3097,
+      "step": 7857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014294069811942954,
+      "loss": 2.2825,
+      "step": 7858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014292662833741875,
+      "loss": 1.2771,
+      "step": 7859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014291255751360159,
+      "loss": 1.0085,
+      "step": 7860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014289848564831955,
+      "loss": 1.0792,
+      "step": 7861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014288441274191418,
+      "loss": 1.126,
+      "step": 7862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014287033879472702,
+      "loss": 1.1531,
+      "step": 7863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014285626380709963,
+      "loss": 1.1752,
+      "step": 7864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014284218777937363,
+      "loss": 1.2061,
+      "step": 7865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001428281107118906,
+      "loss": 1.0809,
+      "step": 7866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014281403260499225,
+      "loss": 1.2469,
+      "step": 7867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014279995345902014,
+      "loss": 1.0628,
+      "step": 7868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014278587327431602,
+      "loss": 1.2416,
+      "step": 7869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014277179205122163,
+      "loss": 1.0263,
+      "step": 7870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014275770979007866,
+      "loss": 1.2235,
+      "step": 7871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014274362649122893,
+      "loss": 2.3169,
+      "step": 7872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014272954215501418,
+      "loss": 1.1845,
+      "step": 7873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014271545678177635,
+      "loss": 1.1557,
+      "step": 7874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014270137037185714,
+      "loss": 1.0916,
+      "step": 7875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014268728292559846,
+      "loss": 1.28,
+      "step": 7876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014267319444334224,
+      "loss": 1.1975,
+      "step": 7877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001426591049254304,
+      "loss": 1.169,
+      "step": 7878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014264501437220484,
+      "loss": 1.0844,
+      "step": 7879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014263092278400758,
+      "loss": 1.207,
+      "step": 7880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014261683016118058,
+      "loss": 2.3357,
+      "step": 7881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001426027365040659,
+      "loss": 1.107,
+      "step": 7882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014258864181300554,
+      "loss": 1.1372,
+      "step": 7883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001425745460883416,
+      "loss": 0.9792,
+      "step": 7884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014256044933041615,
+      "loss": 1.2388,
+      "step": 7885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001425463515395713,
+      "loss": 1.0777,
+      "step": 7886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014253225271614926,
+      "loss": 1.1692,
+      "step": 7887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001425181528604922,
+      "loss": 1.144,
+      "step": 7888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014250405197294223,
+      "loss": 1.0701,
+      "step": 7889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014248995005384162,
+      "loss": 1.0946,
+      "step": 7890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014247584710353264,
+      "loss": 1.0417,
+      "step": 7891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014246174312235752,
+      "loss": 2.2268,
+      "step": 7892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014244763811065855,
+      "loss": 1.1721,
+      "step": 7893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014243353206877808,
+      "loss": 1.0806,
+      "step": 7894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014241942499705848,
+      "loss": 1.1186,
+      "step": 7895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014240531689584208,
+      "loss": 1.0952,
+      "step": 7896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014239120776547132,
+      "loss": 1.137,
+      "step": 7897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014237709760628854,
+      "loss": 1.0292,
+      "step": 7898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014236298641863627,
+      "loss": 1.2213,
+      "step": 7899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014234887420285693,
+      "loss": 1.1355,
+      "step": 7900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000142334760959293,
+      "loss": 1.1676,
+      "step": 7901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014232064668828707,
+      "loss": 1.2235,
+      "step": 7902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014230653139018164,
+      "loss": 1.1994,
+      "step": 7903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014229241506531933,
+      "loss": 1.0644,
+      "step": 7904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014227829771404265,
+      "loss": 1.1587,
+      "step": 7905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014226417933669428,
+      "loss": 1.3137,
+      "step": 7906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014225005993361681,
+      "loss": 1.1981,
+      "step": 7907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000142235939505153,
+      "loss": 1.155,
+      "step": 7908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014222181805164546,
+      "loss": 1.2946,
+      "step": 7909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014220769557343697,
+      "loss": 1.1164,
+      "step": 7910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014219357207087026,
+      "loss": 1.0477,
+      "step": 7911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001421794475442881,
+      "loss": 1.0611,
+      "step": 7912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014216532199403328,
+      "loss": 1.0326,
+      "step": 7913
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014215119542044857,
+      "loss": 0.9808,
+      "step": 7914
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014213706782387688,
+      "loss": 1.0323,
+      "step": 7915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014212293920466107,
+      "loss": 1.074,
+      "step": 7916
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014210880956314398,
+      "loss": 1.0971,
+      "step": 7917
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014209467889966863,
+      "loss": 1.1167,
+      "step": 7918
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001420805472145779,
+      "loss": 1.0948,
+      "step": 7919
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014206641450821474,
+      "loss": 0.9846,
+      "step": 7920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014205228078092218,
+      "loss": 1.2211,
+      "step": 7921
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014203814603304323,
+      "loss": 1.1425,
+      "step": 7922
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014202401026492092,
+      "loss": 1.1353,
+      "step": 7923
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014200987347689833,
+      "loss": 1.0649,
+      "step": 7924
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014199573566931855,
+      "loss": 1.2278,
+      "step": 7925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014198159684252468,
+      "loss": 1.1633,
+      "step": 7926
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014196745699685993,
+      "loss": 1.1997,
+      "step": 7927
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014195331613266733,
+      "loss": 1.0425,
+      "step": 7928
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014193917425029022,
+      "loss": 1.1668,
+      "step": 7929
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001419250313500717,
+      "loss": 1.136,
+      "step": 7930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001419108874323551,
+      "loss": 1.2614,
+      "step": 7931
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014189674249748363,
+      "loss": 1.1831,
+      "step": 7932
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014188259654580063,
+      "loss": 1.2694,
+      "step": 7933
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014186844957764932,
+      "loss": 1.1188,
+      "step": 7934
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014185430159337317,
+      "loss": 1.2679,
+      "step": 7935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014184015259331544,
+      "loss": 1.2069,
+      "step": 7936
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014182600257781956,
+      "loss": 0.9902,
+      "step": 7937
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001418118515472289,
+      "loss": 1.3908,
+      "step": 7938
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014179769950188698,
+      "loss": 1.1928,
+      "step": 7939
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001417835464421372,
+      "loss": 1.0628,
+      "step": 7940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014176939236832308,
+      "loss": 1.132,
+      "step": 7941
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001417552372807881,
+      "loss": 1.0636,
+      "step": 7942
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014174108117987583,
+      "loss": 1.1106,
+      "step": 7943
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001417269240659298,
+      "loss": 1.2073,
+      "step": 7944
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001417127659392936,
+      "loss": 0.9617,
+      "step": 7945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014169860680031087,
+      "loss": 1.1831,
+      "step": 7946
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014168444664932519,
+      "loss": 1.1239,
+      "step": 7947
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014167028548668032,
+      "loss": 1.2526,
+      "step": 7948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014165612331271983,
+      "loss": 1.3025,
+      "step": 7949
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014164196012778746,
+      "loss": 1.079,
+      "step": 7950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000141627795932227,
+      "loss": 1.155,
+      "step": 7951
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014161363072638213,
+      "loss": 2.3477,
+      "step": 7952
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001415994645105967,
+      "loss": 1.2709,
+      "step": 7953
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014158529728521447,
+      "loss": 1.1631,
+      "step": 7954
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014157112905057928,
+      "loss": 1.1119,
+      "step": 7955
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000141556959807035,
+      "loss": 1.0606,
+      "step": 7956
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014154278955492552,
+      "loss": 1.0859,
+      "step": 7957
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014152861829459467,
+      "loss": 1.2303,
+      "step": 7958
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014151444602638648,
+      "loss": 1.2493,
+      "step": 7959
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014150027275064485,
+      "loss": 1.1723,
+      "step": 7960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014148609846771377,
+      "loss": 1.0684,
+      "step": 7961
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001414719231779372,
+      "loss": 1.1823,
+      "step": 7962
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014145774688165927,
+      "loss": 2.4009,
+      "step": 7963
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014144356957922392,
+      "loss": 1.0539,
+      "step": 7964
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001414293912709753,
+      "loss": 1.2521,
+      "step": 7965
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014141521195725745,
+      "loss": 1.1045,
+      "step": 7966
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014140103163841453,
+      "loss": 1.1867,
+      "step": 7967
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014138685031479068,
+      "loss": 1.0113,
+      "step": 7968
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001413726679867301,
+      "loss": 1.1052,
+      "step": 7969
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000141358484654577,
+      "loss": 1.0766,
+      "step": 7970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014134430031867548,
+      "loss": 0.8992,
+      "step": 7971
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014133011497936994,
+      "loss": 1.2127,
+      "step": 7972
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014131592863700454,
+      "loss": 1.2156,
+      "step": 7973
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014130174129192366,
+      "loss": 1.0669,
+      "step": 7974
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014128755294447157,
+      "loss": 1.1247,
+      "step": 7975
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001412733635949926,
+      "loss": 0.9677,
+      "step": 7976
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014125917324383114,
+      "loss": 0.9512,
+      "step": 7977
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014124498189133162,
+      "loss": 1.0513,
+      "step": 7978
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001412307895378384,
+      "loss": 0.9213,
+      "step": 7979
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014121659618369591,
+      "loss": 1.1631,
+      "step": 7980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001412024018292487,
+      "loss": 1.181,
+      "step": 7981
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014118820647484115,
+      "loss": 1.0873,
+      "step": 7982
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014117401012081783,
+      "loss": 1.1494,
+      "step": 7983
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001411598127675233,
+      "loss": 1.1216,
+      "step": 7984
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001411456144153021,
+      "loss": 1.1782,
+      "step": 7985
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014113141506449882,
+      "loss": 0.8492,
+      "step": 7986
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014111721471545804,
+      "loss": 1.3365,
+      "step": 7987
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001411030133685244,
+      "loss": 1.0494,
+      "step": 7988
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014108881102404264,
+      "loss": 1.1127,
+      "step": 7989
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014107460768235735,
+      "loss": 1.3213,
+      "step": 7990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014106040334381327,
+      "loss": 1.1544,
+      "step": 7991
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014104619800875515,
+      "loss": 2.449,
+      "step": 7992
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001410319916775277,
+      "loss": 1.3044,
+      "step": 7993
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014101778435047574,
+      "loss": 1.3378,
+      "step": 7994
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014100357602794404,
+      "loss": 1.3381,
+      "step": 7995
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014098936671027746,
+      "loss": 1.2406,
+      "step": 7996
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014097515639782085,
+      "loss": 1.2189,
+      "step": 7997
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001409609450909191,
+      "loss": 1.0657,
+      "step": 7998
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014094673278991707,
+      "loss": 1.1524,
+      "step": 7999
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001409325194951597,
+      "loss": 0.9737,
+      "step": 8000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014091830520699197,
+      "loss": 1.1593,
+      "step": 8001
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001409040899257588,
+      "loss": 1.2135,
+      "step": 8002
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014088987365180522,
+      "loss": 1.1548,
+      "step": 8003
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014087565638547624,
+      "loss": 1.1821,
+      "step": 8004
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001408614381271169,
+      "loss": 1.0301,
+      "step": 8005
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014084721887707233,
+      "loss": 1.1464,
+      "step": 8006
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014083299863568754,
+      "loss": 1.0704,
+      "step": 8007
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001408187774033077,
+      "loss": 1.2067,
+      "step": 8008
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001408045551802779,
+      "loss": 1.0829,
+      "step": 8009
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014079033196694337,
+      "loss": 1.1049,
+      "step": 8010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014077610776364928,
+      "loss": 1.2294,
+      "step": 8011
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001407618825707408,
+      "loss": 2.3491,
+      "step": 8012
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001407476563885632,
+      "loss": 1.1029,
+      "step": 8013
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014073342921746177,
+      "loss": 1.0999,
+      "step": 8014
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014071920105778176,
+      "loss": 1.1993,
+      "step": 8015
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001407049719098685,
+      "loss": 1.065,
+      "step": 8016
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014069074177406728,
+      "loss": 0.9004,
+      "step": 8017
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001406765106507235,
+      "loss": 1.2736,
+      "step": 8018
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014066227854018254,
+      "loss": 0.9717,
+      "step": 8019
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014064804544278978,
+      "loss": 1.2176,
+      "step": 8020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001406338113588907,
+      "loss": 1.2194,
+      "step": 8021
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001406195762888307,
+      "loss": 1.1849,
+      "step": 8022
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014060534023295528,
+      "loss": 1.3351,
+      "step": 8023
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014059110319160998,
+      "loss": 1.1844,
+      "step": 8024
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014057686516514024,
+      "loss": 1.1735,
+      "step": 8025
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014056262615389168,
+      "loss": 1.2055,
+      "step": 8026
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001405483861582098,
+      "loss": 1.0364,
+      "step": 8027
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014053414517844028,
+      "loss": 1.1862,
+      "step": 8028
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001405199032149287,
+      "loss": 1.2653,
+      "step": 8029
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014050566026802073,
+      "loss": 1.1142,
+      "step": 8030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000140491416338062,
+      "loss": 1.2089,
+      "step": 8031
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014047717142539823,
+      "loss": 1.1175,
+      "step": 8032
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001404629255303751,
+      "loss": 1.2029,
+      "step": 8033
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001404486786533384,
+      "loss": 1.1436,
+      "step": 8034
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014043443079463387,
+      "loss": 1.2293,
+      "step": 8035
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014042018195460732,
+      "loss": 1.0634,
+      "step": 8036
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014040593213360454,
+      "loss": 1.1116,
+      "step": 8037
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014039168133197136,
+      "loss": 2.3557,
+      "step": 8038
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014037742955005364,
+      "loss": 1.2155,
+      "step": 8039
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001403631767881973,
+      "loss": 1.2428,
+      "step": 8040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014034892304674816,
+      "loss": 1.3056,
+      "step": 8041
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014033466832605223,
+      "loss": 1.1841,
+      "step": 8042
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001403204126264555,
+      "loss": 1.0836,
+      "step": 8043
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014030615594830384,
+      "loss": 1.0669,
+      "step": 8044
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001402918982919433,
+      "loss": 1.0971,
+      "step": 8045
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014027763965771992,
+      "loss": 1.1803,
+      "step": 8046
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014026338004597972,
+      "loss": 1.2832,
+      "step": 8047
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001402491194570688,
+      "loss": 0.9973,
+      "step": 8048
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014023485789133326,
+      "loss": 1.3377,
+      "step": 8049
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001402205953491192,
+      "loss": 1.1284,
+      "step": 8050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001402063318307728,
+      "loss": 1.1398,
+      "step": 8051
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014019206733664015,
+      "loss": 1.1634,
+      "step": 8052
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001401778018670675,
+      "loss": 1.2175,
+      "step": 8053
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001401635354224011,
+      "loss": 1.1616,
+      "step": 8054
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001401492680029871,
+      "loss": 1.1969,
+      "step": 8055
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00014013499960917187,
+      "loss": 1.2248,
+      "step": 8056
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00014012073024130157,
+      "loss": 0.9806,
+      "step": 8057
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001401064598997226,
+      "loss": 1.1635,
+      "step": 8058
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001400921885847813,
+      "loss": 1.3041,
+      "step": 8059
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00014007791629682396,
+      "loss": 2.3945,
+      "step": 8060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000140063643036197,
+      "loss": 1.15,
+      "step": 8061
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00014004936880324684,
+      "loss": 1.1044,
+      "step": 8062
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00014003509359831985,
+      "loss": 1.2193,
+      "step": 8063
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00014002081742176258,
+      "loss": 1.2758,
+      "step": 8064
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00014000654027392141,
+      "loss": 0.9823,
+      "step": 8065
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013999226215514286,
+      "loss": 1.0265,
+      "step": 8066
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001399779830657735,
+      "loss": 1.304,
+      "step": 8067
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013996370300615985,
+      "loss": 1.1611,
+      "step": 8068
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013994942197664845,
+      "loss": 1.0643,
+      "step": 8069
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001399351399775859,
+      "loss": 1.1929,
+      "step": 8070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013992085700931885,
+      "loss": 0.9268,
+      "step": 8071
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013990657307219395,
+      "loss": 1.2084,
+      "step": 8072
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001398922881665578,
+      "loss": 2.1365,
+      "step": 8073
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013987800229275715,
+      "loss": 1.2571,
+      "step": 8074
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013986371545113868,
+      "loss": 1.3022,
+      "step": 8075
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001398494276420491,
+      "loss": 1.1147,
+      "step": 8076
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001398351388658352,
+      "loss": 1.1859,
+      "step": 8077
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013982084912284376,
+      "loss": 1.1339,
+      "step": 8078
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001398065584134216,
+      "loss": 0.9796,
+      "step": 8079
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001397922667379155,
+      "loss": 1.2432,
+      "step": 8080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013977797409667238,
+      "loss": 1.1397,
+      "step": 8081
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013976368049003902,
+      "loss": 1.1279,
+      "step": 8082
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013974938591836242,
+      "loss": 1.0935,
+      "step": 8083
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013973509038198943,
+      "loss": 2.3771,
+      "step": 8084
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000139720793881267,
+      "loss": 1.1606,
+      "step": 8085
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013970649641654217,
+      "loss": 0.9088,
+      "step": 8086
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013969219798816183,
+      "loss": 1.1093,
+      "step": 8087
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013967789859647305,
+      "loss": 1.2157,
+      "step": 8088
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001396635982418229,
+      "loss": 1.1307,
+      "step": 8089
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001396492969245584,
+      "loss": 1.0624,
+      "step": 8090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001396349946450266,
+      "loss": 1.2789,
+      "step": 8091
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001396206914035747,
+      "loss": 1.095,
+      "step": 8092
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013960638720054972,
+      "loss": 1.2592,
+      "step": 8093
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001395920820362989,
+      "loss": 0.9079,
+      "step": 8094
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013957777591116946,
+      "loss": 1.2303,
+      "step": 8095
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013956346882550846,
+      "loss": 1.0327,
+      "step": 8096
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013954916077966325,
+      "loss": 1.3314,
+      "step": 8097
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000139534851773981,
+      "loss": 2.2883,
+      "step": 8098
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013952054180880902,
+      "loss": 0.9297,
+      "step": 8099
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001395062308844946,
+      "loss": 1.1024,
+      "step": 8100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013949191900138502,
+      "loss": 1.317,
+      "step": 8101
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001394776061598277,
+      "loss": 1.147,
+      "step": 8102
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013946329236016997,
+      "loss": 1.2285,
+      "step": 8103
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001394489776027592,
+      "loss": 0.9836,
+      "step": 8104
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013943466188794283,
+      "loss": 1.0353,
+      "step": 8105
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013942034521606823,
+      "loss": 1.203,
+      "step": 8106
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013940602758748293,
+      "loss": 1.1429,
+      "step": 8107
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001393917090025344,
+      "loss": 1.0174,
+      "step": 8108
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013937738946157013,
+      "loss": 1.0549,
+      "step": 8109
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013936306896493767,
+      "loss": 1.012,
+      "step": 8110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013934874751298453,
+      "loss": 1.0652,
+      "step": 8111
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001393344251060583,
+      "loss": 1.0452,
+      "step": 8112
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013932010174450658,
+      "loss": 1.1016,
+      "step": 8113
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013930577742867699,
+      "loss": 0.9852,
+      "step": 8114
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013929145215891718,
+      "loss": 1.1343,
+      "step": 8115
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001392771259355748,
+      "loss": 1.0978,
+      "step": 8116
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013926279875899758,
+      "loss": 1.0012,
+      "step": 8117
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013924847062953317,
+      "loss": 1.1008,
+      "step": 8118
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013923414154752938,
+      "loss": 1.1493,
+      "step": 8119
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001392198115133339,
+      "loss": 1.2495,
+      "step": 8120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013920548052729457,
+      "loss": 1.0959,
+      "step": 8121
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001391911485897591,
+      "loss": 1.1029,
+      "step": 8122
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013917681570107545,
+      "loss": 1.1103,
+      "step": 8123
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013916248186159142,
+      "loss": 1.1007,
+      "step": 8124
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001391481470716548,
+      "loss": 1.0518,
+      "step": 8125
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013913381133161363,
+      "loss": 1.0748,
+      "step": 8126
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013911947464181573,
+      "loss": 1.0548,
+      "step": 8127
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013910513700260906,
+      "loss": 1.1293,
+      "step": 8128
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013909079841434164,
+      "loss": 0.9374,
+      "step": 8129
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013907645887736137,
+      "loss": 1.1801,
+      "step": 8130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013906211839201634,
+      "loss": 1.1495,
+      "step": 8131
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001390477769586546,
+      "loss": 2.1672,
+      "step": 8132
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013903343457762413,
+      "loss": 1.0644,
+      "step": 8133
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013901909124927305,
+      "loss": 1.2012,
+      "step": 8134
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013900474697394946,
+      "loss": 1.1435,
+      "step": 8135
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001389904017520015,
+      "loss": 1.1959,
+      "step": 8136
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013897605558377732,
+      "loss": 1.0338,
+      "step": 8137
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001389617084696251,
+      "loss": 1.0374,
+      "step": 8138
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000138947360409893,
+      "loss": 2.0016,
+      "step": 8139
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001389330114049293,
+      "loss": 1.3254,
+      "step": 8140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013891866145508217,
+      "loss": 1.1302,
+      "step": 8141
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001389043105606999,
+      "loss": 1.1502,
+      "step": 8142
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013888995872213085,
+      "loss": 2.29,
+      "step": 8143
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013887560593972323,
+      "loss": 1.065,
+      "step": 8144
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013886125221382542,
+      "loss": 1.213,
+      "step": 8145
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001388468975447858,
+      "loss": 1.2362,
+      "step": 8146
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001388325419329527,
+      "loss": 1.1532,
+      "step": 8147
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013881818537867454,
+      "loss": 1.1333,
+      "step": 8148
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013880382788229978,
+      "loss": 1.1118,
+      "step": 8149
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001387894694441768,
+      "loss": 1.1865,
+      "step": 8150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013877511006465417,
+      "loss": 1.2463,
+      "step": 8151
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013876074974408028,
+      "loss": 1.2131,
+      "step": 8152
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013874638848280374,
+      "loss": 1.1742,
+      "step": 8153
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013873202628117306,
+      "loss": 1.1959,
+      "step": 8154
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013871766313953673,
+      "loss": 1.2594,
+      "step": 8155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013870329905824345,
+      "loss": 2.5032,
+      "step": 8156
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001386889340376417,
+      "loss": 1.1731,
+      "step": 8157
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013867456807808024,
+      "loss": 1.1821,
+      "step": 8158
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013866020117990764,
+      "loss": 0.9601,
+      "step": 8159
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00013864583334347264,
+      "loss": 1.1077,
+      "step": 8160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013863146456912387,
+      "loss": 2.4875,
+      "step": 8161
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013861709485721014,
+      "loss": 2.2449,
+      "step": 8162
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013860272420808013,
+      "loss": 1.2095,
+      "step": 8163
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013858835262208258,
+      "loss": 1.0166,
+      "step": 8164
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013857398009956635,
+      "loss": 1.2563,
+      "step": 8165
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001385596066408802,
+      "loss": 1.1302,
+      "step": 8166
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013854523224637305,
+      "loss": 1.19,
+      "step": 8167
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013853085691639365,
+      "loss": 1.1454,
+      "step": 8168
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013851648065129092,
+      "loss": 1.1559,
+      "step": 8169
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013850210345141384,
+      "loss": 1.1118,
+      "step": 8170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001384877253171112,
+      "loss": 1.178,
+      "step": 8171
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013847334624873204,
+      "loss": 0.9948,
+      "step": 8172
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013845896624662535,
+      "loss": 1.0516,
+      "step": 8173
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013844458531114007,
+      "loss": 1.354,
+      "step": 8174
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013843020344262526,
+      "loss": 1.0052,
+      "step": 8175
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013841582064142991,
+      "loss": 1.0964,
+      "step": 8176
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001384014369079031,
+      "loss": 1.2529,
+      "step": 8177
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013838705224239397,
+      "loss": 1.0237,
+      "step": 8178
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013837266664525156,
+      "loss": 1.1281,
+      "step": 8179
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013835828011682502,
+      "loss": 2.2832,
+      "step": 8180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013834389265746352,
+      "loss": 1.2124,
+      "step": 8181
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001383295042675162,
+      "loss": 1.1211,
+      "step": 8182
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013831511494733233,
+      "loss": 1.1151,
+      "step": 8183
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013830072469726105,
+      "loss": 0.9373,
+      "step": 8184
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013828633351765164,
+      "loss": 1.11,
+      "step": 8185
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001382719414088534,
+      "loss": 1.1577,
+      "step": 8186
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013825754837121554,
+      "loss": 1.1056,
+      "step": 8187
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013824315440508743,
+      "loss": 1.0038,
+      "step": 8188
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001382287595108184,
+      "loss": 1.2278,
+      "step": 8189
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013821436368875777,
+      "loss": 1.1403,
+      "step": 8190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013819996693925494,
+      "loss": 1.1643,
+      "step": 8191
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013818556926265934,
+      "loss": 1.2793,
+      "step": 8192
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013817117065932036,
+      "loss": 1.0201,
+      "step": 8193
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013815677112958747,
+      "loss": 1.0127,
+      "step": 8194
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001381423706738101,
+      "loss": 1.1387,
+      "step": 8195
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013812796929233776,
+      "loss": 1.2031,
+      "step": 8196
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013811356698552001,
+      "loss": 2.3987,
+      "step": 8197
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013809916375370632,
+      "loss": 1.0164,
+      "step": 8198
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013808475959724628,
+      "loss": 1.1914,
+      "step": 8199
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013807035451648947,
+      "loss": 0.9796,
+      "step": 8200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013805594851178546,
+      "loss": 1.0672,
+      "step": 8201
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013804154158348394,
+      "loss": 1.2213,
+      "step": 8202
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001380271337319345,
+      "loss": 1.0671,
+      "step": 8203
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013801272495748683,
+      "loss": 2.3303,
+      "step": 8204
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013799831526049066,
+      "loss": 2.2869,
+      "step": 8205
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013798390464129568,
+      "loss": 1.3172,
+      "step": 8206
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013796949310025157,
+      "loss": 1.1527,
+      "step": 8207
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001379550806377082,
+      "loss": 1.157,
+      "step": 8208
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013794066725401525,
+      "loss": 1.2542,
+      "step": 8209
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001379262529495226,
+      "loss": 0.9433,
+      "step": 8210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013791183772458,
+      "loss": 1.1474,
+      "step": 8211
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001378974215795374,
+      "loss": 0.9957,
+      "step": 8212
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001378830045147446,
+      "loss": 1.1686,
+      "step": 8213
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013786858653055153,
+      "loss": 1.2361,
+      "step": 8214
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013785416762730806,
+      "loss": 1.2262,
+      "step": 8215
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013783974780536416,
+      "loss": 1.1393,
+      "step": 8216
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001378253270650698,
+      "loss": 1.1664,
+      "step": 8217
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013781090540677498,
+      "loss": 1.2773,
+      "step": 8218
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013779648283082966,
+      "loss": 0.8546,
+      "step": 8219
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001377820593375839,
+      "loss": 1.0754,
+      "step": 8220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013776763492738772,
+      "loss": 1.1287,
+      "step": 8221
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001377532096005912,
+      "loss": 1.0753,
+      "step": 8222
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013773878335754448,
+      "loss": 1.126,
+      "step": 8223
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013772435619859765,
+      "loss": 1.1124,
+      "step": 8224
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013770992812410081,
+      "loss": 0.8154,
+      "step": 8225
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001376954991344042,
+      "loss": 2.259,
+      "step": 8226
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013768106922985792,
+      "loss": 1.2019,
+      "step": 8227
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013766663841081224,
+      "loss": 1.1754,
+      "step": 8228
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013765220667761737,
+      "loss": 1.1014,
+      "step": 8229
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001376377740306235,
+      "loss": 1.1009,
+      "step": 8230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013762334047018099,
+      "loss": 1.1019,
+      "step": 8231
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013760890599664012,
+      "loss": 1.2256,
+      "step": 8232
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013759447061035117,
+      "loss": 1.2233,
+      "step": 8233
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013758003431166448,
+      "loss": 1.1866,
+      "step": 8234
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013756559710093047,
+      "loss": 1.1539,
+      "step": 8235
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013755115897849947,
+      "loss": 2.3557,
+      "step": 8236
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013753671994472185,
+      "loss": 1.1611,
+      "step": 8237
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013752227999994814,
+      "loss": 1.1566,
+      "step": 8238
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013750783914452872,
+      "loss": 0.9597,
+      "step": 8239
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013749339737881408,
+      "loss": 1.1535,
+      "step": 8240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001374789547031547,
+      "loss": 1.1626,
+      "step": 8241
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013746451111790113,
+      "loss": 1.0703,
+      "step": 8242
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001374500666234039,
+      "loss": 1.1487,
+      "step": 8243
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001374356212200135,
+      "loss": 1.147,
+      "step": 8244
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013742117490808057,
+      "loss": 0.9658,
+      "step": 8245
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013740672768795576,
+      "loss": 1.101,
+      "step": 8246
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013739227955998962,
+      "loss": 1.0148,
+      "step": 8247
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013737783052453282,
+      "loss": 1.063,
+      "step": 8248
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013736338058193605,
+      "loss": 1.271,
+      "step": 8249
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013734892973254998,
+      "loss": 2.3022,
+      "step": 8250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013733447797672536,
+      "loss": 1.3489,
+      "step": 8251
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013732002531481287,
+      "loss": 1.0753,
+      "step": 8252
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001373055717471633,
+      "loss": 1.0343,
+      "step": 8253
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013729111727412745,
+      "loss": 1.0265,
+      "step": 8254
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013727666189605606,
+      "loss": 0.8176,
+      "step": 8255
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013726220561330004,
+      "loss": 1.0757,
+      "step": 8256
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013724774842621018,
+      "loss": 1.0128,
+      "step": 8257
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013723329033513734,
+      "loss": 1.0757,
+      "step": 8258
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013721883134043246,
+      "loss": 1.1605,
+      "step": 8259
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001372043714424464,
+      "loss": 1.1746,
+      "step": 8260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001371899106415301,
+      "loss": 1.1256,
+      "step": 8261
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013717544893803457,
+      "loss": 1.1213,
+      "step": 8262
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013716098633231074,
+      "loss": 1.2206,
+      "step": 8263
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00013714652282470962,
+      "loss": 1.0115,
+      "step": 8264
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013713205841558223,
+      "loss": 1.2222,
+      "step": 8265
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013711759310527962,
+      "loss": 1.312,
+      "step": 8266
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013710312689415284,
+      "loss": 1.2405,
+      "step": 8267
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013708865978255297,
+      "loss": 1.0959,
+      "step": 8268
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013707419177083114,
+      "loss": 1.2013,
+      "step": 8269
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001370597228593385,
+      "loss": 2.2766,
+      "step": 8270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013704525304842618,
+      "loss": 1.0882,
+      "step": 8271
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013703078233844534,
+      "loss": 1.1862,
+      "step": 8272
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013701631072974722,
+      "loss": 1.1144,
+      "step": 8273
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013700183822268295,
+      "loss": 1.0942,
+      "step": 8274
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013698736481760387,
+      "loss": 1.2798,
+      "step": 8275
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001369728905148612,
+      "loss": 1.0576,
+      "step": 8276
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013695841531480619,
+      "loss": 1.1322,
+      "step": 8277
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013694393921779023,
+      "loss": 1.0984,
+      "step": 8278
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013692946222416459,
+      "loss": 1.0822,
+      "step": 8279
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001369149843342806,
+      "loss": 1.0763,
+      "step": 8280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013690050554848967,
+      "loss": 1.1413,
+      "step": 8281
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001368860258671432,
+      "loss": 1.2063,
+      "step": 8282
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013687154529059257,
+      "loss": 1.0653,
+      "step": 8283
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013685706381918919,
+      "loss": 1.0922,
+      "step": 8284
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013684258145328458,
+      "loss": 1.0113,
+      "step": 8285
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013682809819323024,
+      "loss": 1.0048,
+      "step": 8286
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001368136140393776,
+      "loss": 1.1881,
+      "step": 8287
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013679912899207817,
+      "loss": 1.1274,
+      "step": 8288
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013678464305168358,
+      "loss": 1.124,
+      "step": 8289
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013677015621854532,
+      "loss": 1.286,
+      "step": 8290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013675566849301505,
+      "loss": 1.058,
+      "step": 8291
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013674117987544428,
+      "loss": 0.9589,
+      "step": 8292
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013672669036618473,
+      "loss": 1.0027,
+      "step": 8293
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013671219996558805,
+      "loss": 1.1788,
+      "step": 8294
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001366977086740058,
+      "loss": 1.1937,
+      "step": 8295
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013668321649178985,
+      "loss": 1.0775,
+      "step": 8296
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013666872341929178,
+      "loss": 2.3735,
+      "step": 8297
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001366542294568634,
+      "loss": 1.1516,
+      "step": 8298
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013663973460485641,
+      "loss": 1.2087,
+      "step": 8299
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013662523886362268,
+      "loss": 1.1087,
+      "step": 8300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013661074223351394,
+      "loss": 1.1697,
+      "step": 8301
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013659624471488207,
+      "loss": 0.9696,
+      "step": 8302
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013658174630807883,
+      "loss": 0.8458,
+      "step": 8303
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013656724701345617,
+      "loss": 1.2042,
+      "step": 8304
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013655274683136597,
+      "loss": 1.1732,
+      "step": 8305
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001365382457621601,
+      "loss": 1.0751,
+      "step": 8306
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013652374380619054,
+      "loss": 1.1262,
+      "step": 8307
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013650924096380925,
+      "loss": 1.1418,
+      "step": 8308
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013649473723536813,
+      "loss": 1.0707,
+      "step": 8309
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013648023262121924,
+      "loss": 1.1377,
+      "step": 8310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013646572712171465,
+      "loss": 1.1342,
+      "step": 8311
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013645122073720626,
+      "loss": 1.1576,
+      "step": 8312
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013643671346804626,
+      "loss": 1.2854,
+      "step": 8313
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013642220531458668,
+      "loss": 1.1855,
+      "step": 8314
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001364076962771796,
+      "loss": 1.1705,
+      "step": 8315
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013639318635617721,
+      "loss": 1.0101,
+      "step": 8316
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013637867555193163,
+      "loss": 1.0695,
+      "step": 8317
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013636416386479498,
+      "loss": 1.2591,
+      "step": 8318
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013634965129511955,
+      "loss": 1.2251,
+      "step": 8319
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013633513784325744,
+      "loss": 1.1324,
+      "step": 8320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013632062350956095,
+      "loss": 1.2217,
+      "step": 8321
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013630610829438235,
+      "loss": 1.1663,
+      "step": 8322
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001362915921980739,
+      "loss": 1.1574,
+      "step": 8323
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001362770752209879,
+      "loss": 0.9823,
+      "step": 8324
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013626255736347664,
+      "loss": 1.0432,
+      "step": 8325
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013624803862589247,
+      "loss": 1.2449,
+      "step": 8326
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001362335190085878,
+      "loss": 1.1256,
+      "step": 8327
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013621899851191493,
+      "loss": 1.271,
+      "step": 8328
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013620447713622632,
+      "loss": 1.2677,
+      "step": 8329
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013618995488187444,
+      "loss": 1.2228,
+      "step": 8330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013617543174921162,
+      "loss": 1.239,
+      "step": 8331
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001361609077385904,
+      "loss": 1.2772,
+      "step": 8332
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001361463828503633,
+      "loss": 1.1417,
+      "step": 8333
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013613185708488275,
+      "loss": 1.1227,
+      "step": 8334
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013611733044250138,
+      "loss": 1.1504,
+      "step": 8335
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013610280292357163,
+      "loss": 1.2703,
+      "step": 8336
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013608827452844617,
+      "loss": 1.0976,
+      "step": 8337
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001360737452574776,
+      "loss": 2.4119,
+      "step": 8338
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013605921511101841,
+      "loss": 0.9988,
+      "step": 8339
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013604468408942138,
+      "loss": 1.1302,
+      "step": 8340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001360301521930391,
+      "loss": 1.0951,
+      "step": 8341
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001360156194222243,
+      "loss": 1.1227,
+      "step": 8342
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013600108577732964,
+      "loss": 1.1689,
+      "step": 8343
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013598655125870784,
+      "loss": 1.0681,
+      "step": 8344
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001359720158667117,
+      "loss": 1.206,
+      "step": 8345
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013595747960169392,
+      "loss": 1.209,
+      "step": 8346
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001359429424640073,
+      "loss": 1.2164,
+      "step": 8347
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001359284044540047,
+      "loss": 0.925,
+      "step": 8348
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001359138655720389,
+      "loss": 1.2457,
+      "step": 8349
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013589932581846274,
+      "loss": 1.2582,
+      "step": 8350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013588478519362913,
+      "loss": 1.1459,
+      "step": 8351
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013587024369789095,
+      "loss": 2.2708,
+      "step": 8352
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013585570133160113,
+      "loss": 1.0494,
+      "step": 8353
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001358411580951126,
+      "loss": 1.0547,
+      "step": 8354
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013582661398877827,
+      "loss": 1.1366,
+      "step": 8355
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001358120690129512,
+      "loss": 1.2004,
+      "step": 8356
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013579752316798428,
+      "loss": 1.1881,
+      "step": 8357
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001357829764542306,
+      "loss": 1.3389,
+      "step": 8358
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013576842887204327,
+      "loss": 0.9924,
+      "step": 8359
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013575388042177518,
+      "loss": 1.1827,
+      "step": 8360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013573933110377958,
+      "loss": 1.0677,
+      "step": 8361
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013572478091840946,
+      "loss": 1.225,
+      "step": 8362
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.000135710229866018,
+      "loss": 1.3102,
+      "step": 8363
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013569567794695833,
+      "loss": 1.1075,
+      "step": 8364
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013568112516158362,
+      "loss": 1.0384,
+      "step": 8365
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00013566657151024706,
+      "loss": 1.1731,
+      "step": 8366
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001356520169933019,
+      "loss": 1.1118,
+      "step": 8367
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001356374616111013,
+      "loss": 1.392,
+      "step": 8368
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013562290536399852,
+      "loss": 1.2449,
+      "step": 8369
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013560834825234686,
+      "loss": 1.0949,
+      "step": 8370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001355937902764996,
+      "loss": 1.0849,
+      "step": 8371
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001355792314368101,
+      "loss": 2.4021,
+      "step": 8372
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013556467173363163,
+      "loss": 1.1154,
+      "step": 8373
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013555011116731758,
+      "loss": 2.228,
+      "step": 8374
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001355355497382213,
+      "loss": 1.2897,
+      "step": 8375
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013552098744669625,
+      "loss": 1.2266,
+      "step": 8376
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013550642429309575,
+      "loss": 1.1573,
+      "step": 8377
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013549186027777334,
+      "loss": 1.1858,
+      "step": 8378
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001354772954010824,
+      "loss": 1.2037,
+      "step": 8379
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013546272966337647,
+      "loss": 1.2045,
+      "step": 8380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013544816306500904,
+      "loss": 1.1287,
+      "step": 8381
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013543359560633364,
+      "loss": 1.0692,
+      "step": 8382
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013541902728770377,
+      "loss": 1.3518,
+      "step": 8383
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013540445810947304,
+      "loss": 1.0474,
+      "step": 8384
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013538988807199498,
+      "loss": 1.3022,
+      "step": 8385
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001353753171756233,
+      "loss": 0.9554,
+      "step": 8386
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013536074542071154,
+      "loss": 1.2003,
+      "step": 8387
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013534617280761335,
+      "loss": 0.9823,
+      "step": 8388
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013533159933668248,
+      "loss": 1.0265,
+      "step": 8389
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013531702500827253,
+      "loss": 1.0877,
+      "step": 8390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013530244982273723,
+      "loss": 1.1464,
+      "step": 8391
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013528787378043038,
+      "loss": 1.1779,
+      "step": 8392
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013527329688170562,
+      "loss": 1.2456,
+      "step": 8393
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013525871912691684,
+      "loss": 1.0695,
+      "step": 8394
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013524414051641775,
+      "loss": 1.2142,
+      "step": 8395
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013522956105056219,
+      "loss": 1.1356,
+      "step": 8396
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013521498072970403,
+      "loss": 1.0255,
+      "step": 8397
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013520039955419705,
+      "loss": 1.0784,
+      "step": 8398
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001351858175243952,
+      "loss": 0.9258,
+      "step": 8399
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013517123464065232,
+      "loss": 1.1573,
+      "step": 8400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001351566509033224,
+      "loss": 1.1408,
+      "step": 8401
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013514206631275934,
+      "loss": 1.1998,
+      "step": 8402
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013512748086931707,
+      "loss": 1.0913,
+      "step": 8403
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013511289457334964,
+      "loss": 1.0453,
+      "step": 8404
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.000135098307425211,
+      "loss": 1.1824,
+      "step": 8405
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013508371942525516,
+      "loss": 1.1495,
+      "step": 8406
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001350691305738362,
+      "loss": 1.2111,
+      "step": 8407
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013505454087130817,
+      "loss": 1.2816,
+      "step": 8408
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013503995031802517,
+      "loss": 2.3059,
+      "step": 8409
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013502535891434127,
+      "loss": 1.0399,
+      "step": 8410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013501076666061064,
+      "loss": 2.6682,
+      "step": 8411
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013499617355718739,
+      "loss": 1.2689,
+      "step": 8412
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001349815796044257,
+      "loss": 1.1488,
+      "step": 8413
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013496698480267976,
+      "loss": 1.2815,
+      "step": 8414
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013495238915230376,
+      "loss": 1.2887,
+      "step": 8415
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013493779265365196,
+      "loss": 1.286,
+      "step": 8416
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013492319530707857,
+      "loss": 1.2321,
+      "step": 8417
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013490859711293794,
+      "loss": 1.2073,
+      "step": 8418
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013489399807158425,
+      "loss": 1.1312,
+      "step": 8419
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013487939818337185,
+      "loss": 1.0081,
+      "step": 8420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013486479744865515,
+      "loss": 1.1673,
+      "step": 8421
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013485019586778838,
+      "loss": 1.2367,
+      "step": 8422
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013483559344112597,
+      "loss": 1.1241,
+      "step": 8423
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013482099016902234,
+      "loss": 1.0659,
+      "step": 8424
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013480638605183187,
+      "loss": 1.2049,
+      "step": 8425
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.000134791781089909,
+      "loss": 1.178,
+      "step": 8426
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001347771752836082,
+      "loss": 1.2811,
+      "step": 8427
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013476256863328393,
+      "loss": 1.0323,
+      "step": 8428
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013474796113929066,
+      "loss": 1.1313,
+      "step": 8429
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013473335280198294,
+      "loss": 1.1,
+      "step": 8430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001347187436217153,
+      "loss": 1.0779,
+      "step": 8431
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001347041335988423,
+      "loss": 1.2264,
+      "step": 8432
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013468952273371855,
+      "loss": 1.3023,
+      "step": 8433
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013467491102669854,
+      "loss": 1.1488,
+      "step": 8434
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.000134660298478137,
+      "loss": 1.3845,
+      "step": 8435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001346456850883885,
+      "loss": 1.0386,
+      "step": 8436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013463107085780779,
+      "loss": 1.1226,
+      "step": 8437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001346164557867494,
+      "loss": 1.1911,
+      "step": 8438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013460183987556816,
+      "loss": 1.3104,
+      "step": 8439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013458722312461875,
+      "loss": 1.267,
+      "step": 8440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013457260553425591,
+      "loss": 1.0414,
+      "step": 8441
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013455798710483441,
+      "loss": 1.0673,
+      "step": 8442
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013454336783670902,
+      "loss": 1.1056,
+      "step": 8443
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013452874773023452,
+      "loss": 1.193,
+      "step": 8444
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013451412678576574,
+      "loss": 1.3412,
+      "step": 8445
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013449950500365757,
+      "loss": 1.1532,
+      "step": 8446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001344848823842648,
+      "loss": 1.041,
+      "step": 8447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013447025892794243,
+      "loss": 1.0259,
+      "step": 8448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013445563463504524,
+      "loss": 1.1039,
+      "step": 8449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001344410095059282,
+      "loss": 1.1653,
+      "step": 8450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013442638354094625,
+      "loss": 1.0198,
+      "step": 8451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013441175674045437,
+      "loss": 1.2357,
+      "step": 8452
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013439712910480751,
+      "loss": 1.244,
+      "step": 8453
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013438250063436074,
+      "loss": 1.0505,
+      "step": 8454
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.000134367871329469,
+      "loss": 1.231,
+      "step": 8455
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001343532411904874,
+      "loss": 1.0563,
+      "step": 8456
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.000134338610217771,
+      "loss": 1.1335,
+      "step": 8457
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013432397841167483,
+      "loss": 1.1559,
+      "step": 8458
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013430934577255408,
+      "loss": 1.0934,
+      "step": 8459
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001342947123007638,
+      "loss": 1.3126,
+      "step": 8460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013428007799665915,
+      "loss": 1.2076,
+      "step": 8461
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001342654428605954,
+      "loss": 1.2137,
+      "step": 8462
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013425080689292757,
+      "loss": 1.196,
+      "step": 8463
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013423617009401099,
+      "loss": 1.3732,
+      "step": 8464
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013422153246420084,
+      "loss": 1.0965,
+      "step": 8465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001342068940038524,
+      "loss": 1.3861,
+      "step": 8466
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013419225471332088,
+      "loss": 1.2541,
+      "step": 8467
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001341776145929616,
+      "loss": 1.1414,
+      "step": 8468
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013416297364312983,
+      "loss": 1.1371,
+      "step": 8469
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013414833186418102,
+      "loss": 1.2049,
+      "step": 8470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013413368925647038,
+      "loss": 1.2333,
+      "step": 8471
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00013411904582035335,
+      "loss": 1.2731,
+      "step": 8472
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013410440155618532,
+      "loss": 1.0414,
+      "step": 8473
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013408975646432163,
+      "loss": 1.1888,
+      "step": 8474
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001340751105451178,
+      "loss": 1.0785,
+      "step": 8475
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001340604637989292,
+      "loss": 1.067,
+      "step": 8476
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001340458162261114,
+      "loss": 1.0213,
+      "step": 8477
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013403116782701976,
+      "loss": 1.1579,
+      "step": 8478
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013401651860200987,
+      "loss": 1.2431,
+      "step": 8479
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013400186855143725,
+      "loss": 2.4565,
+      "step": 8480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013398721767565746,
+      "loss": 2.2178,
+      "step": 8481
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.000133972565975026,
+      "loss": 1.1866,
+      "step": 8482
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013395791344989857,
+      "loss": 1.3503,
+      "step": 8483
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013394326010063068,
+      "loss": 1.2665,
+      "step": 8484
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013392860592757798,
+      "loss": 1.1553,
+      "step": 8485
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013391395093109617,
+      "loss": 1.2133,
+      "step": 8486
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013389929511154088,
+      "loss": 1.1368,
+      "step": 8487
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013388463846926777,
+      "loss": 0.9123,
+      "step": 8488
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013386998100463262,
+      "loss": 1.0303,
+      "step": 8489
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001338553227179911,
+      "loss": 1.1807,
+      "step": 8490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.000133840663609699,
+      "loss": 1.2449,
+      "step": 8491
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013382600368011207,
+      "loss": 1.3246,
+      "step": 8492
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013381134292958608,
+      "loss": 2.2373,
+      "step": 8493
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013379668135847687,
+      "loss": 2.428,
+      "step": 8494
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001337820189671402,
+      "loss": 1.0954,
+      "step": 8495
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013376735575593203,
+      "loss": 1.098,
+      "step": 8496
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013375269172520818,
+      "loss": 0.9728,
+      "step": 8497
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013373802687532447,
+      "loss": 1.1101,
+      "step": 8498
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013372336120663693,
+      "loss": 1.2331,
+      "step": 8499
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013370869471950142,
+      "loss": 1.1224,
+      "step": 8500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013369402741427385,
+      "loss": 1.2042,
+      "step": 8501
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001336793592913103,
+      "loss": 1.1386,
+      "step": 8502
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001336646903509666,
+      "loss": 1.1398,
+      "step": 8503
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013365002059359887,
+      "loss": 1.1575,
+      "step": 8504
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013363535001956314,
+      "loss": 1.2447,
+      "step": 8505
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013362067862921547,
+      "loss": 1.3273,
+      "step": 8506
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001336060064229118,
+      "loss": 2.2502,
+      "step": 8507
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013359133340100837,
+      "loss": 0.971,
+      "step": 8508
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013357665956386115,
+      "loss": 1.0101,
+      "step": 8509
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001335619849118264,
+      "loss": 1.1407,
+      "step": 8510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013354730944526016,
+      "loss": 1.2236,
+      "step": 8511
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013353263316451865,
+      "loss": 1.0749,
+      "step": 8512
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013351795606995808,
+      "loss": 1.1897,
+      "step": 8513
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013350327816193458,
+      "loss": 1.3061,
+      "step": 8514
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013348859944080441,
+      "loss": 1.2309,
+      "step": 8515
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013347391990692383,
+      "loss": 1.1243,
+      "step": 8516
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001334592395606491,
+      "loss": 1.2923,
+      "step": 8517
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013344455840233647,
+      "loss": 1.2256,
+      "step": 8518
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001334298764323423,
+      "loss": 1.1208,
+      "step": 8519
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013341519365102285,
+      "loss": 1.1924,
+      "step": 8520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013340051005873452,
+      "loss": 0.9749,
+      "step": 8521
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013338582565583369,
+      "loss": 1.1889,
+      "step": 8522
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013337114044267663,
+      "loss": 1.0883,
+      "step": 8523
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013335645441961985,
+      "loss": 1.1885,
+      "step": 8524
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013334176758701975,
+      "loss": 1.2432,
+      "step": 8525
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013332707994523273,
+      "loss": 2.4998,
+      "step": 8526
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001333123914946153,
+      "loss": 1.2186,
+      "step": 8527
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001332977022355239,
+      "loss": 1.0918,
+      "step": 8528
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013328301216831507,
+      "loss": 1.1937,
+      "step": 8529
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013326832129334533,
+      "loss": 0.9491,
+      "step": 8530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013325362961097117,
+      "loss": 1.1445,
+      "step": 8531
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013323893712154918,
+      "loss": 1.3135,
+      "step": 8532
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013322424382543595,
+      "loss": 1.1236,
+      "step": 8533
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013320954972298804,
+      "loss": 1.2698,
+      "step": 8534
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013319485481456212,
+      "loss": 0.929,
+      "step": 8535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013318015910051482,
+      "loss": 1.2779,
+      "step": 8536
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013316546258120276,
+      "loss": 1.257,
+      "step": 8537
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013315076525698266,
+      "loss": 2.343,
+      "step": 8538
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013313606712821117,
+      "loss": 1.0396,
+      "step": 8539
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013312136819524503,
+      "loss": 1.1517,
+      "step": 8540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013310666845844094,
+      "loss": 2.207,
+      "step": 8541
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013309196791815576,
+      "loss": 1.2025,
+      "step": 8542
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013307726657474616,
+      "loss": 1.2008,
+      "step": 8543
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013306256442856897,
+      "loss": 1.166,
+      "step": 8544
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013304786147998098,
+      "loss": 1.0448,
+      "step": 8545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013303315772933907,
+      "loss": 1.2572,
+      "step": 8546
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013301845317700004,
+      "loss": 1.1582,
+      "step": 8547
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013300374782332082,
+      "loss": 1.1579,
+      "step": 8548
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001329890416686582,
+      "loss": 2.0564,
+      "step": 8549
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013297433471336922,
+      "loss": 1.2191,
+      "step": 8550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013295962695781077,
+      "loss": 1.0068,
+      "step": 8551
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013294491840233974,
+      "loss": 1.105,
+      "step": 8552
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013293020904731311,
+      "loss": 1.2411,
+      "step": 8553
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013291549889308794,
+      "loss": 1.2556,
+      "step": 8554
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013290078794002118,
+      "loss": 2.1338,
+      "step": 8555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013288607618846987,
+      "loss": 2.1887,
+      "step": 8556
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013287136363879102,
+      "loss": 1.0868,
+      "step": 8557
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001328566502913418,
+      "loss": 1.0788,
+      "step": 8558
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013284193614647918,
+      "loss": 1.1504,
+      "step": 8559
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001328272212045603,
+      "loss": 1.1967,
+      "step": 8560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001328125054659423,
+      "loss": 1.0566,
+      "step": 8561
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013279778893098235,
+      "loss": 1.194,
+      "step": 8562
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013278307160003756,
+      "loss": 1.1091,
+      "step": 8563
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001327683534734651,
+      "loss": 1.1589,
+      "step": 8564
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001327536345516223,
+      "loss": 1.161,
+      "step": 8565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001327389148348662,
+      "loss": 1.2333,
+      "step": 8566
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013272419432355415,
+      "loss": 1.0594,
+      "step": 8567
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013270947301804336,
+      "loss": 1.2643,
+      "step": 8568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013269475091869112,
+      "loss": 1.1268,
+      "step": 8569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013268002802585479,
+      "loss": 0.9762,
+      "step": 8570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001326653043398916,
+      "loss": 1.2108,
+      "step": 8571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013265057986115893,
+      "loss": 1.0772,
+      "step": 8572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013263585459001413,
+      "loss": 1.1524,
+      "step": 8573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013262112852681455,
+      "loss": 1.0573,
+      "step": 8574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001326064016719176,
+      "loss": 1.0444,
+      "step": 8575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00013259167402568075,
+      "loss": 1.0309,
+      "step": 8576
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001325769455884613,
+      "loss": 1.1954,
+      "step": 8577
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013256221636061684,
+      "loss": 1.1451,
+      "step": 8578
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013254748634250474,
+      "loss": 1.0465,
+      "step": 8579
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013253275553448256,
+      "loss": 1.1188,
+      "step": 8580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001325180239369078,
+      "loss": 1.0358,
+      "step": 8581
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013250329155013792,
+      "loss": 1.0156,
+      "step": 8582
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001324885583745305,
+      "loss": 1.0178,
+      "step": 8583
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013247382441044318,
+      "loss": 1.0683,
+      "step": 8584
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013245908965823345,
+      "loss": 1.1584,
+      "step": 8585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013244435411825898,
+      "loss": 1.1354,
+      "step": 8586
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013242961779087734,
+      "loss": 1.1996,
+      "step": 8587
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013241488067644618,
+      "loss": 1.2839,
+      "step": 8588
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013240014277532321,
+      "loss": 1.0287,
+      "step": 8589
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001323854040878661,
+      "loss": 1.029,
+      "step": 8590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013237066461443245,
+      "loss": 1.3046,
+      "step": 8591
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013235592435538012,
+      "loss": 1.1075,
+      "step": 8592
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013234118331106678,
+      "loss": 1.147,
+      "step": 8593
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013232644148185018,
+      "loss": 2.3123,
+      "step": 8594
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013231169886808813,
+      "loss": 1.0975,
+      "step": 8595
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001322969554701384,
+      "loss": 1.2001,
+      "step": 8596
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013228221128835883,
+      "loss": 1.1116,
+      "step": 8597
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001322674663231072,
+      "loss": 0.9859,
+      "step": 8598
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001322527205747414,
+      "loss": 1.0643,
+      "step": 8599
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001322379740436193,
+      "loss": 1.2475,
+      "step": 8600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001322232267300988,
+      "loss": 1.1395,
+      "step": 8601
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001322084786345378,
+      "loss": 2.1379,
+      "step": 8602
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013219372975729424,
+      "loss": 1.1863,
+      "step": 8603
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.000132178980098726,
+      "loss": 1.0702,
+      "step": 8604
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013216422965919115,
+      "loss": 1.0574,
+      "step": 8605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001321494784390476,
+      "loss": 1.1943,
+      "step": 8606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013213472643865338,
+      "loss": 1.1855,
+      "step": 8607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013211997365836654,
+      "loss": 2.3235,
+      "step": 8608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013210522009854506,
+      "loss": 1.1555,
+      "step": 8609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013209046575954708,
+      "loss": 0.9995,
+      "step": 8610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001320757106417306,
+      "loss": 1.2342,
+      "step": 8611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013206095474545374,
+      "loss": 1.1306,
+      "step": 8612
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013204619807107468,
+      "loss": 1.0733,
+      "step": 8613
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013203144061895146,
+      "loss": 1.0948,
+      "step": 8614
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013201668238944228,
+      "loss": 2.1917,
+      "step": 8615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013200192338290537,
+      "loss": 1.0428,
+      "step": 8616
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013198716359969886,
+      "loss": 1.156,
+      "step": 8617
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013197240304018096,
+      "loss": 1.2151,
+      "step": 8618
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013195764170470992,
+      "loss": 1.0565,
+      "step": 8619
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013194287959364395,
+      "loss": 1.1719,
+      "step": 8620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001319281167073414,
+      "loss": 1.1207,
+      "step": 8621
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013191335304616047,
+      "loss": 1.1769,
+      "step": 8622
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013189858861045953,
+      "loss": 1.0697,
+      "step": 8623
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013188382340059693,
+      "loss": 1.0099,
+      "step": 8624
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001318690574169309,
+      "loss": 1.1923,
+      "step": 8625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013185429065981988,
+      "loss": 1.041,
+      "step": 8626
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013183952312962224,
+      "loss": 1.1744,
+      "step": 8627
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013182475482669638,
+      "loss": 1.1275,
+      "step": 8628
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013180998575140074,
+      "loss": 1.0396,
+      "step": 8629
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013179521590409368,
+      "loss": 1.1575,
+      "step": 8630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001317804452851338,
+      "loss": 1.0319,
+      "step": 8631
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001317656738948794,
+      "loss": 1.1304,
+      "step": 8632
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001317509017336891,
+      "loss": 1.1134,
+      "step": 8633
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013173612880192136,
+      "loss": 0.9416,
+      "step": 8634
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013172135509993476,
+      "loss": 1.0306,
+      "step": 8635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013170658062808778,
+      "loss": 1.1035,
+      "step": 8636
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013169180538673904,
+      "loss": 1.003,
+      "step": 8637
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001316770293762471,
+      "loss": 1.0598,
+      "step": 8638
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013166225259697058,
+      "loss": 0.9412,
+      "step": 8639
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001316474750492681,
+      "loss": 1.2626,
+      "step": 8640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001316326967334983,
+      "loss": 1.0276,
+      "step": 8641
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013161791765001982,
+      "loss": 1.1737,
+      "step": 8642
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013160313779919143,
+      "loss": 1.175,
+      "step": 8643
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001315883571813717,
+      "loss": 2.1399,
+      "step": 8644
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013157357579691946,
+      "loss": 1.0455,
+      "step": 8645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013155879364619338,
+      "loss": 1.2931,
+      "step": 8646
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013154401072955222,
+      "loss": 0.9495,
+      "step": 8647
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001315292270473548,
+      "loss": 1.0921,
+      "step": 8648
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013151444259995983,
+      "loss": 1.133,
+      "step": 8649
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013149965738772621,
+      "loss": 1.2014,
+      "step": 8650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013148487141101273,
+      "loss": 1.3043,
+      "step": 8651
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013147008467017822,
+      "loss": 1.1314,
+      "step": 8652
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013145529716558158,
+      "loss": 1.1749,
+      "step": 8653
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013144050889758168,
+      "loss": 1.2224,
+      "step": 8654
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001314257198665374,
+      "loss": 2.3401,
+      "step": 8655
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001314109300728077,
+      "loss": 1.0255,
+      "step": 8656
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001313961395167515,
+      "loss": 1.1934,
+      "step": 8657
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013138134819872774,
+      "loss": 0.9684,
+      "step": 8658
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013136655611909546,
+      "loss": 1.1852,
+      "step": 8659
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013135176327821357,
+      "loss": 1.1199,
+      "step": 8660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013133696967644118,
+      "loss": 1.1148,
+      "step": 8661
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013132217531413725,
+      "loss": 1.2125,
+      "step": 8662
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013130738019166084,
+      "loss": 0.9781,
+      "step": 8663
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013129258430937105,
+      "loss": 0.9,
+      "step": 8664
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013127778766762696,
+      "loss": 1.0006,
+      "step": 8665
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013126299026678764,
+      "loss": 1.1812,
+      "step": 8666
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001312481921072123,
+      "loss": 1.0621,
+      "step": 8667
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013123339318925998,
+      "loss": 1.1299,
+      "step": 8668
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001312185935132899,
+      "loss": 1.3018,
+      "step": 8669
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013120379307966125,
+      "loss": 1.0106,
+      "step": 8670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013118899188873314,
+      "loss": 1.3257,
+      "step": 8671
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001311741899408649,
+      "loss": 1.3089,
+      "step": 8672
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013115938723641575,
+      "loss": 1.075,
+      "step": 8673
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013114458377574487,
+      "loss": 1.0605,
+      "step": 8674
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001311297795592116,
+      "loss": 1.052,
+      "step": 8675
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013111497458717522,
+      "loss": 1.1884,
+      "step": 8676
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013110016885999499,
+      "loss": 0.9558,
+      "step": 8677
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013108536237803028,
+      "loss": 1.1685,
+      "step": 8678
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00013107055514164045,
+      "loss": 1.0598,
+      "step": 8679
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001310557471511848,
+      "loss": 1.127,
+      "step": 8680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013104093840702282,
+      "loss": 1.2729,
+      "step": 8681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013102612890951378,
+      "loss": 1.1765,
+      "step": 8682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001310113186590172,
+      "loss": 1.0561,
+      "step": 8683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013099650765589245,
+      "loss": 1.17,
+      "step": 8684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013098169590049905,
+      "loss": 2.4395,
+      "step": 8685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013096688339319642,
+      "loss": 1.0848,
+      "step": 8686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013095207013434407,
+      "loss": 1.0852,
+      "step": 8687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001309372561243015,
+      "loss": 1.0822,
+      "step": 8688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013092244136342826,
+      "loss": 1.2598,
+      "step": 8689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001309076258520839,
+      "loss": 1.2505,
+      "step": 8690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013089280959062793,
+      "loss": 1.0349,
+      "step": 8691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013087799257942,
+      "loss": 1.1967,
+      "step": 8692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013086317481881966,
+      "loss": 1.2131,
+      "step": 8693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013084835630918656,
+      "loss": 1.0866,
+      "step": 8694
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013083353705088032,
+      "loss": 1.2071,
+      "step": 8695
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013081871704426063,
+      "loss": 1.2669,
+      "step": 8696
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001308038962896871,
+      "loss": 2.1521,
+      "step": 8697
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013078907478751952,
+      "loss": 1.1818,
+      "step": 8698
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013077425253811748,
+      "loss": 1.1635,
+      "step": 8699
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013075942954184083,
+      "loss": 0.9546,
+      "step": 8700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013074460579904918,
+      "loss": 1.1161,
+      "step": 8701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013072978131010242,
+      "loss": 1.0979,
+      "step": 8702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013071495607536027,
+      "loss": 1.0956,
+      "step": 8703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001307001300951825,
+      "loss": 1.1098,
+      "step": 8704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013068530336992904,
+      "loss": 0.9795,
+      "step": 8705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001306704758999596,
+      "loss": 1.226,
+      "step": 8706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013065564768563413,
+      "loss": 1.1193,
+      "step": 8707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013064081872731244,
+      "loss": 1.2193,
+      "step": 8708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013062598902535444,
+      "loss": 1.0339,
+      "step": 8709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013061115858012005,
+      "loss": 1.1752,
+      "step": 8710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013059632739196922,
+      "loss": 1.1739,
+      "step": 8711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001305814954612618,
+      "loss": 1.0998,
+      "step": 8712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013056666278835787,
+      "loss": 1.1199,
+      "step": 8713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013055182937361735,
+      "loss": 1.112,
+      "step": 8714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013053699521740027,
+      "loss": 1.0906,
+      "step": 8715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001305221603200666,
+      "loss": 1.0807,
+      "step": 8716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001305073246819764,
+      "loss": 1.0539,
+      "step": 8717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001304924883034897,
+      "loss": 1.1998,
+      "step": 8718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013047765118496664,
+      "loss": 1.1221,
+      "step": 8719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013046281332676726,
+      "loss": 0.8862,
+      "step": 8720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013044797472925166,
+      "loss": 1.094,
+      "step": 8721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013043313539278,
+      "loss": 0.9412,
+      "step": 8722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013041829531771235,
+      "loss": 0.9383,
+      "step": 8723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013040345450440895,
+      "loss": 1.0454,
+      "step": 8724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013038861295322992,
+      "loss": 0.9837,
+      "step": 8725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001303737706645355,
+      "loss": 1.1751,
+      "step": 8726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013035892763868593,
+      "loss": 2.2488,
+      "step": 8727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013034408387604135,
+      "loss": 1.1184,
+      "step": 8728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013032923937696206,
+      "loss": 2.2119,
+      "step": 8729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013031439414180838,
+      "loss": 1.0684,
+      "step": 8730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013029954817094048,
+      "loss": 1.1591,
+      "step": 8731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001302847014647188,
+      "loss": 1.1833,
+      "step": 8732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013026985402350352,
+      "loss": 1.0696,
+      "step": 8733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001302550058476551,
+      "loss": 1.2866,
+      "step": 8734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013024015693753382,
+      "loss": 1.2078,
+      "step": 8735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001302253072935001,
+      "loss": 1.0251,
+      "step": 8736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001302104569159143,
+      "loss": 0.9477,
+      "step": 8737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013019560580513685,
+      "loss": 1.149,
+      "step": 8738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013018075396152817,
+      "loss": 2.302,
+      "step": 8739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013016590138544874,
+      "loss": 1.0858,
+      "step": 8740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013015104807725892,
+      "loss": 1.1072,
+      "step": 8741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013013619403731935,
+      "loss": 1.3165,
+      "step": 8742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013012133926599043,
+      "loss": 1.3265,
+      "step": 8743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013010648376363265,
+      "loss": 1.1539,
+      "step": 8744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013009162753060662,
+      "loss": 1.1681,
+      "step": 8745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013007677056727288,
+      "loss": 1.2958,
+      "step": 8746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013006191287399195,
+      "loss": 1.0646,
+      "step": 8747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001300470544511245,
+      "loss": 1.0951,
+      "step": 8748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013003219529903105,
+      "loss": 1.0113,
+      "step": 8749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013001733541807224,
+      "loss": 1.2611,
+      "step": 8750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00013000247480860876,
+      "loss": 1.1643,
+      "step": 8751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012998761347100124,
+      "loss": 1.1255,
+      "step": 8752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012997275140561035,
+      "loss": 1.1956,
+      "step": 8753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012995788861279678,
+      "loss": 1.1176,
+      "step": 8754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012994302509292127,
+      "loss": 1.0801,
+      "step": 8755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012992816084634454,
+      "loss": 1.2572,
+      "step": 8756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012991329587342732,
+      "loss": 1.0231,
+      "step": 8757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001298984301745304,
+      "loss": 1.1713,
+      "step": 8758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012988356375001454,
+      "loss": 2.2766,
+      "step": 8759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012986869660024051,
+      "loss": 1.1092,
+      "step": 8760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001298538287255692,
+      "loss": 0.8351,
+      "step": 8761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012983896012636143,
+      "loss": 1.1905,
+      "step": 8762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012982409080297802,
+      "loss": 1.1056,
+      "step": 8763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012980922075577986,
+      "loss": 1.1554,
+      "step": 8764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012979434998512782,
+      "loss": 2.1975,
+      "step": 8765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012977947849138283,
+      "loss": 1.1958,
+      "step": 8766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001297646062749058,
+      "loss": 1.1539,
+      "step": 8767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012974973333605767,
+      "loss": 1.1542,
+      "step": 8768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001297348596751994,
+      "loss": 1.0647,
+      "step": 8769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012971998529269198,
+      "loss": 1.1932,
+      "step": 8770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001297051101888964,
+      "loss": 1.126,
+      "step": 8771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012969023436417362,
+      "loss": 1.0683,
+      "step": 8772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012967535781888473,
+      "loss": 1.203,
+      "step": 8773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012966048055339078,
+      "loss": 1.2441,
+      "step": 8774
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012964560256805278,
+      "loss": 1.1904,
+      "step": 8775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012963072386323185,
+      "loss": 1.0368,
+      "step": 8776
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012961584443928905,
+      "loss": 0.9513,
+      "step": 8777
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012960096429658557,
+      "loss": 1.041,
+      "step": 8778
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012958608343548246,
+      "loss": 1.0664,
+      "step": 8779
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012957120185634092,
+      "loss": 1.08,
+      "step": 8780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012955631955952212,
+      "loss": 2.1313,
+      "step": 8781
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001295414365453872,
+      "loss": 1.3146,
+      "step": 8782
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001295265528142974,
+      "loss": 2.3413,
+      "step": 8783
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00012951166836661397,
+      "loss": 1.1505,
+      "step": 8784
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012949678320269806,
+      "loss": 1.2413,
+      "step": 8785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012948189732291102,
+      "loss": 1.1947,
+      "step": 8786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012946701072761408,
+      "loss": 0.9544,
+      "step": 8787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012945212341716847,
+      "loss": 1.0554,
+      "step": 8788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001294372353919356,
+      "loss": 1.1496,
+      "step": 8789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012942234665227676,
+      "loss": 1.0386,
+      "step": 8790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012940745719855324,
+      "loss": 2.0444,
+      "step": 8791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001293925670311265,
+      "loss": 1.0636,
+      "step": 8792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001293776761503578,
+      "loss": 1.0592,
+      "step": 8793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012936278455660862,
+      "loss": 1.2188,
+      "step": 8794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012934789225024033,
+      "loss": 2.1997,
+      "step": 8795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012933299923161438,
+      "loss": 1.0493,
+      "step": 8796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001293181055010922,
+      "loss": 1.1156,
+      "step": 8797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012930321105903526,
+      "loss": 1.0218,
+      "step": 8798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012928831590580501,
+      "loss": 1.048,
+      "step": 8799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012927342004176304,
+      "loss": 0.9987,
+      "step": 8800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012925852346727074,
+      "loss": 1.2462,
+      "step": 8801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001292436261826897,
+      "loss": 1.1449,
+      "step": 8802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012922872818838153,
+      "loss": 1.0352,
+      "step": 8803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012921382948470768,
+      "loss": 1.2164,
+      "step": 8804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012919893007202983,
+      "loss": 0.8812,
+      "step": 8805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012918402995070952,
+      "loss": 1.1647,
+      "step": 8806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001291691291211084,
+      "loss": 1.0622,
+      "step": 8807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001291542275835881,
+      "loss": 1.1951,
+      "step": 8808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012913932533851024,
+      "loss": 1.0033,
+      "step": 8809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012912442238623655,
+      "loss": 1.1758,
+      "step": 8810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001291095187271287,
+      "loss": 0.8804,
+      "step": 8811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012909461436154832,
+      "loss": 1.0423,
+      "step": 8812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012907970928985725,
+      "loss": 1.0677,
+      "step": 8813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001290648035124171,
+      "loss": 1.178,
+      "step": 8814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012904989702958977,
+      "loss": 1.1434,
+      "step": 8815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001290349898417369,
+      "loss": 1.0742,
+      "step": 8816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012902008194922034,
+      "loss": 0.8834,
+      "step": 8817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001290051733524019,
+      "loss": 1.2224,
+      "step": 8818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001289902640516434,
+      "loss": 1.1041,
+      "step": 8819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012897535404730667,
+      "loss": 1.2345,
+      "step": 8820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001289604433397536,
+      "loss": 1.2407,
+      "step": 8821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.000128945531929346,
+      "loss": 1.2363,
+      "step": 8822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001289306198164458,
+      "loss": 1.1577,
+      "step": 8823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012891570700141494,
+      "loss": 1.0968,
+      "step": 8824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012890079348461527,
+      "loss": 0.9957,
+      "step": 8825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001288858792664088,
+      "loss": 1.2831,
+      "step": 8826
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001288709643471575,
+      "loss": 1.12,
+      "step": 8827
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012885604872722327,
+      "loss": 1.0664,
+      "step": 8828
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012884113240696818,
+      "loss": 1.1638,
+      "step": 8829
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001288262153867542,
+      "loss": 1.0978,
+      "step": 8830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012881129766694336,
+      "loss": 2.2827,
+      "step": 8831
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012879637924789774,
+      "loss": 1.0589,
+      "step": 8832
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012878146012997932,
+      "loss": 0.9822,
+      "step": 8833
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001287665403135503,
+      "loss": 1.0854,
+      "step": 8834
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012875161979897267,
+      "loss": 1.1732,
+      "step": 8835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012873669858660857,
+      "loss": 1.0575,
+      "step": 8836
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001287217766768202,
+      "loss": 1.154,
+      "step": 8837
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012870685406996964,
+      "loss": 1.2669,
+      "step": 8838
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012869193076641902,
+      "loss": 1.1221,
+      "step": 8839
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012867700676653062,
+      "loss": 1.1237,
+      "step": 8840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012866208207066652,
+      "loss": 1.1744,
+      "step": 8841
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.000128647156679189,
+      "loss": 1.2064,
+      "step": 8842
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012863223059246035,
+      "loss": 1.2277,
+      "step": 8843
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012861730381084273,
+      "loss": 0.9362,
+      "step": 8844
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012860237633469842,
+      "loss": 1.145,
+      "step": 8845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012858744816438973,
+      "loss": 1.1452,
+      "step": 8846
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012857251930027891,
+      "loss": 1.0422,
+      "step": 8847
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012855758974272832,
+      "loss": 1.1466,
+      "step": 8848
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001285426594921003,
+      "loss": 1.1111,
+      "step": 8849
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012852772854875716,
+      "loss": 1.0334,
+      "step": 8850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001285127969130613,
+      "loss": 1.2552,
+      "step": 8851
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012849786458537507,
+      "loss": 1.0937,
+      "step": 8852
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012848293156606088,
+      "loss": 1.1044,
+      "step": 8853
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001284679978554812,
+      "loss": 1.1205,
+      "step": 8854
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012845306345399835,
+      "loss": 1.11,
+      "step": 8855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001284381283619749,
+      "loss": 1.2402,
+      "step": 8856
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012842319257977321,
+      "loss": 1.0854,
+      "step": 8857
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012840825610775587,
+      "loss": 1.0511,
+      "step": 8858
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012839331894628532,
+      "loss": 1.0648,
+      "step": 8859
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012837838109572405,
+      "loss": 1.2454,
+      "step": 8860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012836344255643466,
+      "loss": 1.0516,
+      "step": 8861
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012834850332877967,
+      "loss": 1.1521,
+      "step": 8862
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012833356341312164,
+      "loss": 1.1826,
+      "step": 8863
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012831862280982315,
+      "loss": 1.0246,
+      "step": 8864
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012830368151924682,
+      "loss": 0.9621,
+      "step": 8865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012828873954175523,
+      "loss": 2.353,
+      "step": 8866
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012827379687771107,
+      "loss": 1.066,
+      "step": 8867
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012825885352747696,
+      "loss": 1.1124,
+      "step": 8868
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012824390949141553,
+      "loss": 1.2026,
+      "step": 8869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012822896476988958,
+      "loss": 1.0979,
+      "step": 8870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012821401936326165,
+      "loss": 1.076,
+      "step": 8871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012819907327189458,
+      "loss": 1.1118,
+      "step": 8872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012818412649615106,
+      "loss": 1.2521,
+      "step": 8873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012816917903639384,
+      "loss": 1.1257,
+      "step": 8874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012815423089298567,
+      "loss": 1.0561,
+      "step": 8875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001281392820662894,
+      "loss": 2.0808,
+      "step": 8876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012812433255666775,
+      "loss": 0.938,
+      "step": 8877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012810938236448358,
+      "loss": 1.2735,
+      "step": 8878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001280944314900997,
+      "loss": 0.9879,
+      "step": 8879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012807947993387895,
+      "loss": 1.0248,
+      "step": 8880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001280645276961843,
+      "loss": 1.1853,
+      "step": 8881
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012804957477737848,
+      "loss": 1.0317,
+      "step": 8882
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001280346211778245,
+      "loss": 0.9531,
+      "step": 8883
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012801966689788522,
+      "loss": 0.9828,
+      "step": 8884
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001280047119379236,
+      "loss": 1.1481,
+      "step": 8885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012798975629830256,
+      "loss": 1.162,
+      "step": 8886
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001279747999793851,
+      "loss": 1.2355,
+      "step": 8887
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00012795984298153416,
+      "loss": 1.3192,
+      "step": 8888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012794488530511284,
+      "loss": 1.0764,
+      "step": 8889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.000127929926950484,
+      "loss": 1.001,
+      "step": 8890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012791496791801076,
+      "loss": 1.2273,
+      "step": 8891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001279000082080562,
+      "loss": 1.2013,
+      "step": 8892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012788504782098328,
+      "loss": 1.0649,
+      "step": 8893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001278700867571552,
+      "loss": 1.0995,
+      "step": 8894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012785512501693497,
+      "loss": 1.0847,
+      "step": 8895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012784016260068576,
+      "loss": 1.0719,
+      "step": 8896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012782519950877067,
+      "loss": 1.0579,
+      "step": 8897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001278102357415528,
+      "loss": 1.1797,
+      "step": 8898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001277952712993954,
+      "loss": 1.1575,
+      "step": 8899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012778030618266164,
+      "loss": 1.1178,
+      "step": 8900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012776534039171463,
+      "loss": 1.1251,
+      "step": 8901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012775037392691768,
+      "loss": 1.1022,
+      "step": 8902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012773540678863397,
+      "loss": 1.0239,
+      "step": 8903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012772043897722672,
+      "loss": 1.1988,
+      "step": 8904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012770547049305926,
+      "loss": 0.9609,
+      "step": 8905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001276905013364948,
+      "loss": 2.1282,
+      "step": 8906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012767553150789667,
+      "loss": 0.9543,
+      "step": 8907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012766056100762816,
+      "loss": 1.1743,
+      "step": 8908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012764558983605263,
+      "loss": 1.1922,
+      "step": 8909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012763061799353343,
+      "loss": 1.1954,
+      "step": 8910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012761564548043386,
+      "loss": 1.1922,
+      "step": 8911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001276006722971173,
+      "loss": 1.0525,
+      "step": 8912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001275856984439472,
+      "loss": 0.9504,
+      "step": 8913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012757072392128688,
+      "loss": 1.0729,
+      "step": 8914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012755574872949985,
+      "loss": 1.2683,
+      "step": 8915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012754077286894952,
+      "loss": 0.9627,
+      "step": 8916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012752579633999934,
+      "loss": 0.9683,
+      "step": 8917
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012751081914301278,
+      "loss": 1.2086,
+      "step": 8918
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012749584127835333,
+      "loss": 1.1324,
+      "step": 8919
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001274808627463845,
+      "loss": 1.2814,
+      "step": 8920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001274658835474698,
+      "loss": 1.0862,
+      "step": 8921
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012745090368197282,
+      "loss": 2.177,
+      "step": 8922
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012743592315025702,
+      "loss": 1.08,
+      "step": 8923
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012742094195268604,
+      "loss": 1.2022,
+      "step": 8924
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012740596008962345,
+      "loss": 1.2119,
+      "step": 8925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012739097756143283,
+      "loss": 1.1166,
+      "step": 8926
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012737599436847782,
+      "loss": 1.0775,
+      "step": 8927
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012736101051112205,
+      "loss": 1.1942,
+      "step": 8928
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012734602598972916,
+      "loss": 1.0172,
+      "step": 8929
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012733104080466285,
+      "loss": 1.1148,
+      "step": 8930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012731605495628674,
+      "loss": 1.0939,
+      "step": 8931
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012730106844496461,
+      "loss": 1.1491,
+      "step": 8932
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012728608127106009,
+      "loss": 1.0487,
+      "step": 8933
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012727109343493696,
+      "loss": 1.1343,
+      "step": 8934
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.000127256104936959,
+      "loss": 0.9936,
+      "step": 8935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012724111577748986,
+      "loss": 1.1941,
+      "step": 8936
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012722612595689344,
+      "loss": 1.1709,
+      "step": 8937
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012721113547553347,
+      "loss": 1.0378,
+      "step": 8938
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012719614433377377,
+      "loss": 2.2454,
+      "step": 8939
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001271811525319782,
+      "loss": 2.1458,
+      "step": 8940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012716616007051054,
+      "loss": 1.2678,
+      "step": 8941
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001271511669497347,
+      "loss": 1.116,
+      "step": 8942
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012713617317001457,
+      "loss": 1.1662,
+      "step": 8943
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012712117873171395,
+      "loss": 1.132,
+      "step": 8944
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012710618363519682,
+      "loss": 1.9589,
+      "step": 8945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012709118788082713,
+      "loss": 2.0845,
+      "step": 8946
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012707619146896877,
+      "loss": 1.0852,
+      "step": 8947
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012706119439998567,
+      "loss": 1.0361,
+      "step": 8948
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012704619667424188,
+      "loss": 1.1173,
+      "step": 8949
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001270311982921013,
+      "loss": 1.2011,
+      "step": 8950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.000127016199253928,
+      "loss": 0.9991,
+      "step": 8951
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012700119956008597,
+      "loss": 1.1449,
+      "step": 8952
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012698619921093921,
+      "loss": 0.8776,
+      "step": 8953
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001269711982068519,
+      "loss": 2.1841,
+      "step": 8954
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012695619654818794,
+      "loss": 1.1377,
+      "step": 8955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001269411942353115,
+      "loss": 1.0387,
+      "step": 8956
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001269261912685867,
+      "loss": 1.1108,
+      "step": 8957
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012691118764837758,
+      "loss": 0.9527,
+      "step": 8958
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001268961833750483,
+      "loss": 1.2483,
+      "step": 8959
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012688117844896307,
+      "loss": 1.2712,
+      "step": 8960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001268661728704859,
+      "loss": 1.2014,
+      "step": 8961
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012685116663998117,
+      "loss": 1.0519,
+      "step": 8962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012683615975781293,
+      "loss": 1.2052,
+      "step": 8963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001268211522243454,
+      "loss": 1.1564,
+      "step": 8964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012680614403994287,
+      "loss": 1.1651,
+      "step": 8965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001267911352049695,
+      "loss": 1.1086,
+      "step": 8966
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012677612571978963,
+      "loss": 1.1698,
+      "step": 8967
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012676111558476745,
+      "loss": 0.9908,
+      "step": 8968
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012674610480026732,
+      "loss": 1.1666,
+      "step": 8969
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001267310933666535,
+      "loss": 1.0742,
+      "step": 8970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012671608128429033,
+      "loss": 1.2196,
+      "step": 8971
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001267010685535421,
+      "loss": 1.1743,
+      "step": 8972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012668605517477324,
+      "loss": 1.1777,
+      "step": 8973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012667104114834804,
+      "loss": 0.9469,
+      "step": 8974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012665602647463095,
+      "loss": 2.2258,
+      "step": 8975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012664101115398633,
+      "loss": 1.1446,
+      "step": 8976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012662599518677858,
+      "loss": 1.1115,
+      "step": 8977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012661097857337216,
+      "loss": 1.0619,
+      "step": 8978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001265959613141315,
+      "loss": 1.0778,
+      "step": 8979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012658094340942103,
+      "loss": 1.0703,
+      "step": 8980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012656592485960533,
+      "loss": 1.1518,
+      "step": 8981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012655090566504877,
+      "loss": 1.067,
+      "step": 8982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012653588582611593,
+      "loss": 1.3477,
+      "step": 8983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001265208653431713,
+      "loss": 1.1353,
+      "step": 8984
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012650584421657944,
+      "loss": 1.0661,
+      "step": 8985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012649082244670491,
+      "loss": 1.0489,
+      "step": 8986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012647580003391227,
+      "loss": 1.0186,
+      "step": 8987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001264607769785661,
+      "loss": 1.0984,
+      "step": 8988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.000126445753281031,
+      "loss": 0.891,
+      "step": 8989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00012643072894167156,
+      "loss": 1.2115,
+      "step": 8990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001264157039608525,
+      "loss": 1.1009,
+      "step": 8991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001264006783389384,
+      "loss": 1.1118,
+      "step": 8992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001263856520762939,
+      "loss": 2.4048,
+      "step": 8993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012637062517328372,
+      "loss": 1.0744,
+      "step": 8994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012635559763027261,
+      "loss": 1.0369,
+      "step": 8995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001263405694476252,
+      "loss": 1.1135,
+      "step": 8996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001263255406257062,
+      "loss": 0.9688,
+      "step": 8997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012631051116488042,
+      "loss": 0.9539,
+      "step": 8998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012629548106551258,
+      "loss": 1.0225,
+      "step": 8999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012628045032796748,
+      "loss": 1.0698,
+      "step": 9000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012626541895260985,
+      "loss": 1.1306,
+      "step": 9001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012625038693980453,
+      "loss": 1.1952,
+      "step": 9002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001262353542899164,
+      "loss": 1.1215,
+      "step": 9003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012622032100331016,
+      "loss": 1.0416,
+      "step": 9004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012620528708035079,
+      "loss": 0.9796,
+      "step": 9005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012619025252140308,
+      "loss": 2.2146,
+      "step": 9006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001261752173268319,
+      "loss": 1.0641,
+      "step": 9007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001261601814970022,
+      "loss": 1.0605,
+      "step": 9008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001261451450322789,
+      "loss": 1.1361,
+      "step": 9009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012613010793302681,
+      "loss": 1.2748,
+      "step": 9010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012611507019961104,
+      "loss": 1.1157,
+      "step": 9011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012610003183239642,
+      "loss": 1.2195,
+      "step": 9012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000126084992831748,
+      "loss": 1.0479,
+      "step": 9013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001260699531980307,
+      "loss": 1.1389,
+      "step": 9014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012605491293160958,
+      "loss": 1.1469,
+      "step": 9015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012603987203284965,
+      "loss": 1.2127,
+      "step": 9016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012602483050211592,
+      "loss": 1.0841,
+      "step": 9017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012600978833977343,
+      "loss": 1.2312,
+      "step": 9018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001259947455461873,
+      "loss": 1.1425,
+      "step": 9019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012597970212172257,
+      "loss": 1.1258,
+      "step": 9020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012596465806674436,
+      "loss": 2.269,
+      "step": 9021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012594961338161778,
+      "loss": 1.09,
+      "step": 9022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001259345680667079,
+      "loss": 1.1283,
+      "step": 9023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012591952212237992,
+      "loss": 1.162,
+      "step": 9024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012590447554899895,
+      "loss": 1.0495,
+      "step": 9025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012588942834693023,
+      "loss": 1.2019,
+      "step": 9026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012587438051653893,
+      "loss": 1.229,
+      "step": 9027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012585933205819022,
+      "loss": 1.136,
+      "step": 9028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001258442829722493,
+      "loss": 1.153,
+      "step": 9029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001258292332590815,
+      "loss": 1.1103,
+      "step": 9030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012581418291905195,
+      "loss": 1.1992,
+      "step": 9031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000125799131952526,
+      "loss": 1.0196,
+      "step": 9032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012578408035986887,
+      "loss": 2.125,
+      "step": 9033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012576902814144587,
+      "loss": 1.0031,
+      "step": 9034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012575397529762235,
+      "loss": 1.0115,
+      "step": 9035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001257389218287636,
+      "loss": 1.1432,
+      "step": 9036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012572386773523496,
+      "loss": 1.0849,
+      "step": 9037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001257088130174018,
+      "loss": 1.1198,
+      "step": 9038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012569375767562945,
+      "loss": 1.1045,
+      "step": 9039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012567870171028338,
+      "loss": 1.1548,
+      "step": 9040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012566364512172888,
+      "loss": 1.1609,
+      "step": 9041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012564858791033143,
+      "loss": 1.0869,
+      "step": 9042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001256335300764565,
+      "loss": 1.0294,
+      "step": 9043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012561847162046943,
+      "loss": 1.0566,
+      "step": 9044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012560341254273574,
+      "loss": 1.1526,
+      "step": 9045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001255883528436209,
+      "loss": 1.0533,
+      "step": 9046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012557329252349043,
+      "loss": 1.0362,
+      "step": 9047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001255582315827098,
+      "loss": 2.3188,
+      "step": 9048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012554317002164454,
+      "loss": 2.1326,
+      "step": 9049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012552810784066017,
+      "loss": 1.0075,
+      "step": 9050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001255130450401223,
+      "loss": 1.0082,
+      "step": 9051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001254979816203964,
+      "loss": 1.0696,
+      "step": 9052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012548291758184813,
+      "loss": 1.1904,
+      "step": 9053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012546785292484307,
+      "loss": 1.2628,
+      "step": 9054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001254527876497468,
+      "loss": 0.9767,
+      "step": 9055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012543772175692498,
+      "loss": 1.0465,
+      "step": 9056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012542265524674324,
+      "loss": 1.1173,
+      "step": 9057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012540758811956725,
+      "loss": 0.9394,
+      "step": 9058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012539252037576266,
+      "loss": 1.0726,
+      "step": 9059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012537745201569514,
+      "loss": 2.1531,
+      "step": 9060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012536238303973042,
+      "loss": 1.1231,
+      "step": 9061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012534731344823424,
+      "loss": 1.003,
+      "step": 9062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012533224324157226,
+      "loss": 1.0957,
+      "step": 9063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001253171724201103,
+      "loss": 1.1171,
+      "step": 9064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001253021009842141,
+      "loss": 0.9324,
+      "step": 9065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012528702893424936,
+      "loss": 1.2477,
+      "step": 9066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012527195627058197,
+      "loss": 1.198,
+      "step": 9067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012525688299357772,
+      "loss": 1.1859,
+      "step": 9068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012524180910360237,
+      "loss": 1.8921,
+      "step": 9069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012522673460102184,
+      "loss": 1.2354,
+      "step": 9070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012521165948620189,
+      "loss": 2.2346,
+      "step": 9071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012519658375950845,
+      "loss": 1.1009,
+      "step": 9072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001251815074213074,
+      "loss": 1.2169,
+      "step": 9073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012516643047196457,
+      "loss": 1.0594,
+      "step": 9074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012515135291184595,
+      "loss": 1.1859,
+      "step": 9075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001251362747413174,
+      "loss": 1.0708,
+      "step": 9076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001251211959607449,
+      "loss": 1.222,
+      "step": 9077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001251061165704944,
+      "loss": 1.0521,
+      "step": 9078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012509103657093183,
+      "loss": 2.1562,
+      "step": 9079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012507595596242325,
+      "loss": 1.1943,
+      "step": 9080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012506087474533462,
+      "loss": 1.1826,
+      "step": 9081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012504579292003189,
+      "loss": 1.1483,
+      "step": 9082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012503071048688115,
+      "loss": 1.011,
+      "step": 9083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012501562744624846,
+      "loss": 0.8891,
+      "step": 9084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012500054379849983,
+      "loss": 1.0711,
+      "step": 9085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012498545954400141,
+      "loss": 1.1008,
+      "step": 9086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012497037468311918,
+      "loss": 1.0908,
+      "step": 9087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001249552892162193,
+      "loss": 1.1281,
+      "step": 9088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001249402031436679,
+      "loss": 1.2136,
+      "step": 9089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012492511646583108,
+      "loss": 1.1461,
+      "step": 9090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012491002918307498,
+      "loss": 1.0222,
+      "step": 9091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012489494129576577,
+      "loss": 1.0486,
+      "step": 9092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012487985280426966,
+      "loss": 2.1672,
+      "step": 9093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001248647637089528,
+      "loss": 1.0876,
+      "step": 9094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00012484967401018142,
+      "loss": 1.0662,
+      "step": 9095
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001248345837083217,
+      "loss": 1.3708,
+      "step": 9096
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012481949280373993,
+      "loss": 1.1408,
+      "step": 9097
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001248044012968023,
+      "loss": 1.0181,
+      "step": 9098
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012478930918787505,
+      "loss": 2.2581,
+      "step": 9099
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012477421647732459,
+      "loss": 1.0153,
+      "step": 9100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012475912316551707,
+      "loss": 1.0502,
+      "step": 9101
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001247440292528189,
+      "loss": 1.072,
+      "step": 9102
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012472893473959636,
+      "loss": 1.1287,
+      "step": 9103
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012471383962621573,
+      "loss": 1.112,
+      "step": 9104
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012469874391304346,
+      "loss": 1.1385,
+      "step": 9105
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012468364760044584,
+      "loss": 1.0912,
+      "step": 9106
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012466855068878925,
+      "loss": 0.9386,
+      "step": 9107
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012465345317844018,
+      "loss": 1.2737,
+      "step": 9108
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012463835506976493,
+      "loss": 1.2101,
+      "step": 9109
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012462325636312995,
+      "loss": 1.2336,
+      "step": 9110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012460815705890173,
+      "loss": 1.2435,
+      "step": 9111
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012459305715744665,
+      "loss": 1.0926,
+      "step": 9112
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001245779566591312,
+      "loss": 1.0604,
+      "step": 9113
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012456285556432188,
+      "loss": 2.2517,
+      "step": 9114
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012454775387338518,
+      "loss": 1.1346,
+      "step": 9115
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001245326515866876,
+      "loss": 1.165,
+      "step": 9116
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001245175487045957,
+      "loss": 1.031,
+      "step": 9117
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012450244522747595,
+      "loss": 1.0168,
+      "step": 9118
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012448734115569492,
+      "loss": 1.1533,
+      "step": 9119
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012447223648961924,
+      "loss": 1.2615,
+      "step": 9120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001244571312296154,
+      "loss": 1.0935,
+      "step": 9121
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001244420253760501,
+      "loss": 1.1735,
+      "step": 9122
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001244269189292899,
+      "loss": 1.1558,
+      "step": 9123
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012441181188970144,
+      "loss": 1.1107,
+      "step": 9124
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001243967042576513,
+      "loss": 1.025,
+      "step": 9125
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012438159603350617,
+      "loss": 1.0001,
+      "step": 9126
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012436648721763277,
+      "loss": 1.061,
+      "step": 9127
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012435137781039772,
+      "loss": 1.066,
+      "step": 9128
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012433626781216772,
+      "loss": 1.1405,
+      "step": 9129
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012432115722330953,
+      "loss": 1.0325,
+      "step": 9130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012430604604418984,
+      "loss": 1.1208,
+      "step": 9131
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012429093427517536,
+      "loss": 1.2433,
+      "step": 9132
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012427582191663292,
+      "loss": 0.9393,
+      "step": 9133
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001242607089689292,
+      "loss": 1.2978,
+      "step": 9134
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012424559543243108,
+      "loss": 0.9677,
+      "step": 9135
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001242304813075053,
+      "loss": 0.9264,
+      "step": 9136
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012421536659451863,
+      "loss": 1.2628,
+      "step": 9137
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012420025129383804,
+      "loss": 2.1943,
+      "step": 9138
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001241851354058302,
+      "loss": 1.0891,
+      "step": 9139
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012417001893086206,
+      "loss": 1.3647,
+      "step": 9140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012415490186930048,
+      "loss": 1.1982,
+      "step": 9141
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001241397842215123,
+      "loss": 1.2099,
+      "step": 9142
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001241246659878645,
+      "loss": 1.115,
+      "step": 9143
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012410954716872392,
+      "loss": 1.0465,
+      "step": 9144
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012409442776445748,
+      "loss": 1.0984,
+      "step": 9145
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001240793077754322,
+      "loss": 1.2095,
+      "step": 9146
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012406418720201497,
+      "loss": 1.0764,
+      "step": 9147
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012404906604457272,
+      "loss": 1.1298,
+      "step": 9148
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012403394430347253,
+      "loss": 1.0251,
+      "step": 9149
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012401882197908133,
+      "loss": 1.2137,
+      "step": 9150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012400369907176613,
+      "loss": 1.0868,
+      "step": 9151
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.000123988575581894,
+      "loss": 1.0277,
+      "step": 9152
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012397345150983195,
+      "loss": 1.1926,
+      "step": 9153
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012395832685594706,
+      "loss": 0.9102,
+      "step": 9154
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012394320162060633,
+      "loss": 1.0911,
+      "step": 9155
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012392807580417688,
+      "loss": 1.1038,
+      "step": 9156
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012391294940702587,
+      "loss": 2.272,
+      "step": 9157
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001238978224295203,
+      "loss": 0.9648,
+      "step": 9158
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012388269487202737,
+      "loss": 1.1038,
+      "step": 9159
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012386756673491415,
+      "loss": 1.1171,
+      "step": 9160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001238524380185479,
+      "loss": 1.1683,
+      "step": 9161
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012383730872329567,
+      "loss": 1.1556,
+      "step": 9162
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001238221788495247,
+      "loss": 1.1182,
+      "step": 9163
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012380704839760214,
+      "loss": 1.1285,
+      "step": 9164
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001237919173678953,
+      "loss": 1.0797,
+      "step": 9165
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012377678576077128,
+      "loss": 1.222,
+      "step": 9166
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001237616535765974,
+      "loss": 1.1499,
+      "step": 9167
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012374652081574084,
+      "loss": 1.046,
+      "step": 9168
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012373138747856895,
+      "loss": 1.1504,
+      "step": 9169
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012371625356544894,
+      "loss": 1.1286,
+      "step": 9170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001237011190767481,
+      "loss": 1.0455,
+      "step": 9171
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001236859840128338,
+      "loss": 1.0045,
+      "step": 9172
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001236708483740733,
+      "loss": 1.0591,
+      "step": 9173
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012365571216083393,
+      "loss": 1.2076,
+      "step": 9174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012364057537348307,
+      "loss": 1.1829,
+      "step": 9175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012362543801238812,
+      "loss": 2.1274,
+      "step": 9176
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012361030007791634,
+      "loss": 1.2371,
+      "step": 9177
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012359516157043525,
+      "loss": 1.0486,
+      "step": 9178
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012358002249031214,
+      "loss": 1.1057,
+      "step": 9179
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012356488283791448,
+      "loss": 1.1995,
+      "step": 9180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012354974261360974,
+      "loss": 1.2223,
+      "step": 9181
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012353460181776529,
+      "loss": 2.2183,
+      "step": 9182
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012351946045074865,
+      "loss": 1.1182,
+      "step": 9183
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012350431851292724,
+      "loss": 2.1476,
+      "step": 9184
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012348917600466859,
+      "loss": 1.17,
+      "step": 9185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012347403292634017,
+      "loss": 1.2112,
+      "step": 9186
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012345888927830953,
+      "loss": 1.0745,
+      "step": 9187
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012344374506094416,
+      "loss": 1.2763,
+      "step": 9188
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012342860027461162,
+      "loss": 1.1954,
+      "step": 9189
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001234134549196795,
+      "loss": 1.1465,
+      "step": 9190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001233983089965153,
+      "loss": 1.2531,
+      "step": 9191
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012338316250548665,
+      "loss": 1.127,
+      "step": 9192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001233680154469611,
+      "loss": 1.0894,
+      "step": 9193
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012335286782130632,
+      "loss": 0.9857,
+      "step": 9194
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012333771962888993,
+      "loss": 1.1647,
+      "step": 9195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001233225708700795,
+      "loss": 1.1463,
+      "step": 9196
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001233074215452428,
+      "loss": 0.9736,
+      "step": 9197
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001232922716547474,
+      "loss": 1.1313,
+      "step": 9198
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00012327712119896098,
+      "loss": 1.2191,
+      "step": 9199
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001232619701782513,
+      "loss": 1.084,
+      "step": 9200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.000123246818592986,
+      "loss": 0.9893,
+      "step": 9201
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001232316664435328,
+      "loss": 1.1199,
+      "step": 9202
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012321651373025953,
+      "loss": 1.0537,
+      "step": 9203
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012320136045353383,
+      "loss": 1.1799,
+      "step": 9204
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012318620661372354,
+      "loss": 1.1729,
+      "step": 9205
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012317105221119638,
+      "loss": 1.1662,
+      "step": 9206
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012315589724632014,
+      "loss": 1.0944,
+      "step": 9207
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012314074171946267,
+      "loss": 1.0937,
+      "step": 9208
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012312558563099172,
+      "loss": 1.232,
+      "step": 9209
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012311042898127518,
+      "loss": 0.9699,
+      "step": 9210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001230952717706809,
+      "loss": 1.1284,
+      "step": 9211
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012308011399957667,
+      "loss": 1.0876,
+      "step": 9212
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001230649556683304,
+      "loss": 1.4056,
+      "step": 9213
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012304979677731,
+      "loss": 1.1531,
+      "step": 9214
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012303463732688333,
+      "loss": 1.2005,
+      "step": 9215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012301947731741835,
+      "loss": 1.0744,
+      "step": 9216
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001230043167492829,
+      "loss": 1.0938,
+      "step": 9217
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012298915562284496,
+      "loss": 1.0403,
+      "step": 9218
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012297399393847256,
+      "loss": 1.1648,
+      "step": 9219
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012295883169653355,
+      "loss": 2.1619,
+      "step": 9220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012294366889739595,
+      "loss": 1.0208,
+      "step": 9221
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001229285055414278,
+      "loss": 1.1187,
+      "step": 9222
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.000122913341628997,
+      "loss": 0.9807,
+      "step": 9223
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001228981771604717,
+      "loss": 1.2162,
+      "step": 9224
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012288301213621985,
+      "loss": 1.0634,
+      "step": 9225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001228678465566095,
+      "loss": 1.1555,
+      "step": 9226
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012285268042200876,
+      "loss": 1.9883,
+      "step": 9227
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012283751373278566,
+      "loss": 1.1005,
+      "step": 9228
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012282234648930827,
+      "loss": 1.2113,
+      "step": 9229
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012280717869194476,
+      "loss": 1.0526,
+      "step": 9230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012279201034106317,
+      "loss": 1.2031,
+      "step": 9231
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001227768414370317,
+      "loss": 1.2079,
+      "step": 9232
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012276167198021842,
+      "loss": 1.0659,
+      "step": 9233
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012274650197099154,
+      "loss": 1.1251,
+      "step": 9234
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012273133140971922,
+      "loss": 0.9738,
+      "step": 9235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012271616029676957,
+      "loss": 2.1672,
+      "step": 9236
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001227009886325109,
+      "loss": 1.1634,
+      "step": 9237
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012268581641731134,
+      "loss": 0.8382,
+      "step": 9238
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012267064365153912,
+      "loss": 1.0381,
+      "step": 9239
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001226554703355625,
+      "loss": 1.1726,
+      "step": 9240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012264029646974972,
+      "loss": 1.124,
+      "step": 9241
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012262512205446903,
+      "loss": 0.9954,
+      "step": 9242
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012260994709008873,
+      "loss": 2.0879,
+      "step": 9243
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012259477157697705,
+      "loss": 1.1407,
+      "step": 9244
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012257959551550236,
+      "loss": 1.0706,
+      "step": 9245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012256441890603294,
+      "loss": 1.0735,
+      "step": 9246
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012254924174893712,
+      "loss": 1.2055,
+      "step": 9247
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012253406404458327,
+      "loss": 1.084,
+      "step": 9248
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001225188857933397,
+      "loss": 0.9538,
+      "step": 9249
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001225037069955748,
+      "loss": 1.0002,
+      "step": 9250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012248852765165698,
+      "loss": 1.073,
+      "step": 9251
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012247334776195457,
+      "loss": 1.2244,
+      "step": 9252
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.000122458167326836,
+      "loss": 1.1442,
+      "step": 9253
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012244298634666974,
+      "loss": 1.1121,
+      "step": 9254
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012242780482182413,
+      "loss": 1.1173,
+      "step": 9255
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012241262275266776,
+      "loss": 0.9633,
+      "step": 9256
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012239744013956897,
+      "loss": 1.1562,
+      "step": 9257
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012238225698289626,
+      "loss": 0.9857,
+      "step": 9258
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012236707328301812,
+      "loss": 1.0856,
+      "step": 9259
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012235188904030308,
+      "loss": 1.0225,
+      "step": 9260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001223367042551196,
+      "loss": 1.1267,
+      "step": 9261
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001223215189278363,
+      "loss": 1.1167,
+      "step": 9262
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012230633305882163,
+      "loss": 0.9933,
+      "step": 9263
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012229114664844413,
+      "loss": 1.0916,
+      "step": 9264
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012227595969707247,
+      "loss": 1.2057,
+      "step": 9265
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001222607722050751,
+      "loss": 1.052,
+      "step": 9266
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001222455841728207,
+      "loss": 1.0875,
+      "step": 9267
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001222303956006779,
+      "loss": 0.9752,
+      "step": 9268
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001222152064890152,
+      "loss": 1.2554,
+      "step": 9269
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012220001683820134,
+      "loss": 1.1413,
+      "step": 9270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012218482664860496,
+      "loss": 1.1769,
+      "step": 9271
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012216963592059465,
+      "loss": 1.176,
+      "step": 9272
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012215444465453912,
+      "loss": 1.0992,
+      "step": 9273
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012213925285080706,
+      "loss": 1.2254,
+      "step": 9274
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012212406050976713,
+      "loss": 1.1351,
+      "step": 9275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001221088676317881,
+      "loss": 2.2183,
+      "step": 9276
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012209367421723864,
+      "loss": 1.2608,
+      "step": 9277
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012207848026648751,
+      "loss": 0.9216,
+      "step": 9278
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012206328577990347,
+      "loss": 2.2112,
+      "step": 9279
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012204809075785526,
+      "loss": 1.1102,
+      "step": 9280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012203289520071168,
+      "loss": 1.1508,
+      "step": 9281
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012201769910884146,
+      "loss": 1.124,
+      "step": 9282
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012200250248261346,
+      "loss": 1.0164,
+      "step": 9283
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012198730532239651,
+      "loss": 1.0897,
+      "step": 9284
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012197210762855938,
+      "loss": 1.0566,
+      "step": 9285
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012195690940147093,
+      "loss": 1.1223,
+      "step": 9286
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012194171064150002,
+      "loss": 1.1935,
+      "step": 9287
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001219265113490155,
+      "loss": 1.0203,
+      "step": 9288
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012191131152438629,
+      "loss": 1.119,
+      "step": 9289
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001218961111679812,
+      "loss": 1.3114,
+      "step": 9290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012188091028016921,
+      "loss": 0.9857,
+      "step": 9291
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012186570886131922,
+      "loss": 1.1755,
+      "step": 9292
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012185050691180018,
+      "loss": 1.1487,
+      "step": 9293
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012183530443198097,
+      "loss": 0.9773,
+      "step": 9294
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012182010142223061,
+      "loss": 1.0365,
+      "step": 9295
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012180489788291803,
+      "loss": 1.0857,
+      "step": 9296
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012178969381441224,
+      "loss": 1.1206,
+      "step": 9297
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012177448921708219,
+      "loss": 1.2308,
+      "step": 9298
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012175928409129693,
+      "loss": 1.0874,
+      "step": 9299
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001217440784374255,
+      "loss": 1.1404,
+      "step": 9300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012172887225583685,
+      "loss": 0.989,
+      "step": 9301
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001217136655469001,
+      "loss": 1.2084,
+      "step": 9302
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00012169845831098427,
+      "loss": 1.139,
+      "step": 9303
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012168325054845844,
+      "loss": 1.1378,
+      "step": 9304
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012166804225969173,
+      "loss": 1.0832,
+      "step": 9305
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012165283344505319,
+      "loss": 1.2718,
+      "step": 9306
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012163762410491196,
+      "loss": 1.1023,
+      "step": 9307
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012162241423963713,
+      "loss": 1.1732,
+      "step": 9308
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012160720384959788,
+      "loss": 1.1475,
+      "step": 9309
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001215919929351633,
+      "loss": 0.8925,
+      "step": 9310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001215767814967026,
+      "loss": 1.1249,
+      "step": 9311
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012156156953458492,
+      "loss": 1.0891,
+      "step": 9312
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012154635704917952,
+      "loss": 1.0598,
+      "step": 9313
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012153114404085548,
+      "loss": 1.1602,
+      "step": 9314
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012151593050998211,
+      "loss": 1.067,
+      "step": 9315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001215007164569286,
+      "loss": 1.0154,
+      "step": 9316
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012148550188206415,
+      "loss": 1.1573,
+      "step": 9317
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012147028678575807,
+      "loss": 1.3511,
+      "step": 9318
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001214550711683796,
+      "loss": 0.9295,
+      "step": 9319
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000121439855030298,
+      "loss": 0.9474,
+      "step": 9320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012142463837188259,
+      "loss": 2.199,
+      "step": 9321
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012140942119350263,
+      "loss": 1.1397,
+      "step": 9322
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012139420349552746,
+      "loss": 1.0923,
+      "step": 9323
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012137898527832639,
+      "loss": 1.1542,
+      "step": 9324
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012136376654226877,
+      "loss": 1.2662,
+      "step": 9325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012134854728772393,
+      "loss": 1.1028,
+      "step": 9326
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012133332751506128,
+      "loss": 1.0587,
+      "step": 9327
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012131810722465016,
+      "loss": 0.9794,
+      "step": 9328
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012130288641685994,
+      "loss": 1.1452,
+      "step": 9329
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012128766509206009,
+      "loss": 1.1989,
+      "step": 9330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012127244325061995,
+      "loss": 1.2976,
+      "step": 9331
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012125722089290894,
+      "loss": 1.3002,
+      "step": 9332
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012124199801929659,
+      "loss": 1.1249,
+      "step": 9333
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012122677463015227,
+      "loss": 1.0935,
+      "step": 9334
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012121155072584546,
+      "loss": 1.0317,
+      "step": 9335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012119632630674564,
+      "loss": 0.8564,
+      "step": 9336
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012118110137322232,
+      "loss": 1.1216,
+      "step": 9337
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012116587592564499,
+      "loss": 1.152,
+      "step": 9338
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012115064996438313,
+      "loss": 0.975,
+      "step": 9339
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012113542348980629,
+      "loss": 1.0933,
+      "step": 9340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012112019650228403,
+      "loss": 1.0734,
+      "step": 9341
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012110496900218586,
+      "loss": 1.1524,
+      "step": 9342
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001210897409898814,
+      "loss": 2.1016,
+      "step": 9343
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012107451246574017,
+      "loss": 1.2392,
+      "step": 9344
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012105928343013178,
+      "loss": 1.0264,
+      "step": 9345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012104405388342584,
+      "loss": 1.1308,
+      "step": 9346
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012102882382599192,
+      "loss": 1.122,
+      "step": 9347
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012101359325819967,
+      "loss": 1.1832,
+      "step": 9348
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012099836218041878,
+      "loss": 2.2026,
+      "step": 9349
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012098313059301882,
+      "loss": 1.0424,
+      "step": 9350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012096789849636952,
+      "loss": 1.0962,
+      "step": 9351
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001209526658908405,
+      "loss": 1.0408,
+      "step": 9352
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012093743277680147,
+      "loss": 0.9816,
+      "step": 9353
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012092219915462215,
+      "loss": 2.1387,
+      "step": 9354
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001209069650246722,
+      "loss": 1.1883,
+      "step": 9355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001208917303873214,
+      "loss": 1.1268,
+      "step": 9356
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012087649524293944,
+      "loss": 1.1758,
+      "step": 9357
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012086125959189611,
+      "loss": 1.126,
+      "step": 9358
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012084602343456115,
+      "loss": 1.2015,
+      "step": 9359
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012083078677130435,
+      "loss": 2.3149,
+      "step": 9360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012081554960249546,
+      "loss": 1.1319,
+      "step": 9361
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012080031192850429,
+      "loss": 2.208,
+      "step": 9362
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012078507374970067,
+      "loss": 1.1697,
+      "step": 9363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001207698350664544,
+      "loss": 1.1339,
+      "step": 9364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012075459587913536,
+      "loss": 1.0782,
+      "step": 9365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012073935618811334,
+      "loss": 1.0191,
+      "step": 9366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012072411599375821,
+      "loss": 1.3669,
+      "step": 9367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012070887529643989,
+      "loss": 0.9502,
+      "step": 9368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012069363409652819,
+      "loss": 1.1851,
+      "step": 9369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012067839239439305,
+      "loss": 1.0388,
+      "step": 9370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012066315019040438,
+      "loss": 1.0583,
+      "step": 9371
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012064790748493206,
+      "loss": 0.9377,
+      "step": 9372
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001206326642783461,
+      "loss": 1.1526,
+      "step": 9373
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001206174205710164,
+      "loss": 1.0798,
+      "step": 9374
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012060217636331288,
+      "loss": 1.0812,
+      "step": 9375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012058693165560557,
+      "loss": 1.126,
+      "step": 9376
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001205716864482644,
+      "loss": 1.1319,
+      "step": 9377
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012055644074165941,
+      "loss": 1.0885,
+      "step": 9378
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012054119453616057,
+      "loss": 1.1314,
+      "step": 9379
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012052594783213789,
+      "loss": 1.0929,
+      "step": 9380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012051070062996149,
+      "loss": 1.1177,
+      "step": 9381
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012049545293000126,
+      "loss": 1.0499,
+      "step": 9382
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012048020473262736,
+      "loss": 1.0494,
+      "step": 9383
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012046495603820984,
+      "loss": 1.0702,
+      "step": 9384
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012044970684711876,
+      "loss": 1.0494,
+      "step": 9385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012043445715972424,
+      "loss": 1.1411,
+      "step": 9386
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012041920697639632,
+      "loss": 1.0189,
+      "step": 9387
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012040395629750518,
+      "loss": 1.0292,
+      "step": 9388
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012038870512342092,
+      "loss": 1.1767,
+      "step": 9389
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012037345345451368,
+      "loss": 1.1058,
+      "step": 9390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012035820129115356,
+      "loss": 1.0894,
+      "step": 9391
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012034294863371083,
+      "loss": 1.1199,
+      "step": 9392
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012032769548255555,
+      "loss": 1.0703,
+      "step": 9393
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012031244183805797,
+      "loss": 0.9205,
+      "step": 9394
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012029718770058828,
+      "loss": 1.1707,
+      "step": 9395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012028193307051672,
+      "loss": 1.0353,
+      "step": 9396
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012026667794821344,
+      "loss": 0.9966,
+      "step": 9397
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012025142233404874,
+      "loss": 2.1238,
+      "step": 9398
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001202361662283928,
+      "loss": 1.1908,
+      "step": 9399
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012022090963161595,
+      "loss": 1.0646,
+      "step": 9400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012020565254408838,
+      "loss": 1.2393,
+      "step": 9401
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012019039496618043,
+      "loss": 1.3263,
+      "step": 9402
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012017513689826241,
+      "loss": 1.0915,
+      "step": 9403
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012015987834070457,
+      "loss": 1.0091,
+      "step": 9404
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012014461929387724,
+      "loss": 1.2766,
+      "step": 9405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012012935975815076,
+      "loss": 1.1338,
+      "step": 9406
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012011409973389547,
+      "loss": 1.1055,
+      "step": 9407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00012009883922148172,
+      "loss": 1.1793,
+      "step": 9408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001200835782212799,
+      "loss": 1.0588,
+      "step": 9409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00012006831673366035,
+      "loss": 1.1071,
+      "step": 9410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00012005305475899349,
+      "loss": 1.2412,
+      "step": 9411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00012003779229764966,
+      "loss": 1.2111,
+      "step": 9412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00012002252934999934,
+      "loss": 1.0914,
+      "step": 9413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00012000726591641293,
+      "loss": 1.1265,
+      "step": 9414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011999200199726084,
+      "loss": 1.0788,
+      "step": 9415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011997673759291356,
+      "loss": 1.0758,
+      "step": 9416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011996147270374152,
+      "loss": 1.0682,
+      "step": 9417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011994620733011521,
+      "loss": 1.07,
+      "step": 9418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011993094147240511,
+      "loss": 1.1117,
+      "step": 9419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011991567513098167,
+      "loss": 1.0737,
+      "step": 9420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011990040830621544,
+      "loss": 1.204,
+      "step": 9421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011988514099847697,
+      "loss": 0.9239,
+      "step": 9422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001198698732081367,
+      "loss": 1.1624,
+      "step": 9423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011985460493556524,
+      "loss": 0.9806,
+      "step": 9424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011983933618113315,
+      "loss": 1.0666,
+      "step": 9425
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011982406694521091,
+      "loss": 1.0097,
+      "step": 9426
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001198087972281692,
+      "loss": 1.2003,
+      "step": 9427
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011979352703037854,
+      "loss": 2.2151,
+      "step": 9428
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011977825635220954,
+      "loss": 1.1016,
+      "step": 9429
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011976298519403284,
+      "loss": 1.1072,
+      "step": 9430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011974771355621903,
+      "loss": 1.16,
+      "step": 9431
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011973244143913878,
+      "loss": 1.0347,
+      "step": 9432
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011971716884316272,
+      "loss": 1.0445,
+      "step": 9433
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011970189576866149,
+      "loss": 0.984,
+      "step": 9434
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011968662221600578,
+      "loss": 1.2531,
+      "step": 9435
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011967134818556625,
+      "loss": 1.0678,
+      "step": 9436
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011965607367771358,
+      "loss": 1.1349,
+      "step": 9437
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011964079869281855,
+      "loss": 1.0899,
+      "step": 9438
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001196255232312518,
+      "loss": 1.0202,
+      "step": 9439
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011961024729338411,
+      "loss": 1.1876,
+      "step": 9440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011959497087958618,
+      "loss": 1.1339,
+      "step": 9441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011957969399022874,
+      "loss": 1.1617,
+      "step": 9442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011956441662568263,
+      "loss": 1.1627,
+      "step": 9443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011954913878631855,
+      "loss": 2.002,
+      "step": 9444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001195338604725073,
+      "loss": 1.0988,
+      "step": 9445
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011951858168461972,
+      "loss": 1.2339,
+      "step": 9446
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011950330242302659,
+      "loss": 1.0383,
+      "step": 9447
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011948802268809869,
+      "loss": 1.2455,
+      "step": 9448
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011947274248020694,
+      "loss": 1.0845,
+      "step": 9449
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011945746179972208,
+      "loss": 1.0099,
+      "step": 9450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011944218064701505,
+      "loss": 1.1415,
+      "step": 9451
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011942689902245665,
+      "loss": 1.1719,
+      "step": 9452
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001194116169264178,
+      "loss": 1.0684,
+      "step": 9453
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001193963343592694,
+      "loss": 0.9953,
+      "step": 9454
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011938105132138231,
+      "loss": 1.202,
+      "step": 9455
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011936576781312743,
+      "loss": 1.2599,
+      "step": 9456
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011935048383487574,
+      "loss": 1.0645,
+      "step": 9457
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011933519938699813,
+      "loss": 0.929,
+      "step": 9458
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011931991446986554,
+      "loss": 1.027,
+      "step": 9459
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011930462908384899,
+      "loss": 1.1241,
+      "step": 9460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011928934322931934,
+      "loss": 1.1019,
+      "step": 9461
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011927405690664769,
+      "loss": 1.1122,
+      "step": 9462
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011925877011620497,
+      "loss": 1.2138,
+      "step": 9463
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011924348285836215,
+      "loss": 1.0425,
+      "step": 9464
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001192281951334903,
+      "loss": 1.0915,
+      "step": 9465
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011921290694196041,
+      "loss": 1.0741,
+      "step": 9466
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011919761828414353,
+      "loss": 1.1859,
+      "step": 9467
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011918232916041075,
+      "loss": 1.0274,
+      "step": 9468
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011916703957113304,
+      "loss": 0.9557,
+      "step": 9469
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011915174951668153,
+      "loss": 1.0196,
+      "step": 9470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001191364589974273,
+      "loss": 0.9271,
+      "step": 9471
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001191211680137414,
+      "loss": 0.9511,
+      "step": 9472
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.000119105876565995,
+      "loss": 1.0959,
+      "step": 9473
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011909058465455915,
+      "loss": 1.1047,
+      "step": 9474
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011907529227980503,
+      "loss": 1.1147,
+      "step": 9475
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011905999944210376,
+      "loss": 1.1122,
+      "step": 9476
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011904470614182648,
+      "loss": 1.2465,
+      "step": 9477
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011902941237934432,
+      "loss": 1.0359,
+      "step": 9478
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011901411815502852,
+      "loss": 1.1099,
+      "step": 9479
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011899882346925021,
+      "loss": 0.9924,
+      "step": 9480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011898352832238062,
+      "loss": 2.2961,
+      "step": 9481
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011896823271479092,
+      "loss": 1.3421,
+      "step": 9482
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011895293664685235,
+      "loss": 1.0683,
+      "step": 9483
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011893764011893616,
+      "loss": 1.0896,
+      "step": 9484
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001189223431314135,
+      "loss": 1.1537,
+      "step": 9485
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011890704568465572,
+      "loss": 2.1567,
+      "step": 9486
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011889174777903403,
+      "loss": 1.1032,
+      "step": 9487
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011887644941491969,
+      "loss": 1.1086,
+      "step": 9488
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011886115059268403,
+      "loss": 0.8695,
+      "step": 9489
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011884585131269829,
+      "loss": 1.0197,
+      "step": 9490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011883055157533385,
+      "loss": 1.0375,
+      "step": 9491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011881525138096197,
+      "loss": 1.0768,
+      "step": 9492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011879995072995398,
+      "loss": 1.1138,
+      "step": 9493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011878464962268122,
+      "loss": 1.1629,
+      "step": 9494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011876934805951505,
+      "loss": 1.112,
+      "step": 9495
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001187540460408268,
+      "loss": 1.1734,
+      "step": 9496
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001187387435669879,
+      "loss": 1.0644,
+      "step": 9497
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011872344063836971,
+      "loss": 1.1859,
+      "step": 9498
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001187081372553436,
+      "loss": 1.1517,
+      "step": 9499
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011869283341828101,
+      "loss": 1.0482,
+      "step": 9500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011867752912755333,
+      "loss": 1.1169,
+      "step": 9501
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011866222438353196,
+      "loss": 1.0352,
+      "step": 9502
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011864691918658842,
+      "loss": 2.1843,
+      "step": 9503
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011863161353709406,
+      "loss": 1.1155,
+      "step": 9504
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011861630743542043,
+      "loss": 1.1563,
+      "step": 9505
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011860100088193897,
+      "loss": 1.1005,
+      "step": 9506
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011858569387702113,
+      "loss": 1.106,
+      "step": 9507
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011857038642103844,
+      "loss": 1.022,
+      "step": 9508
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011855507851436236,
+      "loss": 1.2025,
+      "step": 9509
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011853977015736443,
+      "loss": 2.218,
+      "step": 9510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011852446135041624,
+      "loss": 1.0766,
+      "step": 9511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011850915209388921,
+      "loss": 1.1219,
+      "step": 9512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011849384238815498,
+      "loss": 1.1364,
+      "step": 9513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011847853223358507,
+      "loss": 1.1178,
+      "step": 9514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011846322163055104,
+      "loss": 2.1753,
+      "step": 9515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011844791057942448,
+      "loss": 1.0958,
+      "step": 9516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011843259908057698,
+      "loss": 1.1395,
+      "step": 9517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011841728713438014,
+      "loss": 1.124,
+      "step": 9518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001184019747412056,
+      "loss": 0.9672,
+      "step": 9519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011838666190142496,
+      "loss": 1.2015,
+      "step": 9520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011837134861540984,
+      "loss": 1.1259,
+      "step": 9521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011835603488353193,
+      "loss": 1.1177,
+      "step": 9522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011834072070616282,
+      "loss": 1.1801,
+      "step": 9523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011832540608367422,
+      "loss": 1.019,
+      "step": 9524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011831009101643785,
+      "loss": 1.1484,
+      "step": 9525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011829477550482529,
+      "loss": 1.1386,
+      "step": 9526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011827945954920834,
+      "loss": 1.0381,
+      "step": 9527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011826414314995866,
+      "loss": 1.2083,
+      "step": 9528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011824882630744796,
+      "loss": 0.9809,
+      "step": 9529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011823350902204801,
+      "loss": 1.2086,
+      "step": 9530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011821819129413052,
+      "loss": 1.0894,
+      "step": 9531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011820287312406726,
+      "loss": 1.1449,
+      "step": 9532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011818755451223002,
+      "loss": 1.1584,
+      "step": 9533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011817223545899053,
+      "loss": 1.2269,
+      "step": 9534
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011815691596472059,
+      "loss": 0.9217,
+      "step": 9535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.000118141596029792,
+      "loss": 0.9457,
+      "step": 9536
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011812627565457654,
+      "loss": 1.1376,
+      "step": 9537
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001181109548394461,
+      "loss": 1.0438,
+      "step": 9538
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001180956335847724,
+      "loss": 1.0092,
+      "step": 9539
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011808031189092736,
+      "loss": 1.1221,
+      "step": 9540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011806498975828285,
+      "loss": 1.0821,
+      "step": 9541
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011804966718721062,
+      "loss": 2.4155,
+      "step": 9542
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011803434417808266,
+      "loss": 0.995,
+      "step": 9543
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011801902073127078,
+      "loss": 1.1831,
+      "step": 9544
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011800369684714687,
+      "loss": 0.9942,
+      "step": 9545
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011798837252608291,
+      "loss": 1.2117,
+      "step": 9546
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011797304776845068,
+      "loss": 1.1651,
+      "step": 9547
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011795772257462223,
+      "loss": 1.0723,
+      "step": 9548
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011794239694496942,
+      "loss": 2.1663,
+      "step": 9549
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011792707087986426,
+      "loss": 2.417,
+      "step": 9550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011791174437967863,
+      "loss": 1.1368,
+      "step": 9551
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011789641744478455,
+      "loss": 1.0704,
+      "step": 9552
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011788109007555394,
+      "loss": 1.1621,
+      "step": 9553
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011786576227235888,
+      "loss": 1.1346,
+      "step": 9554
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011785043403557126,
+      "loss": 1.0948,
+      "step": 9555
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011783510536556314,
+      "loss": 1.3534,
+      "step": 9556
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011781977626270661,
+      "loss": 0.9725,
+      "step": 9557
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011780444672737356,
+      "loss": 1.2012,
+      "step": 9558
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011778911675993612,
+      "loss": 1.085,
+      "step": 9559
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011777378636076632,
+      "loss": 1.2705,
+      "step": 9560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011775845553023619,
+      "loss": 1.1104,
+      "step": 9561
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011774312426871788,
+      "loss": 1.047,
+      "step": 9562
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011772779257658338,
+      "loss": 0.9116,
+      "step": 9563
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011771246045420481,
+      "loss": 1.2405,
+      "step": 9564
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011769712790195434,
+      "loss": 1.18,
+      "step": 9565
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011768179492020399,
+      "loss": 1.1573,
+      "step": 9566
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011766646150932592,
+      "loss": 0.9513,
+      "step": 9567
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011765112766969228,
+      "loss": 1.2196,
+      "step": 9568
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011763579340167519,
+      "loss": 1.0871,
+      "step": 9569
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011762045870564682,
+      "loss": 1.0484,
+      "step": 9570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011760512358197931,
+      "loss": 1.2358,
+      "step": 9571
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011758978803104488,
+      "loss": 1.1776,
+      "step": 9572
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011757445205321568,
+      "loss": 0.9361,
+      "step": 9573
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001175591156488639,
+      "loss": 0.9759,
+      "step": 9574
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011754377881836177,
+      "loss": 1.1232,
+      "step": 9575
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011752844156208151,
+      "loss": 1.0873,
+      "step": 9576
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011751310388039532,
+      "loss": 1.0251,
+      "step": 9577
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011749776577367546,
+      "loss": 1.105,
+      "step": 9578
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001174824272422942,
+      "loss": 1.1679,
+      "step": 9579
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011746708828662372,
+      "loss": 0.9622,
+      "step": 9580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011745174890703636,
+      "loss": 0.9778,
+      "step": 9581
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011743640910390436,
+      "loss": 1.1772,
+      "step": 9582
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011742106887760002,
+      "loss": 1.0429,
+      "step": 9583
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011740572822849567,
+      "loss": 1.9817,
+      "step": 9584
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011739038715696355,
+      "loss": 1.0442,
+      "step": 9585
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011737504566337607,
+      "loss": 0.9002,
+      "step": 9586
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011735970374810553,
+      "loss": 1.1224,
+      "step": 9587
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011734436141152418,
+      "loss": 1.0158,
+      "step": 9588
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001173290186540045,
+      "loss": 1.0608,
+      "step": 9589
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011731367547591876,
+      "loss": 1.1997,
+      "step": 9590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011729833187763934,
+      "loss": 1.2065,
+      "step": 9591
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011728298785953871,
+      "loss": 0.9208,
+      "step": 9592
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011726764342198915,
+      "loss": 1.0978,
+      "step": 9593
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011725229856536313,
+      "loss": 1.0677,
+      "step": 9594
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011723695329003304,
+      "loss": 1.0565,
+      "step": 9595
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011722160759637127,
+      "loss": 1.0676,
+      "step": 9596
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011720626148475029,
+      "loss": 0.9816,
+      "step": 9597
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011719091495554256,
+      "loss": 1.1863,
+      "step": 9598
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011717556800912048,
+      "loss": 0.9912,
+      "step": 9599
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011716022064585655,
+      "loss": 1.1591,
+      "step": 9600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011714487286612324,
+      "loss": 1.0179,
+      "step": 9601
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011712952467029301,
+      "loss": 1.0847,
+      "step": 9602
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011711417605873838,
+      "loss": 1.0897,
+      "step": 9603
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001170988270318318,
+      "loss": 1.1445,
+      "step": 9604
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011708347758994584,
+      "loss": 1.03,
+      "step": 9605
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011706812773345302,
+      "loss": 1.0607,
+      "step": 9606
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011705277746272582,
+      "loss": 1.0076,
+      "step": 9607
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011703742677813685,
+      "loss": 1.0712,
+      "step": 9608
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011702207568005865,
+      "loss": 1.1945,
+      "step": 9609
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011700672416886371,
+      "loss": 1.119,
+      "step": 9610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011699137224492471,
+      "loss": 1.1049,
+      "step": 9611
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011697601990861416,
+      "loss": 1.1099,
+      "step": 9612
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011696066716030465,
+      "loss": 1.0536,
+      "step": 9613
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011694531400036884,
+      "loss": 1.1346,
+      "step": 9614
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011692996042917928,
+      "loss": 1.1318,
+      "step": 9615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011691460644710865,
+      "loss": 1.0585,
+      "step": 9616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011689925205452956,
+      "loss": 1.0325,
+      "step": 9617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011688389725181463,
+      "loss": 1.1289,
+      "step": 9618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011686854203933654,
+      "loss": 0.9386,
+      "step": 9619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011685318641746794,
+      "loss": 1.1232,
+      "step": 9620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001168378303865815,
+      "loss": 1.1593,
+      "step": 9621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011682247394704993,
+      "loss": 1.1227,
+      "step": 9622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011680711709924588,
+      "loss": 1.0892,
+      "step": 9623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011679175984354209,
+      "loss": 1.1065,
+      "step": 9624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011677640218031128,
+      "loss": 1.077,
+      "step": 9625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001167610441099261,
+      "loss": 1.1402,
+      "step": 9626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011674568563275936,
+      "loss": 1.1369,
+      "step": 9627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011673032674918376,
+      "loss": 1.106,
+      "step": 9628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011671496745957204,
+      "loss": 1.106,
+      "step": 9629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011669960776429705,
+      "loss": 1.209,
+      "step": 9630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011668424766373147,
+      "loss": 1.1213,
+      "step": 9631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011666888715824811,
+      "loss": 1.0509,
+      "step": 9632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011665352624821977,
+      "loss": 1.0258,
+      "step": 9633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011663816493401922,
+      "loss": 0.9516,
+      "step": 9634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001166228032160193,
+      "loss": 1.0839,
+      "step": 9635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011660744109459283,
+      "loss": 1.1849,
+      "step": 9636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011659207857011263,
+      "loss": 1.1888,
+      "step": 9637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011657671564295159,
+      "loss": 0.819,
+      "step": 9638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011656135231348244,
+      "loss": 1.0108,
+      "step": 9639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011654598858207815,
+      "loss": 1.051,
+      "step": 9640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011653062444911156,
+      "loss": 1.1628,
+      "step": 9641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011651525991495553,
+      "loss": 1.0614,
+      "step": 9642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000116499894979983,
+      "loss": 1.1224,
+      "step": 9643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011648452964456678,
+      "loss": 2.1372,
+      "step": 9644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011646916390907984,
+      "loss": 1.1028,
+      "step": 9645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011645379777389514,
+      "loss": 1.0145,
+      "step": 9646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001164384312393855,
+      "loss": 1.1131,
+      "step": 9647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011642306430592391,
+      "loss": 1.1256,
+      "step": 9648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011640769697388336,
+      "loss": 1.0562,
+      "step": 9649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011639232924363674,
+      "loss": 1.2874,
+      "step": 9650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011637696111555707,
+      "loss": 1.0475,
+      "step": 9651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011636159259001726,
+      "loss": 1.1917,
+      "step": 9652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011634622366739038,
+      "loss": 1.0082,
+      "step": 9653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011633085434804938,
+      "loss": 1.0143,
+      "step": 9654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011631548463236724,
+      "loss": 1.1332,
+      "step": 9655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000116300114520717,
+      "loss": 1.0294,
+      "step": 9656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011628474401347171,
+      "loss": 1.1388,
+      "step": 9657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011626937311100437,
+      "loss": 0.9655,
+      "step": 9658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011625400181368804,
+      "loss": 1.1959,
+      "step": 9659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011623863012189575,
+      "loss": 1.0184,
+      "step": 9660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011622325803600061,
+      "loss": 1.1042,
+      "step": 9661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011620788555637563,
+      "loss": 1.0466,
+      "step": 9662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011619251268339395,
+      "loss": 1.1151,
+      "step": 9663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001161771394174286,
+      "loss": 1.2623,
+      "step": 9664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011616176575885276,
+      "loss": 1.2829,
+      "step": 9665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011614639170803947,
+      "loss": 1.2123,
+      "step": 9666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011613101726536186,
+      "loss": 1.006,
+      "step": 9667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011611564243119313,
+      "loss": 1.1587,
+      "step": 9668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011610026720590634,
+      "loss": 0.981,
+      "step": 9669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011608489158987464,
+      "loss": 1.1257,
+      "step": 9670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011606951558347125,
+      "loss": 1.1228,
+      "step": 9671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011605413918706927,
+      "loss": 0.8747,
+      "step": 9672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011603876240104192,
+      "loss": 1.0369,
+      "step": 9673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011602338522576236,
+      "loss": 1.1664,
+      "step": 9674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011600800766160382,
+      "loss": 1.2673,
+      "step": 9675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011599262970893949,
+      "loss": 0.9211,
+      "step": 9676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011597725136814257,
+      "loss": 1.0441,
+      "step": 9677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011596187263958627,
+      "loss": 1.0441,
+      "step": 9678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011594649352364386,
+      "loss": 1.0538,
+      "step": 9679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011593111402068857,
+      "loss": 1.075,
+      "step": 9680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011591573413109366,
+      "loss": 1.093,
+      "step": 9681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011590035385523239,
+      "loss": 0.9317,
+      "step": 9682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000115884973193478,
+      "loss": 1.1018,
+      "step": 9683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011586959214620383,
+      "loss": 1.0243,
+      "step": 9684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011585421071378308,
+      "loss": 1.2508,
+      "step": 9685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011583882889658913,
+      "loss": 1.1262,
+      "step": 9686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011582344669499528,
+      "loss": 1.1817,
+      "step": 9687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011580806410937482,
+      "loss": 1.0999,
+      "step": 9688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011579268114010109,
+      "loss": 1.1603,
+      "step": 9689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011577729778754743,
+      "loss": 1.1415,
+      "step": 9690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011576191405208714,
+      "loss": 1.1888,
+      "step": 9691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011574652993409368,
+      "loss": 1.1329,
+      "step": 9692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011573114543394027,
+      "loss": 1.2895,
+      "step": 9693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011571576055200039,
+      "loss": 0.973,
+      "step": 9694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011570037528864743,
+      "loss": 1.1586,
+      "step": 9695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011568498964425471,
+      "loss": 2.342,
+      "step": 9696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001156696036191957,
+      "loss": 1.0168,
+      "step": 9697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011565421721384377,
+      "loss": 1.1071,
+      "step": 9698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011563883042857235,
+      "loss": 1.1835,
+      "step": 9699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011562344326375487,
+      "loss": 1.128,
+      "step": 9700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011560805571976473,
+      "loss": 1.1872,
+      "step": 9701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011559266779697545,
+      "loss": 1.1107,
+      "step": 9702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011557727949576045,
+      "loss": 1.0399,
+      "step": 9703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001155618908164932,
+      "loss": 0.991,
+      "step": 9704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011554650175954713,
+      "loss": 1.0389,
+      "step": 9705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011553111232529584,
+      "loss": 1.1142,
+      "step": 9706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011551572251411267,
+      "loss": 1.1597,
+      "step": 9707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011550033232637126,
+      "loss": 2.1284,
+      "step": 9708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011548494176244503,
+      "loss": 1.1259,
+      "step": 9709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011546955082270752,
+      "loss": 1.1248,
+      "step": 9710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011545415950753231,
+      "loss": 1.1457,
+      "step": 9711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001154387678172929,
+      "loss": 1.0546,
+      "step": 9712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011542337575236284,
+      "loss": 1.2103,
+      "step": 9713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011540798331311568,
+      "loss": 1.2452,
+      "step": 9714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000115392590499925,
+      "loss": 1.0525,
+      "step": 9715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011537719731316436,
+      "loss": 1.2474,
+      "step": 9716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011536180375320735,
+      "loss": 1.0929,
+      "step": 9717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011534640982042757,
+      "loss": 1.2098,
+      "step": 9718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011533101551519865,
+      "loss": 1.1148,
+      "step": 9719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011531562083789413,
+      "loss": 1.1365,
+      "step": 9720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011530022578888768,
+      "loss": 1.1863,
+      "step": 9721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011528483036855296,
+      "loss": 1.0865,
+      "step": 9722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011526943457726354,
+      "loss": 2.1807,
+      "step": 9723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001152540384153931,
+      "loss": 1.1887,
+      "step": 9724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011523864188331528,
+      "loss": 1.0831,
+      "step": 9725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011522324498140379,
+      "loss": 2.1912,
+      "step": 9726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011520784771003225,
+      "loss": 2.2402,
+      "step": 9727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001151924500695744,
+      "loss": 1.2314,
+      "step": 9728
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011517705206040385,
+      "loss": 1.2366,
+      "step": 9729
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011516165368289443,
+      "loss": 1.0122,
+      "step": 9730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011514625493741971,
+      "loss": 1.132,
+      "step": 9731
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011513085582435347,
+      "loss": 1.0542,
+      "step": 9732
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011511545634406946,
+      "loss": 0.992,
+      "step": 9733
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011510005649694142,
+      "loss": 1.0737,
+      "step": 9734
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011508465628334304,
+      "loss": 1.019,
+      "step": 9735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011506925570364812,
+      "loss": 0.9707,
+      "step": 9736
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001150538547582304,
+      "loss": 1.1635,
+      "step": 9737
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001150384534474637,
+      "loss": 1.293,
+      "step": 9738
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011502305177172174,
+      "loss": 1.0181,
+      "step": 9739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011500764973137834,
+      "loss": 1.0754,
+      "step": 9740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011499224732680734,
+      "loss": 0.9476,
+      "step": 9741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011497684455838245,
+      "loss": 0.9491,
+      "step": 9742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011496144142647754,
+      "loss": 1.026,
+      "step": 9743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011494603793146646,
+      "loss": 1.3333,
+      "step": 9744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011493063407372302,
+      "loss": 1.0642,
+      "step": 9745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011491522985362106,
+      "loss": 1.1303,
+      "step": 9746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011489982527153442,
+      "loss": 1.0378,
+      "step": 9747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011488442032783702,
+      "loss": 1.1266,
+      "step": 9748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011486901502290268,
+      "loss": 0.9875,
+      "step": 9749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011485360935710527,
+      "loss": 2.0244,
+      "step": 9750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011483820333081867,
+      "loss": 1.1578,
+      "step": 9751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011482279694441685,
+      "loss": 1.0522,
+      "step": 9752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011480739019827362,
+      "loss": 1.0756,
+      "step": 9753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011479198309276298,
+      "loss": 1.0336,
+      "step": 9754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011477657562825879,
+      "loss": 1.2507,
+      "step": 9755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011476116780513502,
+      "loss": 1.2802,
+      "step": 9756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001147457596237656,
+      "loss": 2.1267,
+      "step": 9757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011473035108452442,
+      "loss": 1.0804,
+      "step": 9758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011471494218778549,
+      "loss": 1.1477,
+      "step": 9759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011469953293392282,
+      "loss": 1.0952,
+      "step": 9760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011468412332331029,
+      "loss": 1.0803,
+      "step": 9761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011466871335632196,
+      "loss": 1.1078,
+      "step": 9762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011465330303333179,
+      "loss": 1.1407,
+      "step": 9763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011463789235471378,
+      "loss": 1.0767,
+      "step": 9764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011462248132084195,
+      "loss": 1.289,
+      "step": 9765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011460706993209027,
+      "loss": 1.004,
+      "step": 9766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011459165818883282,
+      "loss": 1.2888,
+      "step": 9767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011457624609144365,
+      "loss": 1.3922,
+      "step": 9768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011456083364029673,
+      "loss": 0.985,
+      "step": 9769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011454542083576619,
+      "loss": 0.897,
+      "step": 9770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011453000767822603,
+      "loss": 0.9817,
+      "step": 9771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011451459416805035,
+      "loss": 1.1074,
+      "step": 9772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011449918030561324,
+      "loss": 1.0699,
+      "step": 9773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011448376609128875,
+      "loss": 1.2089,
+      "step": 9774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011446835152545099,
+      "loss": 1.1203,
+      "step": 9775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011445293660847407,
+      "loss": 1.2012,
+      "step": 9776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011443752134073209,
+      "loss": 1.17,
+      "step": 9777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011442210572259921,
+      "loss": 1.0389,
+      "step": 9778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011440668975444952,
+      "loss": 1.1572,
+      "step": 9779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011439127343665714,
+      "loss": 0.9707,
+      "step": 9780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011437585676959626,
+      "loss": 2.0391,
+      "step": 9781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000114360439753641,
+      "loss": 1.1545,
+      "step": 9782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011434502238916552,
+      "loss": 1.205,
+      "step": 9783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011432960467654403,
+      "loss": 1.1066,
+      "step": 9784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001143141866161507,
+      "loss": 1.0187,
+      "step": 9785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001142987682083597,
+      "loss": 1.4119,
+      "step": 9786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011428334945354522,
+      "loss": 1.1068,
+      "step": 9787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011426793035208148,
+      "loss": 1.0182,
+      "step": 9788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011425251090434271,
+      "loss": 1.0764,
+      "step": 9789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011423709111070309,
+      "loss": 1.0285,
+      "step": 9790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011422167097153687,
+      "loss": 1.117,
+      "step": 9791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011420625048721831,
+      "loss": 1.0388,
+      "step": 9792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011419082965812164,
+      "loss": 1.2285,
+      "step": 9793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001141754084846211,
+      "loss": 1.1763,
+      "step": 9794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011415998696709098,
+      "loss": 1.0727,
+      "step": 9795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011414456510590552,
+      "loss": 0.9673,
+      "step": 9796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011412914290143902,
+      "loss": 0.9316,
+      "step": 9797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011411372035406578,
+      "loss": 1.1439,
+      "step": 9798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011409829746416008,
+      "loss": 1.0536,
+      "step": 9799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011408287423209624,
+      "loss": 1.0918,
+      "step": 9800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011406745065824857,
+      "loss": 1.3723,
+      "step": 9801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011405202674299138,
+      "loss": 0.9696,
+      "step": 9802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000114036602486699,
+      "loss": 1.0373,
+      "step": 9803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011402117788974575,
+      "loss": 1.0636,
+      "step": 9804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011400575295250604,
+      "loss": 1.0748,
+      "step": 9805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011399032767535417,
+      "loss": 1.0728,
+      "step": 9806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011397490205866454,
+      "loss": 1.1627,
+      "step": 9807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011395947610281146,
+      "loss": 1.0559,
+      "step": 9808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001139440498081694,
+      "loss": 1.1625,
+      "step": 9809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011392862317511265,
+      "loss": 1.0818,
+      "step": 9810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001139131962040157,
+      "loss": 1.0963,
+      "step": 9811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011389776889525287,
+      "loss": 0.9814,
+      "step": 9812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011388234124919864,
+      "loss": 1.0737,
+      "step": 9813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011386691326622738,
+      "loss": 1.0593,
+      "step": 9814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011385148494671356,
+      "loss": 0.9398,
+      "step": 9815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011383605629103158,
+      "loss": 1.1215,
+      "step": 9816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011382062729955593,
+      "loss": 1.2379,
+      "step": 9817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011380519797266101,
+      "loss": 1.1662,
+      "step": 9818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011378976831072135,
+      "loss": 0.9604,
+      "step": 9819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011377433831411133,
+      "loss": 1.0938,
+      "step": 9820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011375890798320547,
+      "loss": 0.9906,
+      "step": 9821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011374347731837833,
+      "loss": 1.0981,
+      "step": 9822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00011372804632000429,
+      "loss": 1.1274,
+      "step": 9823
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011371261498845789,
+      "loss": 0.9988,
+      "step": 9824
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011369718332411365,
+      "loss": 0.9333,
+      "step": 9825
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001136817513273461,
+      "loss": 1.083,
+      "step": 9826
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011366631899852976,
+      "loss": 1.0354,
+      "step": 9827
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011365088633803913,
+      "loss": 1.1425,
+      "step": 9828
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011363545334624881,
+      "loss": 1.0272,
+      "step": 9829
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011362002002353331,
+      "loss": 1.1549,
+      "step": 9830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011360458637026718,
+      "loss": 1.1465,
+      "step": 9831
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011358915238682501,
+      "loss": 1.1525,
+      "step": 9832
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001135737180735814,
+      "loss": 1.0058,
+      "step": 9833
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011355828343091086,
+      "loss": 1.2445,
+      "step": 9834
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011354284845918806,
+      "loss": 1.1519,
+      "step": 9835
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011352741315878755,
+      "loss": 1.0942,
+      "step": 9836
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011351197753008393,
+      "loss": 1.1365,
+      "step": 9837
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011349654157345187,
+      "loss": 1.1476,
+      "step": 9838
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011348110528926592,
+      "loss": 1.1068,
+      "step": 9839
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011346566867790074,
+      "loss": 0.9899,
+      "step": 9840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000113450231739731,
+      "loss": 1.1489,
+      "step": 9841
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011343479447513128,
+      "loss": 1.1467,
+      "step": 9842
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011341935688447631,
+      "loss": 2.0232,
+      "step": 9843
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011340391896814073,
+      "loss": 1.0133,
+      "step": 9844
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011338848072649914,
+      "loss": 1.0062,
+      "step": 9845
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011337304215992635,
+      "loss": 1.1107,
+      "step": 9846
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001133576032687969,
+      "loss": 1.1586,
+      "step": 9847
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011334216405348557,
+      "loss": 1.0978,
+      "step": 9848
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011332672451436706,
+      "loss": 1.0956,
+      "step": 9849
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011331128465181605,
+      "loss": 1.1561,
+      "step": 9850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001132958444662073,
+      "loss": 1.1481,
+      "step": 9851
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011328040395791548,
+      "loss": 1.2755,
+      "step": 9852
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011326496312731533,
+      "loss": 1.2023,
+      "step": 9853
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011324952197478165,
+      "loss": 1.1136,
+      "step": 9854
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011323408050068912,
+      "loss": 1.1652,
+      "step": 9855
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011321863870541252,
+      "loss": 0.7936,
+      "step": 9856
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011320319658932665,
+      "loss": 2.1331,
+      "step": 9857
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011318775415280623,
+      "loss": 1.1429,
+      "step": 9858
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011317231139622606,
+      "loss": 1.1757,
+      "step": 9859
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011315686831996095,
+      "loss": 1.0814,
+      "step": 9860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011314142492438564,
+      "loss": 1.3271,
+      "step": 9861
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011312598120987499,
+      "loss": 1.1143,
+      "step": 9862
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011311053717680378,
+      "loss": 0.9121,
+      "step": 9863
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011309509282554682,
+      "loss": 1.0419,
+      "step": 9864
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011307964815647897,
+      "loss": 1.1043,
+      "step": 9865
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011306420316997505,
+      "loss": 1.0941,
+      "step": 9866
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011304875786640989,
+      "loss": 0.9203,
+      "step": 9867
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011303331224615835,
+      "loss": 1.1924,
+      "step": 9868
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011301786630959527,
+      "loss": 1.0651,
+      "step": 9869
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011300242005709551,
+      "loss": 1.0898,
+      "step": 9870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000112986973489034,
+      "loss": 1.0131,
+      "step": 9871
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011297152660578556,
+      "loss": 1.1222,
+      "step": 9872
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001129560794077251,
+      "loss": 1.0074,
+      "step": 9873
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011294063189522752,
+      "loss": 2.292,
+      "step": 9874
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001129251840686677,
+      "loss": 1.1718,
+      "step": 9875
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001129097359284206,
+      "loss": 1.1028,
+      "step": 9876
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011289428747486106,
+      "loss": 1.0868,
+      "step": 9877
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011287883870836407,
+      "loss": 2.0344,
+      "step": 9878
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011286338962930454,
+      "loss": 0.8929,
+      "step": 9879
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001128479402380574,
+      "loss": 1.0367,
+      "step": 9880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011283249053499765,
+      "loss": 0.9348,
+      "step": 9881
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011281704052050019,
+      "loss": 1.1377,
+      "step": 9882
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011280159019493999,
+      "loss": 0.975,
+      "step": 9883
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011278613955869206,
+      "loss": 1.0912,
+      "step": 9884
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011277068861213131,
+      "loss": 0.9508,
+      "step": 9885
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001127552373556328,
+      "loss": 1.0234,
+      "step": 9886
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011273978578957149,
+      "loss": 1.097,
+      "step": 9887
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001127243339143224,
+      "loss": 1.108,
+      "step": 9888
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011270888173026048,
+      "loss": 1.064,
+      "step": 9889
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011269342923776083,
+      "loss": 1.1117,
+      "step": 9890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001126779764371984,
+      "loss": 1.2217,
+      "step": 9891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011266252332894826,
+      "loss": 1.0291,
+      "step": 9892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011264706991338543,
+      "loss": 1.0914,
+      "step": 9893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011263161619088498,
+      "loss": 1.1126,
+      "step": 9894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011261616216182198,
+      "loss": 0.9364,
+      "step": 9895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011260070782657144,
+      "loss": 2.0938,
+      "step": 9896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011258525318550843,
+      "loss": 1.1041,
+      "step": 9897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011256979823900808,
+      "loss": 1.1721,
+      "step": 9898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011255434298744543,
+      "loss": 2.0999,
+      "step": 9899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011253888743119558,
+      "loss": 1.0829,
+      "step": 9900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011252343157063361,
+      "loss": 1.0983,
+      "step": 9901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011250797540613465,
+      "loss": 1.0935,
+      "step": 9902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011249251893807386,
+      "loss": 1.0575,
+      "step": 9903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011247706216682625,
+      "loss": 0.9557,
+      "step": 9904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011246160509276702,
+      "loss": 0.9838,
+      "step": 9905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011244614771627129,
+      "loss": 0.8013,
+      "step": 9906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001124306900377142,
+      "loss": 1.0876,
+      "step": 9907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011241523205747091,
+      "loss": 1.0443,
+      "step": 9908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011239977377591657,
+      "loss": 1.063,
+      "step": 9909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011238431519342633,
+      "loss": 0.8969,
+      "step": 9910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011236885631037541,
+      "loss": 1.0265,
+      "step": 9911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011235339712713893,
+      "loss": 0.9923,
+      "step": 9912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011233793764409207,
+      "loss": 1.0397,
+      "step": 9913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001123224778616101,
+      "loss": 1.1074,
+      "step": 9914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011230701778006813,
+      "loss": 1.1289,
+      "step": 9915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011229155739984145,
+      "loss": 1.1932,
+      "step": 9916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011227609672130523,
+      "loss": 0.9556,
+      "step": 9917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011226063574483469,
+      "loss": 1.142,
+      "step": 9918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011224517447080507,
+      "loss": 1.2002,
+      "step": 9919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011222971289959161,
+      "loss": 1.0401,
+      "step": 9920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011221425103156953,
+      "loss": 1.1419,
+      "step": 9921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011219878886711413,
+      "loss": 1.1089,
+      "step": 9922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011218332640660061,
+      "loss": 1.1666,
+      "step": 9923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001121678636504043,
+      "loss": 1.1583,
+      "step": 9924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011215240059890043,
+      "loss": 1.0107,
+      "step": 9925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011213693725246426,
+      "loss": 2.2139,
+      "step": 9926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00011212147361147113,
+      "loss": 1.2675,
+      "step": 9927
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011210600967629629,
+      "loss": 1.1711,
+      "step": 9928
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011209054544731507,
+      "loss": 1.1107,
+      "step": 9929
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011207508092490279,
+      "loss": 0.9907,
+      "step": 9930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011205961610943472,
+      "loss": 1.2396,
+      "step": 9931
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011204415100128622,
+      "loss": 1.0635,
+      "step": 9932
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011202868560083264,
+      "loss": 1.1621,
+      "step": 9933
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011201321990844922,
+      "loss": 1.2021,
+      "step": 9934
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001119977539245114,
+      "loss": 0.9667,
+      "step": 9935
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011198228764939451,
+      "loss": 0.9042,
+      "step": 9936
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011196682108347387,
+      "loss": 1.2288,
+      "step": 9937
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001119513542271249,
+      "loss": 1.1545,
+      "step": 9938
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011193588708072293,
+      "loss": 1.0785,
+      "step": 9939
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011192041964464339,
+      "loss": 1.0552,
+      "step": 9940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001119049519192616,
+      "loss": 1.1296,
+      "step": 9941
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.000111889483904953,
+      "loss": 1.0269,
+      "step": 9942
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011187401560209293,
+      "loss": 0.9539,
+      "step": 9943
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011185854701105692,
+      "loss": 2.1316,
+      "step": 9944
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011184307813222026,
+      "loss": 1.0765,
+      "step": 9945
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011182760896595843,
+      "loss": 1.0222,
+      "step": 9946
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011181213951264686,
+      "loss": 1.0615,
+      "step": 9947
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011179666977266094,
+      "loss": 0.9976,
+      "step": 9948
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011178119974637617,
+      "loss": 1.1548,
+      "step": 9949
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011176572943416796,
+      "loss": 1.1707,
+      "step": 9950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011175025883641177,
+      "loss": 0.9056,
+      "step": 9951
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011173478795348311,
+      "loss": 1.1472,
+      "step": 9952
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001117193167857574,
+      "loss": 1.2267,
+      "step": 9953
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011170384533361013,
+      "loss": 1.0089,
+      "step": 9954
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001116883735974168,
+      "loss": 1.2172,
+      "step": 9955
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011167290157755286,
+      "loss": 1.1567,
+      "step": 9956
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011165742927439384,
+      "loss": 1.0889,
+      "step": 9957
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011164195668831523,
+      "loss": 1.0732,
+      "step": 9958
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011162648381969254,
+      "loss": 1.1589,
+      "step": 9959
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011161101066890134,
+      "loss": 1.1668,
+      "step": 9960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011159553723631706,
+      "loss": 1.1371,
+      "step": 9961
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011158006352231534,
+      "loss": 1.0938,
+      "step": 9962
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011156458952727165,
+      "loss": 0.9382,
+      "step": 9963
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011154911525156151,
+      "loss": 1.2103,
+      "step": 9964
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011153364069556056,
+      "loss": 1.0969,
+      "step": 9965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011151816585964427,
+      "loss": 1.0577,
+      "step": 9966
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011150269074418824,
+      "loss": 1.1299,
+      "step": 9967
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011148721534956809,
+      "loss": 1.0934,
+      "step": 9968
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011147173967615936,
+      "loss": 0.9834,
+      "step": 9969
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001114562637243376,
+      "loss": 1.0159,
+      "step": 9970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011144078749447846,
+      "loss": 1.2082,
+      "step": 9971
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011142531098695752,
+      "loss": 1.1404,
+      "step": 9972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011140983420215038,
+      "loss": 1.1377,
+      "step": 9973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011139435714043265,
+      "loss": 1.0172,
+      "step": 9974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011137887980217995,
+      "loss": 0.96,
+      "step": 9975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011136340218776795,
+      "loss": 1.1861,
+      "step": 9976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011134792429757221,
+      "loss": 1.2097,
+      "step": 9977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011133244613196843,
+      "loss": 1.2313,
+      "step": 9978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011131696769133225,
+      "loss": 2.2883,
+      "step": 9979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011130148897603928,
+      "loss": 1.2396,
+      "step": 9980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001112860099864652,
+      "loss": 1.0649,
+      "step": 9981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001112705307229857,
+      "loss": 0.9869,
+      "step": 9982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011125505118597643,
+      "loss": 1.075,
+      "step": 9983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011123957137581313,
+      "loss": 1.0109,
+      "step": 9984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011122409129287139,
+      "loss": 1.1317,
+      "step": 9985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011120861093752695,
+      "loss": 1.1832,
+      "step": 9986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011119313031015552,
+      "loss": 1.2036,
+      "step": 9987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011117764941113278,
+      "loss": 0.9825,
+      "step": 9988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011116216824083448,
+      "loss": 2.1147,
+      "step": 9989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001111466867996363,
+      "loss": 1.9971,
+      "step": 9990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.000111131205087914,
+      "loss": 1.1389,
+      "step": 9991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001111157231060433,
+      "loss": 1.1414,
+      "step": 9992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011110024085439992,
+      "loss": 1.025,
+      "step": 9993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001110847583333596,
+      "loss": 1.131,
+      "step": 9994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011106927554329818,
+      "loss": 1.1872,
+      "step": 9995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011105379248459132,
+      "loss": 1.0747,
+      "step": 9996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011103830915761482,
+      "loss": 1.1105,
+      "step": 9997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011102282556274446,
+      "loss": 1.0196,
+      "step": 9998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.000111007341700356,
+      "loss": 1.1293,
+      "step": 9999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011099185757082522,
+      "loss": 1.1342,
+      "step": 10000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011097637317452796,
+      "loss": 1.0486,
+      "step": 10001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011096088851183997,
+      "loss": 1.0236,
+      "step": 10002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011094540358313707,
+      "loss": 1.3112,
+      "step": 10003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011092991838879507,
+      "loss": 1.0242,
+      "step": 10004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011091443292918979,
+      "loss": 1.025,
+      "step": 10005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011089894720469709,
+      "loss": 1.3002,
+      "step": 10006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001108834612156927,
+      "loss": 1.0984,
+      "step": 10007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011086797496255254,
+      "loss": 1.019,
+      "step": 10008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011085248844565246,
+      "loss": 1.004,
+      "step": 10009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011083700166536825,
+      "loss": 1.1527,
+      "step": 10010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011082151462207582,
+      "loss": 1.0655,
+      "step": 10011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011080602731615099,
+      "loss": 1.1704,
+      "step": 10012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011079053974796967,
+      "loss": 1.1972,
+      "step": 10013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011077505191790771,
+      "loss": 1.1747,
+      "step": 10014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011075956382634099,
+      "loss": 1.1236,
+      "step": 10015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011074407547364538,
+      "loss": 1.1299,
+      "step": 10016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011072858686019682,
+      "loss": 1.2277,
+      "step": 10017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011071309798637117,
+      "loss": 1.2017,
+      "step": 10018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011069760885254439,
+      "loss": 1.1162,
+      "step": 10019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011068211945909233,
+      "loss": 1.0807,
+      "step": 10020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011066662980639093,
+      "loss": 1.1667,
+      "step": 10021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011065113989481614,
+      "loss": 0.9737,
+      "step": 10022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011063564972474386,
+      "loss": 1.0155,
+      "step": 10023
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011062015929655003,
+      "loss": 1.0942,
+      "step": 10024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011060466861061063,
+      "loss": 1.3909,
+      "step": 10025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011058917766730157,
+      "loss": 1.0773,
+      "step": 10026
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011057368646699886,
+      "loss": 1.2505,
+      "step": 10027
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001105581950100784,
+      "loss": 1.1805,
+      "step": 10028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00011054270329691619,
+      "loss": 1.0789,
+      "step": 10029
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001105272113278882,
+      "loss": 1.0291,
+      "step": 10030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001105117191033704,
+      "loss": 1.0326,
+      "step": 10031
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001104962266237388,
+      "loss": 1.2271,
+      "step": 10032
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011048073388936939,
+      "loss": 1.2226,
+      "step": 10033
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011046524090063815,
+      "loss": 0.876,
+      "step": 10034
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011044974765792113,
+      "loss": 1.0483,
+      "step": 10035
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011043425416159432,
+      "loss": 0.9484,
+      "step": 10036
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011041876041203371,
+      "loss": 1.0803,
+      "step": 10037
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011040326640961536,
+      "loss": 0.9892,
+      "step": 10038
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001103877721547153,
+      "loss": 0.9662,
+      "step": 10039
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011037227764770952,
+      "loss": 1.1724,
+      "step": 10040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011035678288897413,
+      "loss": 1.0499,
+      "step": 10041
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011034128787888518,
+      "loss": 1.1857,
+      "step": 10042
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011032579261781864,
+      "loss": 0.9061,
+      "step": 10043
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011031029710615066,
+      "loss": 1.0009,
+      "step": 10044
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011029480134425727,
+      "loss": 0.9949,
+      "step": 10045
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011027930533251455,
+      "loss": 1.0005,
+      "step": 10046
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011026380907129859,
+      "loss": 1.0763,
+      "step": 10047
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011024831256098544,
+      "loss": 0.8855,
+      "step": 10048
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011023281580195125,
+      "loss": 0.9778,
+      "step": 10049
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011021731879457207,
+      "loss": 1.0578,
+      "step": 10050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000110201821539224,
+      "loss": 1.056,
+      "step": 10051
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011018632403628322,
+      "loss": 1.0237,
+      "step": 10052
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011017082628612574,
+      "loss": 0.9855,
+      "step": 10053
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011015532828912778,
+      "loss": 1.0353,
+      "step": 10054
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011013983004566539,
+      "loss": 1.0192,
+      "step": 10055
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011012433155611475,
+      "loss": 1.0604,
+      "step": 10056
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011010883282085205,
+      "loss": 1.0935,
+      "step": 10057
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011009333384025331,
+      "loss": 2.1677,
+      "step": 10058
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011007783461469474,
+      "loss": 1.1394,
+      "step": 10059
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011006233514455255,
+      "loss": 1.0921,
+      "step": 10060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011004683543020284,
+      "loss": 1.0139,
+      "step": 10061
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001100313354720218,
+      "loss": 1.0529,
+      "step": 10062
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011001583527038561,
+      "loss": 1.0172,
+      "step": 10063
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00011000033482567045,
+      "loss": 1.1189,
+      "step": 10064
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010998483413825248,
+      "loss": 1.0262,
+      "step": 10065
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010996933320850798,
+      "loss": 1.1413,
+      "step": 10066
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010995383203681302,
+      "loss": 1.043,
+      "step": 10067
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010993833062354393,
+      "loss": 0.9695,
+      "step": 10068
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010992282896907682,
+      "loss": 0.9389,
+      "step": 10069
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010990732707378795,
+      "loss": 1.1248,
+      "step": 10070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010989182493805357,
+      "loss": 1.2176,
+      "step": 10071
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010987632256224989,
+      "loss": 1.2488,
+      "step": 10072
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010986081994675311,
+      "loss": 1.1464,
+      "step": 10073
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001098453170919395,
+      "loss": 0.9902,
+      "step": 10074
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001098298139981853,
+      "loss": 1.0551,
+      "step": 10075
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010981431066586679,
+      "loss": 0.9095,
+      "step": 10076
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010979880709536017,
+      "loss": 1.0191,
+      "step": 10077
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010978330328704171,
+      "loss": 1.0965,
+      "step": 10078
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001097677992412878,
+      "loss": 1.2205,
+      "step": 10079
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010975229495847454,
+      "loss": 1.0691,
+      "step": 10080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001097367904389783,
+      "loss": 1.0179,
+      "step": 10081
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010972128568317539,
+      "loss": 1.0419,
+      "step": 10082
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010970578069144204,
+      "loss": 1.0521,
+      "step": 10083
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010969027546415458,
+      "loss": 1.0991,
+      "step": 10084
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001096747700016893,
+      "loss": 1.146,
+      "step": 10085
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010965926430442255,
+      "loss": 1.0797,
+      "step": 10086
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010964375837273061,
+      "loss": 1.1418,
+      "step": 10087
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010962825220698978,
+      "loss": 1.1667,
+      "step": 10088
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010961274580757642,
+      "loss": 1.3029,
+      "step": 10089
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010959723917486686,
+      "loss": 0.9039,
+      "step": 10090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010958173230923743,
+      "loss": 1.144,
+      "step": 10091
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010956622521106448,
+      "loss": 0.9786,
+      "step": 10092
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010955071788072433,
+      "loss": 0.8363,
+      "step": 10093
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010953521031859339,
+      "loss": 1.1141,
+      "step": 10094
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000109519702525048,
+      "loss": 1.0901,
+      "step": 10095
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010950419450046447,
+      "loss": 1.0338,
+      "step": 10096
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010948868624521921,
+      "loss": 1.0224,
+      "step": 10097
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010947317775968864,
+      "loss": 1.1708,
+      "step": 10098
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010945766904424907,
+      "loss": 0.9875,
+      "step": 10099
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010944216009927698,
+      "loss": 1.1028,
+      "step": 10100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010942665092514868,
+      "loss": 1.043,
+      "step": 10101
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010941114152224057,
+      "loss": 1.0854,
+      "step": 10102
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010939563189092911,
+      "loss": 1.0722,
+      "step": 10103
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010938012203159067,
+      "loss": 1.1171,
+      "step": 10104
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010936461194460169,
+      "loss": 1.0196,
+      "step": 10105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010934910163033856,
+      "loss": 0.9692,
+      "step": 10106
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010933359108917773,
+      "loss": 1.1517,
+      "step": 10107
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010931808032149564,
+      "loss": 1.154,
+      "step": 10108
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010930256932766874,
+      "loss": 1.2618,
+      "step": 10109
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001092870581080734,
+      "loss": 1.0601,
+      "step": 10110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010927154666308617,
+      "loss": 1.206,
+      "step": 10111
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010925603499308341,
+      "loss": 0.9105,
+      "step": 10112
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010924052309844164,
+      "loss": 1.0417,
+      "step": 10113
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010922501097953732,
+      "loss": 0.9951,
+      "step": 10114
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001092094986367469,
+      "loss": 0.8894,
+      "step": 10115
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010919398607044689,
+      "loss": 0.9653,
+      "step": 10116
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010917847328101374,
+      "loss": 1.1057,
+      "step": 10117
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010916296026882393,
+      "loss": 1.0011,
+      "step": 10118
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010914744703425399,
+      "loss": 0.9207,
+      "step": 10119
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010913193357768037,
+      "loss": 1.0731,
+      "step": 10120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010911641989947958,
+      "loss": 0.9552,
+      "step": 10121
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001091009060000282,
+      "loss": 1.1245,
+      "step": 10122
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010908539187970268,
+      "loss": 0.9798,
+      "step": 10123
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010906987753887954,
+      "loss": 1.0143,
+      "step": 10124
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010905436297793532,
+      "loss": 1.0668,
+      "step": 10125
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010903884819724655,
+      "loss": 1.1346,
+      "step": 10126
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010902333319718976,
+      "loss": 0.9568,
+      "step": 10127
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010900781797814147,
+      "loss": 1.1836,
+      "step": 10128
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010899230254047828,
+      "loss": 0.8781,
+      "step": 10129
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010897678688457671,
+      "loss": 1.0134,
+      "step": 10130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010896127101081331,
+      "loss": 0.9889,
+      "step": 10131
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010894575491956463,
+      "loss": 1.113,
+      "step": 10132
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010893023861120728,
+      "loss": 1.0096,
+      "step": 10133
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001089147220861178,
+      "loss": 1.1673,
+      "step": 10134
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010889920534467275,
+      "loss": 1.052,
+      "step": 10135
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010888368838724878,
+      "loss": 1.1299,
+      "step": 10136
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010886817121422241,
+      "loss": 0.8977,
+      "step": 10137
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001088526538259703,
+      "loss": 1.1602,
+      "step": 10138
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010883713622286899,
+      "loss": 1.1825,
+      "step": 10139
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010882161840529508,
+      "loss": 1.0296,
+      "step": 10140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010880610037362524,
+      "loss": 1.2024,
+      "step": 10141
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000108790582128236,
+      "loss": 1.0858,
+      "step": 10142
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010877506366950405,
+      "loss": 2.0171,
+      "step": 10143
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000108759544997806,
+      "loss": 0.9891,
+      "step": 10144
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010874402611351851,
+      "loss": 1.0955,
+      "step": 10145
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010872850701701811,
+      "loss": 1.1263,
+      "step": 10146
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010871298770868155,
+      "loss": 2.0867,
+      "step": 10147
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001086974681888854,
+      "loss": 1.1173,
+      "step": 10148
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010868194845800637,
+      "loss": 1.1703,
+      "step": 10149
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010866642851642106,
+      "loss": 1.143,
+      "step": 10150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010865090836450616,
+      "loss": 1.3666,
+      "step": 10151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010863538800263835,
+      "loss": 1.1133,
+      "step": 10152
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001086198674311943,
+      "loss": 1.068,
+      "step": 10153
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010860434665055064,
+      "loss": 1.1272,
+      "step": 10154
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001085888256610841,
+      "loss": 1.0239,
+      "step": 10155
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010857330446317133,
+      "loss": 1.1024,
+      "step": 10156
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010855778305718906,
+      "loss": 1.2361,
+      "step": 10157
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010854226144351397,
+      "loss": 0.9663,
+      "step": 10158
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010852673962252273,
+      "loss": 1.1516,
+      "step": 10159
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010851121759459213,
+      "loss": 1.0984,
+      "step": 10160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010849569536009876,
+      "loss": 1.0048,
+      "step": 10161
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010848017291941943,
+      "loss": 1.0603,
+      "step": 10162
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010846465027293082,
+      "loss": 0.8932,
+      "step": 10163
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010844912742100967,
+      "loss": 0.9542,
+      "step": 10164
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010843360436403273,
+      "loss": 0.9182,
+      "step": 10165
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001084180811023767,
+      "loss": 1.0947,
+      "step": 10166
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010840255763641836,
+      "loss": 0.8976,
+      "step": 10167
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010838703396653442,
+      "loss": 1.1473,
+      "step": 10168
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010837151009310165,
+      "loss": 0.9838,
+      "step": 10169
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010835598601649677,
+      "loss": 1.0315,
+      "step": 10170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010834046173709659,
+      "loss": 1.2209,
+      "step": 10171
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010832493725527788,
+      "loss": 0.9425,
+      "step": 10172
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010830941257141737,
+      "loss": 1.1395,
+      "step": 10173
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010829388768589186,
+      "loss": 1.2096,
+      "step": 10174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010827836259907813,
+      "loss": 1.1641,
+      "step": 10175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010826283731135296,
+      "loss": 1.1575,
+      "step": 10176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010824731182309314,
+      "loss": 1.05,
+      "step": 10177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010823178613467544,
+      "loss": 1.2502,
+      "step": 10178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010821626024647673,
+      "loss": 1.0765,
+      "step": 10179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010820073415887372,
+      "loss": 1.136,
+      "step": 10180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010818520787224334,
+      "loss": 2.0303,
+      "step": 10181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001081696813869623,
+      "loss": 1.0365,
+      "step": 10182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010815415470340747,
+      "loss": 0.9109,
+      "step": 10183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010813862782195565,
+      "loss": 1.0178,
+      "step": 10184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010812310074298367,
+      "loss": 1.0483,
+      "step": 10185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010810757346686835,
+      "loss": 1.1989,
+      "step": 10186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010809204599398661,
+      "loss": 1.1429,
+      "step": 10187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001080765183247152,
+      "loss": 0.9372,
+      "step": 10188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010806099045943102,
+      "loss": 0.8738,
+      "step": 10189
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001080454623985109,
+      "loss": 1.0495,
+      "step": 10190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010802993414233172,
+      "loss": 1.1008,
+      "step": 10191
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001080144056912703,
+      "loss": 1.0653,
+      "step": 10192
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010799887704570355,
+      "loss": 1.0475,
+      "step": 10193
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010798334820600831,
+      "loss": 1.105,
+      "step": 10194
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001079678191725615,
+      "loss": 0.9346,
+      "step": 10195
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010795228994573994,
+      "loss": 0.9443,
+      "step": 10196
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010793676052592059,
+      "loss": 1.0211,
+      "step": 10197
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010792123091348027,
+      "loss": 1.0955,
+      "step": 10198
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010790570110879592,
+      "loss": 1.0195,
+      "step": 10199
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010789017111224442,
+      "loss": 1.2175,
+      "step": 10200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001078746409242027,
+      "loss": 0.989,
+      "step": 10201
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010785911054504763,
+      "loss": 1.1395,
+      "step": 10202
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010784357997515615,
+      "loss": 0.885,
+      "step": 10203
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001078280492149052,
+      "loss": 1.0889,
+      "step": 10204
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010781251826467162,
+      "loss": 1.0562,
+      "step": 10205
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010779698712483243,
+      "loss": 0.9915,
+      "step": 10206
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010778145579576452,
+      "loss": 1.0579,
+      "step": 10207
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010776592427784485,
+      "loss": 1.0403,
+      "step": 10208
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010775039257145033,
+      "loss": 0.8629,
+      "step": 10209
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010773486067695791,
+      "loss": 1.0556,
+      "step": 10210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010771932859474458,
+      "loss": 0.9862,
+      "step": 10211
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010770379632518725,
+      "loss": 1.0445,
+      "step": 10212
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010768826386866291,
+      "loss": 1.0566,
+      "step": 10213
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001076727312255485,
+      "loss": 1.0481,
+      "step": 10214
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010765719839622102,
+      "loss": 1.1106,
+      "step": 10215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010764166538105738,
+      "loss": 1.1545,
+      "step": 10216
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010762613218043465,
+      "loss": 1.2225,
+      "step": 10217
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010761059879472975,
+      "loss": 1.204,
+      "step": 10218
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010759506522431968,
+      "loss": 1.2174,
+      "step": 10219
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010757953146958145,
+      "loss": 1.0749,
+      "step": 10220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010756399753089201,
+      "loss": 2.2114,
+      "step": 10221
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010754846340862841,
+      "loss": 1.1301,
+      "step": 10222
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010753292910316761,
+      "loss": 1.0228,
+      "step": 10223
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010751739461488666,
+      "loss": 1.0379,
+      "step": 10224
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010750185994416258,
+      "loss": 0.9547,
+      "step": 10225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010748632509137235,
+      "loss": 1.0134,
+      "step": 10226
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000107470790056893,
+      "loss": 1.1223,
+      "step": 10227
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010745525484110157,
+      "loss": 1.2935,
+      "step": 10228
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010743971944437507,
+      "loss": 0.9734,
+      "step": 10229
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010742418386709059,
+      "loss": 0.9545,
+      "step": 10230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001074086481096251,
+      "loss": 1.022,
+      "step": 10231
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010739311217235568,
+      "loss": 1.0513,
+      "step": 10232
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010737757605565942,
+      "loss": 1.1118,
+      "step": 10233
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010736203975991328,
+      "loss": 1.0631,
+      "step": 10234
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010734650328549439,
+      "loss": 1.123,
+      "step": 10235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010733096663277978,
+      "loss": 1.0392,
+      "step": 10236
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010731542980214651,
+      "loss": 1.0809,
+      "step": 10237
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001072998927939717,
+      "loss": 1.0093,
+      "step": 10238
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010728435560863237,
+      "loss": 1.1397,
+      "step": 10239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001072688182465056,
+      "loss": 0.9974,
+      "step": 10240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010725328070796856,
+      "loss": 1.1504,
+      "step": 10241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010723774299339821,
+      "loss": 1.1704,
+      "step": 10242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010722220510317171,
+      "loss": 1.0103,
+      "step": 10243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010720666703766618,
+      "loss": 2.2654,
+      "step": 10244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010719112879725865,
+      "loss": 1.2494,
+      "step": 10245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001071755903823263,
+      "loss": 2.3105,
+      "step": 10246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010716005179324618,
+      "loss": 0.9996,
+      "step": 10247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010714451303039546,
+      "loss": 0.928,
+      "step": 10248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010712897409415121,
+      "loss": 1.1472,
+      "step": 10249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010711343498489054,
+      "loss": 1.1175,
+      "step": 10250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001070978957029906,
+      "loss": 1.0103,
+      "step": 10251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010708235624882856,
+      "loss": 0.9443,
+      "step": 10252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010706681662278149,
+      "loss": 0.9655,
+      "step": 10253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010705127682522658,
+      "loss": 1.1369,
+      "step": 10254
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010703573685654092,
+      "loss": 2.1073,
+      "step": 10255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001070201967171017,
+      "loss": 1.0448,
+      "step": 10256
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010700465640728608,
+      "loss": 2.3254,
+      "step": 10257
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010698911592747114,
+      "loss": 0.9609,
+      "step": 10258
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010697357527803409,
+      "loss": 0.9724,
+      "step": 10259
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010695803445935213,
+      "loss": 1.0404,
+      "step": 10260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010694249347180237,
+      "loss": 1.1546,
+      "step": 10261
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010692695231576201,
+      "loss": 1.1135,
+      "step": 10262
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010691141099160821,
+      "loss": 0.9634,
+      "step": 10263
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010689586949971817,
+      "loss": 1.1767,
+      "step": 10264
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010688032784046907,
+      "loss": 0.9794,
+      "step": 10265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010686478601423804,
+      "loss": 1.049,
+      "step": 10266
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010684924402140233,
+      "loss": 1.1372,
+      "step": 10267
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010683370186233916,
+      "loss": 1.3324,
+      "step": 10268
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010681815953742566,
+      "loss": 1.1399,
+      "step": 10269
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010680261704703911,
+      "loss": 1.1195,
+      "step": 10270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010678707439155667,
+      "loss": 0.9416,
+      "step": 10271
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010677153157135553,
+      "loss": 1.1298,
+      "step": 10272
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010675598858681295,
+      "loss": 1.1525,
+      "step": 10273
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010674044543830615,
+      "loss": 2.1941,
+      "step": 10274
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010672490212621232,
+      "loss": 1.0291,
+      "step": 10275
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010670935865090869,
+      "loss": 1.1071,
+      "step": 10276
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010669381501277254,
+      "loss": 1.0667,
+      "step": 10277
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010667827121218107,
+      "loss": 0.9909,
+      "step": 10278
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010666272724951154,
+      "loss": 1.1397,
+      "step": 10279
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010664718312514115,
+      "loss": 1.059,
+      "step": 10280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010663163883944718,
+      "loss": 1.0688,
+      "step": 10281
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010661609439280688,
+      "loss": 0.8804,
+      "step": 10282
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001066005497855975,
+      "loss": 1.0866,
+      "step": 10283
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010658500501819633,
+      "loss": 0.9651,
+      "step": 10284
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010656946009098059,
+      "loss": 1.0539,
+      "step": 10285
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010655391500432754,
+      "loss": 1.0984,
+      "step": 10286
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010653836975861449,
+      "loss": 1.0147,
+      "step": 10287
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001065228243542187,
+      "loss": 1.0192,
+      "step": 10288
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010650727879151743,
+      "loss": 1.0619,
+      "step": 10289
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.000106491733070888,
+      "loss": 1.0561,
+      "step": 10290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010647618719270767,
+      "loss": 0.9583,
+      "step": 10291
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010646064115735372,
+      "loss": 1.0924,
+      "step": 10292
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001064450949652035,
+      "loss": 1.0343,
+      "step": 10293
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010642954861663424,
+      "loss": 1.0454,
+      "step": 10294
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010641400211202329,
+      "loss": 1.0675,
+      "step": 10295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001063984554517479,
+      "loss": 1.1605,
+      "step": 10296
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001063829086361854,
+      "loss": 1.0104,
+      "step": 10297
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010636736166571315,
+      "loss": 0.946,
+      "step": 10298
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010635181454070841,
+      "loss": 1.0173,
+      "step": 10299
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010633626726154858,
+      "loss": 1.0717,
+      "step": 10300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010632071982861089,
+      "loss": 1.0508,
+      "step": 10301
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001063051722422727,
+      "loss": 1.2516,
+      "step": 10302
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010628962450291135,
+      "loss": 1.1863,
+      "step": 10303
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010627407661090418,
+      "loss": 0.9431,
+      "step": 10304
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001062585285666285,
+      "loss": 1.0977,
+      "step": 10305
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001062429803704617,
+      "loss": 1.1216,
+      "step": 10306
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010622743202278111,
+      "loss": 1.0143,
+      "step": 10307
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010621188352396405,
+      "loss": 1.0149,
+      "step": 10308
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010619633487438789,
+      "loss": 0.961,
+      "step": 10309
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010618078607443001,
+      "loss": 1.0533,
+      "step": 10310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010616523712446774,
+      "loss": 1.1801,
+      "step": 10311
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010614968802487846,
+      "loss": 1.1306,
+      "step": 10312
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010613413877603953,
+      "loss": 0.8119,
+      "step": 10313
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010611858937832834,
+      "loss": 0.914,
+      "step": 10314
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010610303983212226,
+      "loss": 1.1216,
+      "step": 10315
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010608749013779864,
+      "loss": 1.2482,
+      "step": 10316
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001060719402957349,
+      "loss": 1.0714,
+      "step": 10317
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010605639030630841,
+      "loss": 1.0556,
+      "step": 10318
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010604084016989657,
+      "loss": 0.8867,
+      "step": 10319
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010602528988687674,
+      "loss": 0.9962,
+      "step": 10320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010600973945762635,
+      "loss": 1.2333,
+      "step": 10321
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010599418888252284,
+      "loss": 1.0508,
+      "step": 10322
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010597863816194352,
+      "loss": 1.1484,
+      "step": 10323
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010596308729626585,
+      "loss": 1.0065,
+      "step": 10324
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010594753628586724,
+      "loss": 0.9307,
+      "step": 10325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001059319851311251,
+      "loss": 1.2552,
+      "step": 10326
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010591643383241687,
+      "loss": 0.9859,
+      "step": 10327
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010590088239011993,
+      "loss": 1.081,
+      "step": 10328
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010588533080461171,
+      "loss": 0.9745,
+      "step": 10329
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001058697790762697,
+      "loss": 1.0273,
+      "step": 10330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010585422720547124,
+      "loss": 0.9944,
+      "step": 10331
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010583867519259382,
+      "loss": 1.0779,
+      "step": 10332
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010582312303801491,
+      "loss": 0.8751,
+      "step": 10333
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010580757074211187,
+      "loss": 1.0953,
+      "step": 10334
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010579201830526222,
+      "loss": 1.123,
+      "step": 10335
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010577646572784339,
+      "loss": 0.9699,
+      "step": 10336
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010576091301023278,
+      "loss": 1.0508,
+      "step": 10337
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001057453601528079,
+      "loss": 2.0978,
+      "step": 10338
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010572980715594622,
+      "loss": 1.0605,
+      "step": 10339
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010571425402002515,
+      "loss": 1.1049,
+      "step": 10340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010569870074542224,
+      "loss": 1.0841,
+      "step": 10341
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010568314733251482,
+      "loss": 0.9791,
+      "step": 10342
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001056675937816805,
+      "loss": 1.106,
+      "step": 10343
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010565204009329675,
+      "loss": 1.2609,
+      "step": 10344
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010563648626774095,
+      "loss": 1.157,
+      "step": 10345
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010562093230539063,
+      "loss": 1.075,
+      "step": 10346
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010560537820662333,
+      "loss": 1.0711,
+      "step": 10347
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010558982397181644,
+      "loss": 1.0809,
+      "step": 10348
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010557426960134756,
+      "loss": 1.0276,
+      "step": 10349
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001055587150955941,
+      "loss": 0.942,
+      "step": 10350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010554316045493362,
+      "loss": 1.084,
+      "step": 10351
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010552760567974358,
+      "loss": 1.0843,
+      "step": 10352
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010551205077040149,
+      "loss": 1.2382,
+      "step": 10353
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010549649572728489,
+      "loss": 0.9478,
+      "step": 10354
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010548094055077129,
+      "loss": 1.1291,
+      "step": 10355
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010546538524123814,
+      "loss": 1.0765,
+      "step": 10356
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010544982979906307,
+      "loss": 0.9644,
+      "step": 10357
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010543427422462351,
+      "loss": 2.1057,
+      "step": 10358
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010541871851829701,
+      "loss": 2.1882,
+      "step": 10359
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010540316268046114,
+      "loss": 0.9866,
+      "step": 10360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010538760671149334,
+      "loss": 1.1599,
+      "step": 10361
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010537205061177122,
+      "loss": 1.1375,
+      "step": 10362
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010535649438167234,
+      "loss": 0.9918,
+      "step": 10363
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010534093802157417,
+      "loss": 1.1546,
+      "step": 10364
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001053253815318543,
+      "loss": 0.9446,
+      "step": 10365
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010530982491289028,
+      "loss": 0.9766,
+      "step": 10366
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001052942681650596,
+      "loss": 0.8663,
+      "step": 10367
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001052787112887399,
+      "loss": 1.2103,
+      "step": 10368
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010526315428430866,
+      "loss": 1.0435,
+      "step": 10369
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010524759715214349,
+      "loss": 1.1199,
+      "step": 10370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010523203989262196,
+      "loss": 1.2061,
+      "step": 10371
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001052164825061216,
+      "loss": 1.1212,
+      "step": 10372
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010520092499302001,
+      "loss": 1.0433,
+      "step": 10373
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010518536735369475,
+      "loss": 1.0578,
+      "step": 10374
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010516980958852336,
+      "loss": 1.0624,
+      "step": 10375
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010515425169788349,
+      "loss": 1.0192,
+      "step": 10376
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010513869368215266,
+      "loss": 1.0558,
+      "step": 10377
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010512313554170848,
+      "loss": 1.2091,
+      "step": 10378
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010510757727692855,
+      "loss": 1.1851,
+      "step": 10379
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010509201888819045,
+      "loss": 0.9333,
+      "step": 10380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010507646037587175,
+      "loss": 1.1282,
+      "step": 10381
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010506090174035009,
+      "loss": 1.0133,
+      "step": 10382
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010504534298200301,
+      "loss": 1.0763,
+      "step": 10383
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010502978410120821,
+      "loss": 1.0418,
+      "step": 10384
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010501422509834317,
+      "loss": 1.0526,
+      "step": 10385
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001049986659737856,
+      "loss": 1.0085,
+      "step": 10386
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001049831067279131,
+      "loss": 0.9697,
+      "step": 10387
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010496754736110323,
+      "loss": 1.0814,
+      "step": 10388
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010495198787373364,
+      "loss": 1.1271,
+      "step": 10389
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010493642826618197,
+      "loss": 1.1915,
+      "step": 10390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010492086853882579,
+      "loss": 1.0501,
+      "step": 10391
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001049053086920428,
+      "loss": 0.9218,
+      "step": 10392
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010488974872621055,
+      "loss": 1.0403,
+      "step": 10393
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001048741886417067,
+      "loss": 1.104,
+      "step": 10394
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010485862843890896,
+      "loss": 1.7992,
+      "step": 10395
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010484306811819487,
+      "loss": 1.4527,
+      "step": 10396
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010482750767994208,
+      "loss": 0.9733,
+      "step": 10397
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010481194712452828,
+      "loss": 0.8505,
+      "step": 10398
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010479638645233108,
+      "loss": 0.8415,
+      "step": 10399
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010478082566372815,
+      "loss": 0.9787,
+      "step": 10400
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010476526475909714,
+      "loss": 1.0643,
+      "step": 10401
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010474970373881567,
+      "loss": 0.9092,
+      "step": 10402
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010473414260326146,
+      "loss": 0.9526,
+      "step": 10403
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010471858135281215,
+      "loss": 0.8269,
+      "step": 10404
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010470301998784535,
+      "loss": 0.8636,
+      "step": 10405
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001046874585087388,
+      "loss": 0.8906,
+      "step": 10406
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001046718969158701,
+      "loss": 0.984,
+      "step": 10407
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010465633520961698,
+      "loss": 0.9082,
+      "step": 10408
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010464077339035709,
+      "loss": 0.9012,
+      "step": 10409
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010462521145846811,
+      "loss": 0.9806,
+      "step": 10410
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010460964941432772,
+      "loss": 0.9855,
+      "step": 10411
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010459408725831361,
+      "loss": 1.0049,
+      "step": 10412
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010457852499080343,
+      "loss": 0.9661,
+      "step": 10413
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010456296261217491,
+      "loss": 1.0367,
+      "step": 10414
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010454740012280571,
+      "loss": 0.7758,
+      "step": 10415
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010453183752307354,
+      "loss": 0.9255,
+      "step": 10416
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010451627481335613,
+      "loss": 0.9035,
+      "step": 10417
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010450071199403111,
+      "loss": 0.9286,
+      "step": 10418
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010448514906547622,
+      "loss": 0.938,
+      "step": 10419
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010446958602806915,
+      "loss": 1.5883,
+      "step": 10420
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010445402288218761,
+      "loss": 1.1549,
+      "step": 10421
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010443845962820935,
+      "loss": 0.9995,
+      "step": 10422
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.000104422896266512,
+      "loss": 0.9035,
+      "step": 10423
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010440733279747333,
+      "loss": 1.0023,
+      "step": 10424
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010439176922147109,
+      "loss": 0.8636,
+      "step": 10425
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010437620553888292,
+      "loss": 0.9639,
+      "step": 10426
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010436064175008657,
+      "loss": 0.8516,
+      "step": 10427
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001043450778554598,
+      "loss": 0.9915,
+      "step": 10428
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010432951385538028,
+      "loss": 0.9146,
+      "step": 10429
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010431394975022579,
+      "loss": 0.9831,
+      "step": 10430
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010429838554037404,
+      "loss": 0.7657,
+      "step": 10431
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010428282122620279,
+      "loss": 1.0136,
+      "step": 10432
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010426725680808974,
+      "loss": 0.9906,
+      "step": 10433
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010425169228641263,
+      "loss": 0.7523,
+      "step": 10434
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010423612766154922,
+      "loss": 0.9772,
+      "step": 10435
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010422056293387728,
+      "loss": 0.9136,
+      "step": 10436
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010420499810377451,
+      "loss": 1.4943,
+      "step": 10437
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001041894331716187,
+      "loss": 1.2933,
+      "step": 10438
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010417386813778758,
+      "loss": 0.8743,
+      "step": 10439
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001041583030026589,
+      "loss": 0.9095,
+      "step": 10440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010414273776661045,
+      "loss": 0.9227,
+      "step": 10441
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010412717243001993,
+      "loss": 0.954,
+      "step": 10442
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010411160699326515,
+      "loss": 0.8365,
+      "step": 10443
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010409604145672389,
+      "loss": 0.7874,
+      "step": 10444
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010408047582077385,
+      "loss": 0.9992,
+      "step": 10445
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010406491008579289,
+      "loss": 1.0105,
+      "step": 10446
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010404934425215871,
+      "loss": 0.8638,
+      "step": 10447
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001040337783202491,
+      "loss": 1.452,
+      "step": 10448
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010401821229044186,
+      "loss": 1.8796,
+      "step": 10449
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010400264616311473,
+      "loss": 1.1799,
+      "step": 10450
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010398707993864551,
+      "loss": 0.7783,
+      "step": 10451
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010397151361741204,
+      "loss": 0.9127,
+      "step": 10452
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000103955947199792,
+      "loss": 0.9887,
+      "step": 10453
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010394038068616327,
+      "loss": 0.9078,
+      "step": 10454
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010392481407690362,
+      "loss": 1.5485,
+      "step": 10455
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010390924737239078,
+      "loss": 1.3318,
+      "step": 10456
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010389368057300263,
+      "loss": 0.9748,
+      "step": 10457
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010387811367911693,
+      "loss": 0.955,
+      "step": 10458
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010386254669111144,
+      "loss": 1.0037,
+      "step": 10459
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010384697960936407,
+      "loss": 0.9946,
+      "step": 10460
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010383141243425251,
+      "loss": 0.8592,
+      "step": 10461
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010381584516615464,
+      "loss": 1.0125,
+      "step": 10462
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010380027780544826,
+      "loss": 0.9305,
+      "step": 10463
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010378471035251111,
+      "loss": 0.8946,
+      "step": 10464
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010376914280772112,
+      "loss": 0.9489,
+      "step": 10465
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010375357517145603,
+      "loss": 1.3106,
+      "step": 10466
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010373800744409366,
+      "loss": 1.8672,
+      "step": 10467
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010372243962601184,
+      "loss": 1.1937,
+      "step": 10468
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010370687171758844,
+      "loss": 0.8546,
+      "step": 10469
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010369130371920122,
+      "loss": 0.9641,
+      "step": 10470
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010367573563122804,
+      "loss": 0.8694,
+      "step": 10471
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001036601674540467,
+      "loss": 0.8687,
+      "step": 10472
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010364459918803504,
+      "loss": 1.0491,
+      "step": 10473
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010362903083357092,
+      "loss": 0.9575,
+      "step": 10474
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010361346239103216,
+      "loss": 0.978,
+      "step": 10475
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010359789386079661,
+      "loss": 0.8632,
+      "step": 10476
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001035823252432421,
+      "loss": 0.7325,
+      "step": 10477
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010356675653874646,
+      "loss": 0.9174,
+      "step": 10478
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010355118774768755,
+      "loss": 0.8795,
+      "step": 10479
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010353561887044319,
+      "loss": 1.0342,
+      "step": 10480
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010352004990739123,
+      "loss": 0.9636,
+      "step": 10481
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010350448085890958,
+      "loss": 1.0577,
+      "step": 10482
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010348891172537607,
+      "loss": 1.5344,
+      "step": 10483
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010347334250716849,
+      "loss": 1.8872,
+      "step": 10484
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010345777320466476,
+      "loss": 1.286,
+      "step": 10485
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010344220381824272,
+      "loss": 0.9575,
+      "step": 10486
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010342663434828024,
+      "loss": 0.9434,
+      "step": 10487
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010341106479515517,
+      "loss": 0.9843,
+      "step": 10488
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010339549515924536,
+      "loss": 0.8618,
+      "step": 10489
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010337992544092871,
+      "loss": 0.7931,
+      "step": 10490
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010336435564058309,
+      "loss": 0.9406,
+      "step": 10491
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010334878575858633,
+      "loss": 0.8233,
+      "step": 10492
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010333321579531636,
+      "loss": 0.8436,
+      "step": 10493
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000103317645751151,
+      "loss": 0.9619,
+      "step": 10494
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010330207562646814,
+      "loss": 0.9869,
+      "step": 10495
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010328650542164568,
+      "loss": 1.0456,
+      "step": 10496
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010327093513706147,
+      "loss": 0.957,
+      "step": 10497
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010325536477309347,
+      "loss": 0.9812,
+      "step": 10498
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010323979433011945,
+      "loss": 0.8803,
+      "step": 10499
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010322422380851737,
+      "loss": 1.0949,
+      "step": 10500
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001032086532086651,
+      "loss": 0.9312,
+      "step": 10501
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010319308253094054,
+      "loss": 0.9631,
+      "step": 10502
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010317751177572155,
+      "loss": 1.0182,
+      "step": 10503
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010316194094338604,
+      "loss": 0.8746,
+      "step": 10504
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010314637003431195,
+      "loss": 0.8242,
+      "step": 10505
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010313079904887712,
+      "loss": 1.4159,
+      "step": 10506
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010311522798745947,
+      "loss": 1.1976,
+      "step": 10507
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001030996568504369,
+      "loss": 0.9736,
+      "step": 10508
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010308408563818733,
+      "loss": 0.7128,
+      "step": 10509
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001030685143510886,
+      "loss": 1.0072,
+      "step": 10510
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010305294298951871,
+      "loss": 0.948,
+      "step": 10511
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010303737155385551,
+      "loss": 0.9363,
+      "step": 10512
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010302180004447692,
+      "loss": 1.0021,
+      "step": 10513
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010300622846176089,
+      "loss": 1.0483,
+      "step": 10514
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010299065680608526,
+      "loss": 0.9075,
+      "step": 10515
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010297508507782798,
+      "loss": 0.9725,
+      "step": 10516
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010295951327736701,
+      "loss": 0.957,
+      "step": 10517
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010294394140508021,
+      "loss": 1.0103,
+      "step": 10518
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010292836946134554,
+      "loss": 0.9001,
+      "step": 10519
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010291279744654092,
+      "loss": 0.9707,
+      "step": 10520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010289722536104423,
+      "loss": 1.0101,
+      "step": 10521
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010288165320523347,
+      "loss": 0.9488,
+      "step": 10522
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010286608097948646,
+      "loss": 0.821,
+      "step": 10523
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010285050868418122,
+      "loss": 0.7472,
+      "step": 10524
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001028349363196957,
+      "loss": 0.8907,
+      "step": 10525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010281936388640773,
+      "loss": 0.9238,
+      "step": 10526
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010280379138469536,
+      "loss": 0.9815,
+      "step": 10527
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010278821881493643,
+      "loss": 0.8215,
+      "step": 10528
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010277264617750892,
+      "loss": 0.9473,
+      "step": 10529
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001027570734727908,
+      "loss": 0.8454,
+      "step": 10530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010274150070115995,
+      "loss": 1.0316,
+      "step": 10531
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010272592786299433,
+      "loss": 0.9623,
+      "step": 10532
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010271035495867194,
+      "loss": 1.0508,
+      "step": 10533
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010269478198857065,
+      "loss": 1.3986,
+      "step": 10534
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010267920895306847,
+      "loss": 1.2574,
+      "step": 10535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010266363585254331,
+      "loss": 0.8209,
+      "step": 10536
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001026480626873731,
+      "loss": 0.867,
+      "step": 10537
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010263248945793584,
+      "loss": 0.8604,
+      "step": 10538
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010261691616460946,
+      "loss": 1.3889,
+      "step": 10539
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010260134280777194,
+      "loss": 1.2397,
+      "step": 10540
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010258576938780123,
+      "loss": 0.9644,
+      "step": 10541
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010257019590507526,
+      "loss": 0.9538,
+      "step": 10542
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000102554622359972,
+      "loss": 0.8974,
+      "step": 10543
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010253904875286943,
+      "loss": 0.9186,
+      "step": 10544
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001025234750841455,
+      "loss": 0.7678,
+      "step": 10545
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010250790135417816,
+      "loss": 0.9758,
+      "step": 10546
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010249232756334543,
+      "loss": 0.9255,
+      "step": 10547
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001024767537120252,
+      "loss": 0.9628,
+      "step": 10548
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010246117980059553,
+      "loss": 0.9972,
+      "step": 10549
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010244560582943432,
+      "loss": 1.4306,
+      "step": 10550
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010243003179891954,
+      "loss": 1.2254,
+      "step": 10551
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010241445770942924,
+      "loss": 1.5038,
+      "step": 10552
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010239888356134128,
+      "loss": 1.1955,
+      "step": 10553
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010238330935503371,
+      "loss": 0.9907,
+      "step": 10554
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010236773509088453,
+      "loss": 0.8472,
+      "step": 10555
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010235216076927163,
+      "loss": 0.878,
+      "step": 10556
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010233658639057309,
+      "loss": 0.938,
+      "step": 10557
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010232101195516686,
+      "loss": 1.013,
+      "step": 10558
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010230543746343087,
+      "loss": 0.8894,
+      "step": 10559
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010228986291574316,
+      "loss": 0.7578,
+      "step": 10560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010227428831248167,
+      "loss": 0.8885,
+      "step": 10561
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010225871365402443,
+      "loss": 1.0654,
+      "step": 10562
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010224313894074944,
+      "loss": 0.7458,
+      "step": 10563
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010222756417303467,
+      "loss": 0.9783,
+      "step": 10564
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010221198935125809,
+      "loss": 0.8374,
+      "step": 10565
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010219641447579771,
+      "loss": 0.8328,
+      "step": 10566
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010218083954703153,
+      "loss": 1.1385,
+      "step": 10567
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010216526456533754,
+      "loss": 0.9131,
+      "step": 10568
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010214968953109372,
+      "loss": 0.9941,
+      "step": 10569
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010213411444467808,
+      "loss": 0.9213,
+      "step": 10570
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010211853930646864,
+      "loss": 0.9434,
+      "step": 10571
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010210296411684339,
+      "loss": 1.0294,
+      "step": 10572
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001020873888761803,
+      "loss": 0.7651,
+      "step": 10573
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010207181358485742,
+      "loss": 0.8577,
+      "step": 10574
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010205623824325272,
+      "loss": 1.0352,
+      "step": 10575
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010204066285174423,
+      "loss": 0.927,
+      "step": 10576
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010202508741070991,
+      "loss": 1.0026,
+      "step": 10577
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010200951192052781,
+      "loss": 0.744,
+      "step": 10578
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010199393638157597,
+      "loss": 0.8909,
+      "step": 10579
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001019783607942323,
+      "loss": 0.8616,
+      "step": 10580
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001019627851588749,
+      "loss": 1.0261,
+      "step": 10581
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010194720947588177,
+      "loss": 1.5942,
+      "step": 10582
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010193163374563087,
+      "loss": 1.2766,
+      "step": 10583
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010191605796850028,
+      "loss": 1.0314,
+      "step": 10584
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010190048214486796,
+      "loss": 0.8939,
+      "step": 10585
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010188490627511195,
+      "loss": 1.0326,
+      "step": 10586
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010186933035961031,
+      "loss": 0.9005,
+      "step": 10587
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010185375439874099,
+      "loss": 0.884,
+      "step": 10588
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010183817839288203,
+      "loss": 0.9122,
+      "step": 10589
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001018226023424115,
+      "loss": 1.0497,
+      "step": 10590
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010180702624770735,
+      "loss": 1.0709,
+      "step": 10591
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010179145010914768,
+      "loss": 0.9461,
+      "step": 10592
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010177587392711043,
+      "loss": 1.0666,
+      "step": 10593
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010176029770197368,
+      "loss": 0.9389,
+      "step": 10594
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010174472143411547,
+      "loss": 0.8615,
+      "step": 10595
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010172914512391378,
+      "loss": 0.8754,
+      "step": 10596
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010171356877174665,
+      "loss": 1.4935,
+      "step": 10597
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010169799237799213,
+      "loss": 1.2027,
+      "step": 10598
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010168241594302825,
+      "loss": 1.0812,
+      "step": 10599
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010166683946723303,
+      "loss": 1.3318,
+      "step": 10600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010165126295098453,
+      "loss": 1.1544,
+      "step": 10601
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010163568639466073,
+      "loss": 0.8906,
+      "step": 10602
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001016201097986397,
+      "loss": 0.8252,
+      "step": 10603
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010160453316329945,
+      "loss": 0.9196,
+      "step": 10604
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010158895648901807,
+      "loss": 0.8193,
+      "step": 10605
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010157337977617356,
+      "loss": 0.9945,
+      "step": 10606
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010155780302514394,
+      "loss": 0.8499,
+      "step": 10607
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010154222623630731,
+      "loss": 0.827,
+      "step": 10608
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010152664941004167,
+      "loss": 0.8852,
+      "step": 10609
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010151107254672503,
+      "loss": 0.9546,
+      "step": 10610
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010149549564673546,
+      "loss": 1.0214,
+      "step": 10611
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010147991871045104,
+      "loss": 0.8334,
+      "step": 10612
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010146434173824976,
+      "loss": 1.0119,
+      "step": 10613
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010144876473050973,
+      "loss": 1.0271,
+      "step": 10614
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010143318768760892,
+      "loss": 0.8342,
+      "step": 10615
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010141761060992542,
+      "loss": 0.9828,
+      "step": 10616
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010140203349783728,
+      "loss": 0.7619,
+      "step": 10617
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001013864563517225,
+      "loss": 0.8748,
+      "step": 10618
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010137087917195918,
+      "loss": 1.0427,
+      "step": 10619
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010135530195892537,
+      "loss": 0.8347,
+      "step": 10620
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010133972471299908,
+      "loss": 0.8965,
+      "step": 10621
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010132414743455843,
+      "loss": 0.9183,
+      "step": 10622
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001013085701239814,
+      "loss": 0.955,
+      "step": 10623
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010129299278164608,
+      "loss": 0.9308,
+      "step": 10624
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010127741540793052,
+      "loss": 0.9631,
+      "step": 10625
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010126183800321274,
+      "loss": 0.8117,
+      "step": 10626
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010124626056787083,
+      "loss": 0.9632,
+      "step": 10627
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010123068310228287,
+      "loss": 0.8721,
+      "step": 10628
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010121510560682687,
+      "loss": 0.8311,
+      "step": 10629
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010119952808188094,
+      "loss": 0.9438,
+      "step": 10630
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010118395052782307,
+      "loss": 0.9478,
+      "step": 10631
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010116837294503135,
+      "loss": 1.4622,
+      "step": 10632
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010115279533388387,
+      "loss": 1.2695,
+      "step": 10633
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010113721769475864,
+      "loss": 0.8211,
+      "step": 10634
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010112164002803373,
+      "loss": 1.0219,
+      "step": 10635
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010110606233408724,
+      "loss": 0.8238,
+      "step": 10636
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001010904846132972,
+      "loss": 0.9447,
+      "step": 10637
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010107490686604168,
+      "loss": 0.815,
+      "step": 10638
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010105932909269878,
+      "loss": 1.0398,
+      "step": 10639
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010104375129364648,
+      "loss": 0.8472,
+      "step": 10640
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010102817346926292,
+      "loss": 0.8759,
+      "step": 10641
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001010125956199261,
+      "loss": 0.901,
+      "step": 10642
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010099701774601415,
+      "loss": 0.7668,
+      "step": 10643
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010098143984790513,
+      "loss": 0.9524,
+      "step": 10644
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010096586192597708,
+      "loss": 1.0116,
+      "step": 10645
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010095028398060806,
+      "loss": 0.9537,
+      "step": 10646
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010093470601217618,
+      "loss": 0.9829,
+      "step": 10647
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010091912802105947,
+      "loss": 0.991,
+      "step": 10648
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010090355000763603,
+      "loss": 0.914,
+      "step": 10649
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010088797197228388,
+      "loss": 0.8471,
+      "step": 10650
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010087239391538113,
+      "loss": 1.4995,
+      "step": 10651
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010085681583730589,
+      "loss": 1.3042,
+      "step": 10652
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010084123773843615,
+      "loss": 1.045,
+      "step": 10653
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010082565961915,
+      "loss": 0.9732,
+      "step": 10654
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010081008147982558,
+      "loss": 0.829,
+      "step": 10655
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010079450332084088,
+      "loss": 0.9999,
+      "step": 10656
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010077892514257403,
+      "loss": 0.8837,
+      "step": 10657
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010076334694540308,
+      "loss": 0.8965,
+      "step": 10658
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010074776872970609,
+      "loss": 1.0568,
+      "step": 10659
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010073219049586118,
+      "loss": 0.9392,
+      "step": 10660
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010071661224424638,
+      "loss": 1.4413,
+      "step": 10661
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010070103397523977,
+      "loss": 1.3528,
+      "step": 10662
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010068545568921947,
+      "loss": 1.0726,
+      "step": 10663
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001006698773865635,
+      "loss": 0.7292,
+      "step": 10664
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010065429906764999,
+      "loss": 0.9804,
+      "step": 10665
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010063872073285696,
+      "loss": 0.9031,
+      "step": 10666
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010062314238256256,
+      "loss": 0.8991,
+      "step": 10667
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010060756401714479,
+      "loss": 0.8388,
+      "step": 10668
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010059198563698178,
+      "loss": 1.0992,
+      "step": 10669
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010057640724245159,
+      "loss": 0.9669,
+      "step": 10670
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010056082883393231,
+      "loss": 0.8611,
+      "step": 10671
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000100545250411802,
+      "loss": 1.0136,
+      "step": 10672
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001005296719764388,
+      "loss": 0.9102,
+      "step": 10673
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001005140935282207,
+      "loss": 0.9144,
+      "step": 10674
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010049851506752583,
+      "loss": 0.9365,
+      "step": 10675
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010048293659473225,
+      "loss": 0.935,
+      "step": 10676
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010046735811021807,
+      "loss": 1.0031,
+      "step": 10677
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010045177961436137,
+      "loss": 0.8815,
+      "step": 10678
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010043620110754022,
+      "loss": 0.8642,
+      "step": 10679
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010042062259013269,
+      "loss": 0.9635,
+      "step": 10680
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010040504406251687,
+      "loss": 1.0521,
+      "step": 10681
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010038946552507089,
+      "loss": 0.845,
+      "step": 10682
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010037388697817276,
+      "loss": 0.9872,
+      "step": 10683
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010035830842220057,
+      "loss": 0.7743,
+      "step": 10684
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010034272985753245,
+      "loss": 0.894,
+      "step": 10685
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010032715128454646,
+      "loss": 0.9153,
+      "step": 10686
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010031157270362069,
+      "loss": 0.9509,
+      "step": 10687
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010029599411513319,
+      "loss": 0.8897,
+      "step": 10688
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001002804155194621,
+      "loss": 0.9795,
+      "step": 10689
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010026483691698548,
+      "loss": 1.0523,
+      "step": 10690
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010024925830808137,
+      "loss": 1.5197,
+      "step": 10691
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010023367969312789,
+      "loss": 1.2777,
+      "step": 10692
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010021810107250317,
+      "loss": 1.0355,
+      "step": 10693
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010020252244658523,
+      "loss": 1.0496,
+      "step": 10694
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001001869438157522,
+      "loss": 0.8878,
+      "step": 10695
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010017136518038215,
+      "loss": 0.9044,
+      "step": 10696
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010015578654085311,
+      "loss": 0.9645,
+      "step": 10697
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010014020789754326,
+      "loss": 0.8695,
+      "step": 10698
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010012462925083062,
+      "loss": 1.0574,
+      "step": 10699
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010010905060109328,
+      "loss": 0.7724,
+      "step": 10700
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010009347194870935,
+      "loss": 0.8348,
+      "step": 10701
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001000778932940569,
+      "loss": 1.5818,
+      "step": 10702
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010006231463751406,
+      "loss": 1.2192,
+      "step": 10703
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010004673597945887,
+      "loss": 0.8657,
+      "step": 10704
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010003115732026938,
+      "loss": 1.4332,
+      "step": 10705
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010001557866032374,
+      "loss": 1.1147,
+      "step": 10706
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8053,
+      "step": 10707
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.998442133967627e-05,
+      "loss": 0.8774,
+      "step": 10708
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.996884267973064e-05,
+      "loss": 1.0008,
+      "step": 10709
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.995326402054117e-05,
+      "loss": 0.9984,
+      "step": 10710
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.993768536248596e-05,
+      "loss": 0.8906,
+      "step": 10711
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.992210670594308e-05,
+      "loss": 0.8245,
+      "step": 10712
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.990652805129064e-05,
+      "loss": 0.9245,
+      "step": 10713
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.989094939890673e-05,
+      "loss": 0.9828,
+      "step": 10714
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.987537074916944e-05,
+      "loss": 0.8291,
+      "step": 10715
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.985979210245678e-05,
+      "loss": 0.9359,
+      "step": 10716
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.984421345914693e-05,
+      "loss": 0.998,
+      "step": 10717
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.98286348196179e-05,
+      "loss": 0.9118,
+      "step": 10718
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.981305618424782e-05,
+      "loss": 1.5291,
+      "step": 10719
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979747755341479e-05,
+      "loss": 1.2095,
+      "step": 10720
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.978189892749684e-05,
+      "loss": 0.8444,
+      "step": 10721
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.976632030687212e-05,
+      "loss": 0.9247,
+      "step": 10722
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.975074169191866e-05,
+      "loss": 0.9704,
+      "step": 10723
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.973516308301456e-05,
+      "loss": 0.854,
+      "step": 10724
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.971958448053793e-05,
+      "loss": 0.9125,
+      "step": 10725
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.970400588486682e-05,
+      "loss": 0.8207,
+      "step": 10726
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.968842729637934e-05,
+      "loss": 0.9473,
+      "step": 10727
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.967284871545355e-05,
+      "loss": 1.0057,
+      "step": 10728
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.965727014246754e-05,
+      "loss": 1.0021,
+      "step": 10729
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.964169157779945e-05,
+      "loss": 0.9529,
+      "step": 10730
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.96261130218273e-05,
+      "loss": 0.9551,
+      "step": 10731
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.961053447492915e-05,
+      "loss": 0.7442,
+      "step": 10732
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.959495593748315e-05,
+      "loss": 0.9117,
+      "step": 10733
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.957937740986733e-05,
+      "loss": 0.8866,
+      "step": 10734
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.956379889245979e-05,
+      "loss": 0.781,
+      "step": 10735
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.954822038563864e-05,
+      "loss": 0.786,
+      "step": 10736
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.953264188978194e-05,
+      "loss": 0.8899,
+      "step": 10737
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.951706340526777e-05,
+      "loss": 1.7062,
+      "step": 10738
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.95014849324742e-05,
+      "loss": 1.3714,
+      "step": 10739
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.948590647177933e-05,
+      "loss": 0.8668,
+      "step": 10740
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.947032802356124e-05,
+      "loss": 0.7965,
+      "step": 10741
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.945474958819799e-05,
+      "loss": 0.9922,
+      "step": 10742
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.943917116606771e-05,
+      "loss": 0.9107,
+      "step": 10743
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.942359275754844e-05,
+      "loss": 1.0271,
+      "step": 10744
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.940801436301827e-05,
+      "loss": 0.8882,
+      "step": 10745
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.939243598285523e-05,
+      "loss": 0.9216,
+      "step": 10746
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.93768576174375e-05,
+      "loss": 0.8827,
+      "step": 10747
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.936127926714306e-05,
+      "loss": 0.8196,
+      "step": 10748
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.934570093235002e-05,
+      "loss": 1.2976,
+      "step": 10749
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.933012261343652e-05,
+      "loss": 1.4158,
+      "step": 10750
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.931454431078055e-05,
+      "loss": 0.819,
+      "step": 10751
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.929896602476024e-05,
+      "loss": 1.3626,
+      "step": 10752
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.928338775575364e-05,
+      "loss": 1.3156,
+      "step": 10753
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.926780950413883e-05,
+      "loss": 0.874,
+      "step": 10754
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.925223127029392e-05,
+      "loss": 0.9009,
+      "step": 10755
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.923665305459693e-05,
+      "loss": 1.1388,
+      "step": 10756
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.9221074857426e-05,
+      "loss": 0.9503,
+      "step": 10757
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.920549667915911e-05,
+      "loss": 1.1242,
+      "step": 10758
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.918991852017446e-05,
+      "loss": 0.9301,
+      "step": 10759
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.917434038085001e-05,
+      "loss": 0.8825,
+      "step": 10760
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91587622615639e-05,
+      "loss": 1.0006,
+      "step": 10761
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.914318416269415e-05,
+      "loss": 0.9495,
+      "step": 10762
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91276060846189e-05,
+      "loss": 0.9224,
+      "step": 10763
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.911202802771614e-05,
+      "loss": 0.8317,
+      "step": 10764
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.909644999236401e-05,
+      "loss": 1.5687,
+      "step": 10765
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.908087197894056e-05,
+      "loss": 1.2214,
+      "step": 10766
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.906529398782383e-05,
+      "loss": 0.9809,
+      "step": 10767
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.904971601939196e-05,
+      "loss": 0.9811,
+      "step": 10768
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.903413807402295e-05,
+      "loss": 1.0435,
+      "step": 10769
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.901856015209488e-05,
+      "loss": 0.8756,
+      "step": 10770
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.900298225398584e-05,
+      "loss": 0.9072,
+      "step": 10771
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.898740438007389e-05,
+      "loss": 0.7986,
+      "step": 10772
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897182653073711e-05,
+      "loss": 1.4161,
+      "step": 10773
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.895624870635357e-05,
+      "loss": 1.1962,
+      "step": 10774
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.894067090730128e-05,
+      "loss": 0.912,
+      "step": 10775
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.892509313395834e-05,
+      "loss": 1.0522,
+      "step": 10776
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.890951538670284e-05,
+      "loss": 0.8611,
+      "step": 10777
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.889393766591279e-05,
+      "loss": 0.9625,
+      "step": 10778
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.88783599719663e-05,
+      "loss": 0.8495,
+      "step": 10779
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.88627823052414e-05,
+      "loss": 0.9253,
+      "step": 10780
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.884720466611615e-05,
+      "loss": 1.0207,
+      "step": 10781
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.883162705496866e-05,
+      "loss": 0.9135,
+      "step": 10782
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.881604947217695e-05,
+      "loss": 0.8385,
+      "step": 10783
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.88004719181191e-05,
+      "loss": 0.8351,
+      "step": 10784
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.878489439317312e-05,
+      "loss": 0.8559,
+      "step": 10785
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.876931689771712e-05,
+      "loss": 0.8079,
+      "step": 10786
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.875373943212915e-05,
+      "loss": 0.9891,
+      "step": 10787
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.87381619967873e-05,
+      "loss": 1.0497,
+      "step": 10788
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.872258459206953e-05,
+      "loss": 0.893,
+      "step": 10789
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.870700721835397e-05,
+      "loss": 1.1027,
+      "step": 10790
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.869142987601863e-05,
+      "loss": 0.9987,
+      "step": 10791
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86758525654416e-05,
+      "loss": 0.9388,
+      "step": 10792
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.866027528700093e-05,
+      "loss": 1.0161,
+      "step": 10793
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.864469804107465e-05,
+      "loss": 0.8209,
+      "step": 10794
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.862912082804083e-05,
+      "loss": 0.8548,
+      "step": 10795
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86135436482775e-05,
+      "loss": 0.9879,
+      "step": 10796
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.859796650216274e-05,
+      "loss": 0.8149,
+      "step": 10797
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.85823893900746e-05,
+      "loss": 0.8936,
+      "step": 10798
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.856681231239109e-05,
+      "loss": 0.8995,
+      "step": 10799
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.855123526949029e-05,
+      "loss": 0.8216,
+      "step": 10800
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.853565826175023e-05,
+      "loss": 0.9438,
+      "step": 10801
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.852008128954894e-05,
+      "loss": 0.9272,
+      "step": 10802
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.850450435326456e-05,
+      "loss": 0.9947,
+      "step": 10803
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.848892745327502e-05,
+      "loss": 0.8958,
+      "step": 10804
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.84733505899584e-05,
+      "loss": 1.3746,
+      "step": 10805
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.845777376369272e-05,
+      "loss": 1.2066,
+      "step": 10806
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.844219697485608e-05,
+      "loss": 1.0665,
+      "step": 10807
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.842662022382646e-05,
+      "loss": 0.8741,
+      "step": 10808
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.841104351098196e-05,
+      "loss": 0.8056,
+      "step": 10809
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.839546683670056e-05,
+      "loss": 0.7715,
+      "step": 10810
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.837989020136031e-05,
+      "loss": 0.9266,
+      "step": 10811
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.83643136053393e-05,
+      "loss": 0.8982,
+      "step": 10812
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.834873704901549e-05,
+      "loss": 0.8639,
+      "step": 10813
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.833316053276699e-05,
+      "loss": 0.9025,
+      "step": 10814
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.831758405697176e-05,
+      "loss": 1.0077,
+      "step": 10815
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.830200762200786e-05,
+      "loss": 0.9489,
+      "step": 10816
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.828643122825336e-05,
+      "loss": 0.945,
+      "step": 10817
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.827085487608627e-05,
+      "loss": 0.8508,
+      "step": 10818
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.825527856588457e-05,
+      "loss": 1.5222,
+      "step": 10819
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.823970229802635e-05,
+      "loss": 1.3248,
+      "step": 10820
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.82241260728896e-05,
+      "loss": 0.9557,
+      "step": 10821
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.820854989085235e-05,
+      "loss": 0.8287,
+      "step": 10822
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.819297375229268e-05,
+      "loss": 0.8567,
+      "step": 10823
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.817739765758852e-05,
+      "loss": 0.8327,
+      "step": 10824
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.816182160711799e-05,
+      "loss": 0.8188,
+      "step": 10825
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.814624560125903e-05,
+      "loss": 0.7612,
+      "step": 10826
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.81306696403897e-05,
+      "loss": 0.9671,
+      "step": 10827
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.811509372488806e-05,
+      "loss": 0.959,
+      "step": 10828
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.809951785513205e-05,
+      "loss": 0.9541,
+      "step": 10829
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.808394203149975e-05,
+      "loss": 0.9174,
+      "step": 10830
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.806836625436914e-05,
+      "loss": 0.9904,
+      "step": 10831
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.805279052411828e-05,
+      "loss": 1.0637,
+      "step": 10832
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.803721484112512e-05,
+      "loss": 0.8318,
+      "step": 10833
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.802163920576773e-05,
+      "loss": 0.9654,
+      "step": 10834
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.800606361842408e-05,
+      "loss": 0.8933,
+      "step": 10835
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.799048807947223e-05,
+      "loss": 0.9501,
+      "step": 10836
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.797491258929011e-05,
+      "loss": 0.91,
+      "step": 10837
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.795933714825581e-05,
+      "loss": 1.0036,
+      "step": 10838
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.794376175674731e-05,
+      "loss": 0.9547,
+      "step": 10839
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.79281864151426e-05,
+      "loss": 1.0021,
+      "step": 10840
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.791261112381972e-05,
+      "loss": 0.9214,
+      "step": 10841
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.789703588315662e-05,
+      "loss": 1.0133,
+      "step": 10842
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.788146069353136e-05,
+      "loss": 0.8306,
+      "step": 10843
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.786588555532193e-05,
+      "loss": 0.718,
+      "step": 10844
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.785031046890629e-05,
+      "loss": 1.4562,
+      "step": 10845
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.783473543466248e-05,
+      "loss": 1.2327,
+      "step": 10846
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.781916045296852e-05,
+      "loss": 0.9697,
+      "step": 10847
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.780358552420232e-05,
+      "loss": 1.0282,
+      "step": 10848
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.778801064874193e-05,
+      "loss": 0.7252,
+      "step": 10849
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.777243582696536e-05,
+      "loss": 0.8113,
+      "step": 10850
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.775686105925057e-05,
+      "loss": 0.9258,
+      "step": 10851
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.774128634597557e-05,
+      "loss": 0.8515,
+      "step": 10852
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.772571168751834e-05,
+      "loss": 0.9297,
+      "step": 10853
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.771013708425687e-05,
+      "loss": 0.9861,
+      "step": 10854
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.769456253656917e-05,
+      "loss": 1.0644,
+      "step": 10855
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.767898804483317e-05,
+      "loss": 0.8812,
+      "step": 10856
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.766341360942692e-05,
+      "loss": 1.078,
+      "step": 10857
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.764783923072835e-05,
+      "loss": 0.8195,
+      "step": 10858
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.763226490911548e-05,
+      "loss": 0.8616,
+      "step": 10859
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.761669064496629e-05,
+      "loss": 0.9728,
+      "step": 10860
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.760111643865871e-05,
+      "loss": 0.9604,
+      "step": 10861
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.758554229057081e-05,
+      "loss": 0.8475,
+      "step": 10862
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.756996820108048e-05,
+      "loss": 1.5508,
+      "step": 10863
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.755439417056572e-05,
+      "loss": 1.1897,
+      "step": 10864
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753882019940449e-05,
+      "loss": 0.9732,
+      "step": 10865
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.75232462879748e-05,
+      "loss": 0.8769,
+      "step": 10866
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.750767243665459e-05,
+      "loss": 0.7425,
+      "step": 10867
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.749209864582185e-05,
+      "loss": 1.0722,
+      "step": 10868
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.747652491585452e-05,
+      "loss": 1.6047,
+      "step": 10869
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.746095124713058e-05,
+      "loss": 1.184,
+      "step": 10870
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.744537764002803e-05,
+      "loss": 1.3265,
+      "step": 10871
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.742980409492476e-05,
+      "loss": 1.2657,
+      "step": 10872
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.74142306121988e-05,
+      "loss": 0.7397,
+      "step": 10873
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.739865719222806e-05,
+      "loss": 0.9991,
+      "step": 10874
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.738308383539052e-05,
+      "loss": 1.4966,
+      "step": 10875
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.736751054206417e-05,
+      "loss": 1.2719,
+      "step": 10876
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.735193731262694e-05,
+      "loss": 0.8301,
+      "step": 10877
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.733636414745674e-05,
+      "loss": 0.8691,
+      "step": 10878
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.732079104693155e-05,
+      "loss": 0.7975,
+      "step": 10879
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.730521801142938e-05,
+      "loss": 0.8332,
+      "step": 10880
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.728964504132809e-05,
+      "loss": 0.9155,
+      "step": 10881
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.727407213700568e-05,
+      "loss": 0.8566,
+      "step": 10882
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.725849929884008e-05,
+      "loss": 0.9215,
+      "step": 10883
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.724292652720922e-05,
+      "loss": 0.8332,
+      "step": 10884
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.722735382249109e-05,
+      "loss": 0.8877,
+      "step": 10885
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.721178118506358e-05,
+      "loss": 0.9337,
+      "step": 10886
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.719620861530467e-05,
+      "loss": 0.8871,
+      "step": 10887
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.718063611359226e-05,
+      "loss": 0.8752,
+      "step": 10888
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.716506368030431e-05,
+      "loss": 0.8716,
+      "step": 10889
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.714949131581877e-05,
+      "loss": 0.8976,
+      "step": 10890
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.713391902051357e-05,
+      "loss": 0.9109,
+      "step": 10891
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.71183467947666e-05,
+      "loss": 0.821,
+      "step": 10892
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.710277463895581e-05,
+      "loss": 0.877,
+      "step": 10893
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.708720255345913e-05,
+      "loss": 0.8936,
+      "step": 10894
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.707163053865447e-05,
+      "loss": 0.9797,
+      "step": 10895
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.705605859491981e-05,
+      "loss": 0.8845,
+      "step": 10896
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.7040486722633e-05,
+      "loss": 1.0251,
+      "step": 10897
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.702491492217203e-05,
+      "loss": 0.9143,
+      "step": 10898
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.700934319391475e-05,
+      "loss": 0.9613,
+      "step": 10899
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.699377153823912e-05,
+      "loss": 1.0003,
+      "step": 10900
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.697819995552309e-05,
+      "loss": 0.8247,
+      "step": 10901
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.69626284461445e-05,
+      "loss": 0.8857,
+      "step": 10902
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.69470570104813e-05,
+      "loss": 0.8878,
+      "step": 10903
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.69314856489114e-05,
+      "loss": 1.0267,
+      "step": 10904
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.691591436181268e-05,
+      "loss": 0.8949,
+      "step": 10905
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.690034314956314e-05,
+      "loss": 1.0513,
+      "step": 10906
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.688477201254057e-05,
+      "loss": 0.9297,
+      "step": 10907
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.686920095112291e-05,
+      "loss": 0.9306,
+      "step": 10908
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.685362996568808e-05,
+      "loss": 1.0184,
+      "step": 10909
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.683805905661397e-05,
+      "loss": 0.9383,
+      "step": 10910
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.682248822427846e-05,
+      "loss": 0.8648,
+      "step": 10911
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.68069174690595e-05,
+      "loss": 0.9579,
+      "step": 10912
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.679134679133491e-05,
+      "loss": 1.0354,
+      "step": 10913
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.677577619148266e-05,
+      "loss": 0.8867,
+      "step": 10914
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.676020566988057e-05,
+      "loss": 1.0006,
+      "step": 10915
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.674463522690655e-05,
+      "loss": 0.9748,
+      "step": 10916
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.672906486293852e-05,
+      "loss": 0.8884,
+      "step": 10917
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.671349457835433e-05,
+      "loss": 0.8265,
+      "step": 10918
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.669792437353187e-05,
+      "loss": 0.8641,
+      "step": 10919
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.668235424884901e-05,
+      "loss": 0.8636,
+      "step": 10920
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.666678420468369e-05,
+      "loss": 0.9465,
+      "step": 10921
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.665121424141369e-05,
+      "loss": 0.9457,
+      "step": 10922
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.663564435941695e-05,
+      "loss": 1.0141,
+      "step": 10923
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.66200745590713e-05,
+      "loss": 0.9025,
+      "step": 10924
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.660450484075468e-05,
+      "loss": 1.033,
+      "step": 10925
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.658893520484487e-05,
+      "loss": 0.9127,
+      "step": 10926
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.657336565171978e-05,
+      "loss": 0.9369,
+      "step": 10927
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.655779618175731e-05,
+      "loss": 0.8635,
+      "step": 10928
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.654222679533525e-05,
+      "loss": 0.9767,
+      "step": 10929
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.652665749283153e-05,
+      "loss": 0.903,
+      "step": 10930
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.651108827462396e-05,
+      "loss": 0.9135,
+      "step": 10931
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.64955191410904e-05,
+      "loss": 0.8327,
+      "step": 10932
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.647995009260876e-05,
+      "loss": 0.9102,
+      "step": 10933
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.646438112955682e-05,
+      "loss": 0.8571,
+      "step": 10934
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.64488122523125e-05,
+      "loss": 1.0024,
+      "step": 10935
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.64332434612536e-05,
+      "loss": 1.1543,
+      "step": 10936
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.641767475675795e-05,
+      "loss": 0.9854,
+      "step": 10937
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.640210613920341e-05,
+      "loss": 0.8247,
+      "step": 10938
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.638653760896786e-05,
+      "loss": 0.8126,
+      "step": 10939
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.637096916642909e-05,
+      "loss": 0.9425,
+      "step": 10940
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.635540081196499e-05,
+      "loss": 0.9659,
+      "step": 10941
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.633983254595333e-05,
+      "loss": 0.9399,
+      "step": 10942
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.632426436877199e-05,
+      "loss": 1.0435,
+      "step": 10943
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.630869628079882e-05,
+      "loss": 0.9763,
+      "step": 10944
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.629312828241158e-05,
+      "loss": 0.8476,
+      "step": 10945
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.627756037398817e-05,
+      "loss": 1.4943,
+      "step": 10946
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.626199255590635e-05,
+      "loss": 1.2905,
+      "step": 10947
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.624642482854398e-05,
+      "loss": 1.4294,
+      "step": 10948
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.62308571922789e-05,
+      "loss": 1.2073,
+      "step": 10949
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.621528964748891e-05,
+      "loss": 0.7669,
+      "step": 10950
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.619972219455179e-05,
+      "loss": 0.9604,
+      "step": 10951
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.618415483384538e-05,
+      "loss": 0.8028,
+      "step": 10952
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.616858756574751e-05,
+      "loss": 0.8466,
+      "step": 10953
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.615302039063596e-05,
+      "loss": 0.8487,
+      "step": 10954
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.613745330888857e-05,
+      "loss": 0.972,
+      "step": 10955
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.61218863208831e-05,
+      "loss": 0.8202,
+      "step": 10956
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.610631942699738e-05,
+      "loss": 0.9091,
+      "step": 10957
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.609075262760923e-05,
+      "loss": 0.9399,
+      "step": 10958
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.60751859230964e-05,
+      "loss": 0.8669,
+      "step": 10959
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.605961931383674e-05,
+      "loss": 1.4446,
+      "step": 10960
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.604405280020799e-05,
+      "loss": 1.194,
+      "step": 10961
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.602848638258797e-05,
+      "loss": 0.8405,
+      "step": 10962
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.601292006135447e-05,
+      "loss": 1.007,
+      "step": 10963
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.599735383688526e-05,
+      "loss": 0.9212,
+      "step": 10964
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.598178770955819e-05,
+      "loss": 0.9627,
+      "step": 10965
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.596622167975094e-05,
+      "loss": 0.9525,
+      "step": 10966
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.595065574784134e-05,
+      "loss": 1.0104,
+      "step": 10967
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.593508991420714e-05,
+      "loss": 0.9726,
+      "step": 10968
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.591952417922616e-05,
+      "loss": 1.0299,
+      "step": 10969
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.590395854327614e-05,
+      "loss": 0.8994,
+      "step": 10970
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.588839300673486e-05,
+      "loss": 0.888,
+      "step": 10971
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.58728275699801e-05,
+      "loss": 0.946,
+      "step": 10972
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.585726223338957e-05,
+      "loss": 0.9767,
+      "step": 10973
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.584169699734113e-05,
+      "loss": 1.0418,
+      "step": 10974
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.582613186221243e-05,
+      "loss": 0.9898,
+      "step": 10975
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.581056682838132e-05,
+      "loss": 1.0038,
+      "step": 10976
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.579500189622549e-05,
+      "loss": 0.9654,
+      "step": 10977
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.57794370661227e-05,
+      "loss": 1.0226,
+      "step": 10978
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.57638723384508e-05,
+      "loss": 0.9892,
+      "step": 10979
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.574830771358742e-05,
+      "loss": 0.9164,
+      "step": 10980
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.57327431919103e-05,
+      "loss": 0.8992,
+      "step": 10981
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.571717877379727e-05,
+      "loss": 0.9613,
+      "step": 10982
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.570161445962599e-05,
+      "loss": 1.4103,
+      "step": 10983
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.568605024977422e-05,
+      "loss": 1.1311,
+      "step": 10984
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.567048614461974e-05,
+      "loss": 0.9676,
+      "step": 10985
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.565492214454023e-05,
+      "loss": 0.9202,
+      "step": 10986
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.563935824991346e-05,
+      "loss": 0.8167,
+      "step": 10987
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.562379446111711e-05,
+      "loss": 0.9725,
+      "step": 10988
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.560823077852893e-05,
+      "loss": 0.8467,
+      "step": 10989
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.559266720252667e-05,
+      "loss": 0.9316,
+      "step": 10990
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.557710373348801e-05,
+      "loss": 0.9066,
+      "step": 10991
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.556154037179068e-05,
+      "loss": 0.9744,
+      "step": 10992
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.554597711781238e-05,
+      "loss": 0.9424,
+      "step": 10993
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.553041397193089e-05,
+      "loss": 0.7096,
+      "step": 10994
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.551485093452382e-05,
+      "loss": 0.9079,
+      "step": 10995
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.549928800596893e-05,
+      "loss": 0.9295,
+      "step": 10996
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.548372518664391e-05,
+      "loss": 0.8726,
+      "step": 10997
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.546816247692649e-05,
+      "loss": 0.9764,
+      "step": 10998
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.545259987719431e-05,
+      "loss": 0.9435,
+      "step": 10999
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.543703738782511e-05,
+      "loss": 0.8734,
+      "step": 11000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.54214750091966e-05,
+      "loss": 0.874,
+      "step": 11001
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.540591274168642e-05,
+      "loss": 1.0307,
+      "step": 11002
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.53903505856723e-05,
+      "loss": 1.0533,
+      "step": 11003
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.53747885415319e-05,
+      "loss": 0.921,
+      "step": 11004
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.535922660964292e-05,
+      "loss": 0.8332,
+      "step": 11005
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.534366479038303e-05,
+      "loss": 1.0887,
+      "step": 11006
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.532810308412989e-05,
+      "loss": 0.9422,
+      "step": 11007
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.531254149126122e-05,
+      "loss": 0.9162,
+      "step": 11008
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.529698001215468e-05,
+      "loss": 0.7744,
+      "step": 11009
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.52814186471879e-05,
+      "loss": 0.9021,
+      "step": 11010
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.526585739673857e-05,
+      "loss": 0.8494,
+      "step": 11011
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.525029626118435e-05,
+      "loss": 1.4115,
+      "step": 11012
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.52347352409029e-05,
+      "loss": 1.2302,
+      "step": 11013
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.521917433627187e-05,
+      "loss": 1.0055,
+      "step": 11014
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.520361354766895e-05,
+      "loss": 0.9487,
+      "step": 11015
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.518805287547174e-05,
+      "loss": 0.902,
+      "step": 11016
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.517249232005795e-05,
+      "loss": 0.9756,
+      "step": 11017
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.515693188180517e-05,
+      "loss": 1.075,
+      "step": 11018
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.514137156109105e-05,
+      "loss": 0.8665,
+      "step": 11019
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.512581135829328e-05,
+      "loss": 0.8911,
+      "step": 11020
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.511025127378946e-05,
+      "loss": 0.9591,
+      "step": 11021
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.509469130795723e-05,
+      "loss": 0.9623,
+      "step": 11022
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.50791314611742e-05,
+      "loss": 0.9802,
+      "step": 11023
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.506357173381807e-05,
+      "loss": 0.8981,
+      "step": 11024
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.504801212626637e-05,
+      "loss": 1.0326,
+      "step": 11025
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.50324526388968e-05,
+      "loss": 0.7726,
+      "step": 11026
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.501689327208693e-05,
+      "loss": 1.0417,
+      "step": 11027
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.500133402621441e-05,
+      "loss": 1.0602,
+      "step": 11028
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.498577490165684e-05,
+      "loss": 0.7943,
+      "step": 11029
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.497021589879181e-05,
+      "loss": 0.9572,
+      "step": 11030
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.4954657017997e-05,
+      "loss": 1.0279,
+      "step": 11031
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.493909825964994e-05,
+      "loss": 0.9522,
+      "step": 11032
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.492353962412827e-05,
+      "loss": 0.8802,
+      "step": 11033
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.490798111180957e-05,
+      "loss": 0.8828,
+      "step": 11034
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.489242272307146e-05,
+      "loss": 0.9624,
+      "step": 11035
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.487686445829152e-05,
+      "loss": 1.3368,
+      "step": 11036
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.486130631784735e-05,
+      "loss": 1.6262,
+      "step": 11037
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.484574830211655e-05,
+      "loss": 1.1952,
+      "step": 11038
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.483019041147668e-05,
+      "loss": 0.804,
+      "step": 11039
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.48146326463053e-05,
+      "loss": 0.9347,
+      "step": 11040
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.479907500698002e-05,
+      "loss": 0.9088,
+      "step": 11041
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.478351749387843e-05,
+      "loss": 0.9905,
+      "step": 11042
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.476796010737806e-05,
+      "loss": 0.9417,
+      "step": 11043
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.475240284785653e-05,
+      "loss": 1.3317,
+      "step": 11044
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.473684571569135e-05,
+      "loss": 1.1767,
+      "step": 11045
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.472128871126013e-05,
+      "loss": 0.8659,
+      "step": 11046
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.470573183494041e-05,
+      "loss": 0.9378,
+      "step": 11047
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.469017508710976e-05,
+      "loss": 0.8998,
+      "step": 11048
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.467461846814571e-05,
+      "loss": 0.9551,
+      "step": 11049
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.465906197842584e-05,
+      "loss": 0.9359,
+      "step": 11050
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.464350561832766e-05,
+      "loss": 0.9374,
+      "step": 11051
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.462794938822876e-05,
+      "loss": 0.8103,
+      "step": 11052
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.46123932885067e-05,
+      "loss": 0.8782,
+      "step": 11053
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.459683731953891e-05,
+      "loss": 1.0026,
+      "step": 11054
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.458128148170303e-05,
+      "loss": 0.898,
+      "step": 11055
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.456572577537652e-05,
+      "loss": 0.8844,
+      "step": 11056
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.455017020093696e-05,
+      "loss": 1.4087,
+      "step": 11057
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.453461475876188e-05,
+      "loss": 1.3034,
+      "step": 11058
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.451905944922874e-05,
+      "loss": 0.8392,
+      "step": 11059
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.450350427271512e-05,
+      "loss": 0.9438,
+      "step": 11060
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.448794922959852e-05,
+      "loss": 0.8399,
+      "step": 11061
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.447239432025643e-05,
+      "loss": 1.0543,
+      "step": 11062
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.44568395450664e-05,
+      "loss": 0.9425,
+      "step": 11063
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.444128490440591e-05,
+      "loss": 0.8425,
+      "step": 11064
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.442573039865245e-05,
+      "loss": 0.8783,
+      "step": 11065
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.441017602818355e-05,
+      "loss": 0.9815,
+      "step": 11066
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.439462179337668e-05,
+      "loss": 0.8629,
+      "step": 11067
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.437906769460938e-05,
+      "loss": 0.9196,
+      "step": 11068
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.43635137322591e-05,
+      "loss": 0.925,
+      "step": 11069
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.43479599067033e-05,
+      "loss": 1.0941,
+      "step": 11070
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.433240621831951e-05,
+      "loss": 0.8756,
+      "step": 11071
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.431685266748518e-05,
+      "loss": 0.8461,
+      "step": 11072
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.43012992545778e-05,
+      "loss": 0.947,
+      "step": 11073
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.428574597997488e-05,
+      "loss": 1.032,
+      "step": 11074
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.427019284405381e-05,
+      "loss": 0.9254,
+      "step": 11075
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.425463984719211e-05,
+      "loss": 0.9491,
+      "step": 11076
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.423908698976723e-05,
+      "loss": 0.9069,
+      "step": 11077
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.422353427215664e-05,
+      "loss": 0.8065,
+      "step": 11078
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.420798169473779e-05,
+      "loss": 0.8181,
+      "step": 11079
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.419242925788813e-05,
+      "loss": 0.9924,
+      "step": 11080
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.417687696198511e-05,
+      "loss": 1.0994,
+      "step": 11081
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.416132480740619e-05,
+      "loss": 0.8258,
+      "step": 11082
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.41457727945288e-05,
+      "loss": 0.9736,
+      "step": 11083
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.413022092373034e-05,
+      "loss": 1.5088,
+      "step": 11084
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.411466919538831e-05,
+      "loss": 1.2174,
+      "step": 11085
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.409911760988011e-05,
+      "loss": 0.9678,
+      "step": 11086
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.408356616758315e-05,
+      "loss": 0.8436,
+      "step": 11087
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.406801486887492e-05,
+      "loss": 0.9847,
+      "step": 11088
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.405246371413278e-05,
+      "loss": 0.8208,
+      "step": 11089
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.403691270373418e-05,
+      "loss": 0.9821,
+      "step": 11090
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.402136183805649e-05,
+      "loss": 0.8886,
+      "step": 11091
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.400581111747719e-05,
+      "loss": 0.713,
+      "step": 11092
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.399026054237365e-05,
+      "loss": 0.8365,
+      "step": 11093
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.397471011312326e-05,
+      "loss": 1.0593,
+      "step": 11094
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.395915983010344e-05,
+      "loss": 0.952,
+      "step": 11095
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.394360969369159e-05,
+      "loss": 0.9124,
+      "step": 11096
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.392805970426513e-05,
+      "loss": 0.8246,
+      "step": 11097
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.391250986220139e-05,
+      "loss": 0.9372,
+      "step": 11098
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.389696016787779e-05,
+      "loss": 0.84,
+      "step": 11099
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.388141062167168e-05,
+      "loss": 1.0351,
+      "step": 11100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.386586122396051e-05,
+      "loss": 1.5293,
+      "step": 11101
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.385031197512158e-05,
+      "loss": 1.2919,
+      "step": 11102
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.383476287553227e-05,
+      "loss": 0.9492,
+      "step": 11103
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.381921392557003e-05,
+      "loss": 0.7894,
+      "step": 11104
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.380366512561212e-05,
+      "loss": 1.0461,
+      "step": 11105
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.378811647603599e-05,
+      "loss": 0.9527,
+      "step": 11106
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.377256797721892e-05,
+      "loss": 0.9682,
+      "step": 11107
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.37570196295383e-05,
+      "loss": 0.9223,
+      "step": 11108
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.37414714333715e-05,
+      "loss": 1.045,
+      "step": 11109
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.372592338909583e-05,
+      "loss": 0.9881,
+      "step": 11110
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.371037549708866e-05,
+      "loss": 0.8771,
+      "step": 11111
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.369482775772734e-05,
+      "loss": 0.9314,
+      "step": 11112
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.367928017138915e-05,
+      "loss": 0.9326,
+      "step": 11113
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.366373273845144e-05,
+      "loss": 0.9668,
+      "step": 11114
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.36481854592916e-05,
+      "loss": 0.8319,
+      "step": 11115
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.363263833428686e-05,
+      "loss": 0.9475,
+      "step": 11116
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.361709136381462e-05,
+      "loss": 0.9457,
+      "step": 11117
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.360154454825214e-05,
+      "loss": 1.1474,
+      "step": 11118
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.358599788797675e-05,
+      "loss": 0.9669,
+      "step": 11119
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.35704513833658e-05,
+      "loss": 1.0345,
+      "step": 11120
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.355490503479653e-05,
+      "loss": 1.0419,
+      "step": 11121
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.353935884264628e-05,
+      "loss": 1.002,
+      "step": 11122
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.352381280729234e-05,
+      "loss": 1.0229,
+      "step": 11123
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.350826692911199e-05,
+      "loss": 0.912,
+      "step": 11124
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.349272120848258e-05,
+      "loss": 0.7778,
+      "step": 11125
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.347717564578135e-05,
+      "loss": 0.9598,
+      "step": 11126
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.346163024138555e-05,
+      "loss": 0.8768,
+      "step": 11127
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.344608499567251e-05,
+      "loss": 0.8127,
+      "step": 11128
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.343053990901946e-05,
+      "loss": 0.8778,
+      "step": 11129
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.34149949818037e-05,
+      "loss": 1.0518,
+      "step": 11130
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.339945021440252e-05,
+      "loss": 0.8669,
+      "step": 11131
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.338390560719313e-05,
+      "loss": 0.9818,
+      "step": 11132
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.336836116055286e-05,
+      "loss": 0.9285,
+      "step": 11133
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.335281687485888e-05,
+      "loss": 0.7254,
+      "step": 11134
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.333727275048849e-05,
+      "loss": 1.0017,
+      "step": 11135
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.332172878781895e-05,
+      "loss": 0.953,
+      "step": 11136
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.330618498722746e-05,
+      "loss": 0.917,
+      "step": 11137
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.329064134909132e-05,
+      "loss": 0.8348,
+      "step": 11138
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.327509787378769e-05,
+      "loss": 0.796,
+      "step": 11139
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.325955456169386e-05,
+      "loss": 1.015,
+      "step": 11140
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.324401141318708e-05,
+      "loss": 1.0069,
+      "step": 11141
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.322846842864452e-05,
+      "loss": 0.763,
+      "step": 11142
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.321292560844338e-05,
+      "loss": 0.9433,
+      "step": 11143
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.319738295296094e-05,
+      "loss": 0.9753,
+      "step": 11144
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.318184046257436e-05,
+      "loss": 0.9836,
+      "step": 11145
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.316629813766086e-05,
+      "loss": 0.9626,
+      "step": 11146
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.315075597859768e-05,
+      "loss": 1.0014,
+      "step": 11147
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.313521398576198e-05,
+      "loss": 0.9379,
+      "step": 11148
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.311967215953097e-05,
+      "loss": 0.809,
+      "step": 11149
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.310413050028187e-05,
+      "loss": 0.8763,
+      "step": 11150
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.30885890083918e-05,
+      "loss": 0.8486,
+      "step": 11151
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.307304768423801e-05,
+      "loss": 0.8805,
+      "step": 11152
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.305750652819764e-05,
+      "loss": 0.7944,
+      "step": 11153
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.304196554064787e-05,
+      "loss": 0.7764,
+      "step": 11154
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.30264247219659e-05,
+      "loss": 1.5125,
+      "step": 11155
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.301088407252891e-05,
+      "loss": 1.1019,
+      "step": 11156
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.299534359271397e-05,
+      "loss": 0.8973,
+      "step": 11157
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.297980328289833e-05,
+      "loss": 1.4268,
+      "step": 11158
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.29642631434591e-05,
+      "loss": 1.1202,
+      "step": 11159
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.294872317477345e-05,
+      "loss": 0.8152,
+      "step": 11160
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.293318337721852e-05,
+      "loss": 0.834,
+      "step": 11161
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.291764375117146e-05,
+      "loss": 0.9533,
+      "step": 11162
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.290210429700941e-05,
+      "loss": 0.9568,
+      "step": 11163
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.288656501510948e-05,
+      "loss": 0.9664,
+      "step": 11164
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.287102590584882e-05,
+      "loss": 0.966,
+      "step": 11165
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.285548696960458e-05,
+      "loss": 0.9704,
+      "step": 11166
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.283994820675381e-05,
+      "loss": 0.8396,
+      "step": 11167
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.282440961767373e-05,
+      "loss": 0.8204,
+      "step": 11168
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.280887120274135e-05,
+      "loss": 0.9586,
+      "step": 11169
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.279333296233387e-05,
+      "loss": 0.9365,
+      "step": 11170
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.27777948968283e-05,
+      "loss": 1.0251,
+      "step": 11171
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.276225700660184e-05,
+      "loss": 0.936,
+      "step": 11172
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.274671929203149e-05,
+      "loss": 0.8147,
+      "step": 11173
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.273118175349442e-05,
+      "loss": 0.826,
+      "step": 11174
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.271564439136767e-05,
+      "loss": 1.461,
+      "step": 11175
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.270010720602832e-05,
+      "loss": 1.2346,
+      "step": 11176
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.26845701978535e-05,
+      "loss": 1.0307,
+      "step": 11177
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.266903336722025e-05,
+      "loss": 0.9272,
+      "step": 11178
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.265349671450565e-05,
+      "loss": 0.8714,
+      "step": 11179
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.263796024008675e-05,
+      "loss": 1.031,
+      "step": 11180
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.262242394434059e-05,
+      "loss": 0.7866,
+      "step": 11181
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.260688782764432e-05,
+      "loss": 0.9304,
+      "step": 11182
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.25913518903749e-05,
+      "loss": 1.0113,
+      "step": 11183
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.257581613290944e-05,
+      "loss": 0.9007,
+      "step": 11184
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.256028055562496e-05,
+      "loss": 1.0571,
+      "step": 11185
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.254474515889847e-05,
+      "loss": 0.8514,
+      "step": 11186
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.252920994310703e-05,
+      "loss": 0.8183,
+      "step": 11187
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.25136749086277e-05,
+      "loss": 0.8688,
+      "step": 11188
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.249814005583745e-05,
+      "loss": 0.8461,
+      "step": 11189
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.248260538511336e-05,
+      "loss": 1.0942,
+      "step": 11190
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.24670708968324e-05,
+      "loss": 0.984,
+      "step": 11191
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.245153659137161e-05,
+      "loss": 0.8201,
+      "step": 11192
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.243600246910801e-05,
+      "loss": 0.8935,
+      "step": 11193
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.242046853041858e-05,
+      "loss": 0.9858,
+      "step": 11194
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.240493477568034e-05,
+      "loss": 0.7949,
+      "step": 11195
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.238940120527026e-05,
+      "loss": 0.766,
+      "step": 11196
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.237386781956535e-05,
+      "loss": 0.9277,
+      "step": 11197
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.23583346189426e-05,
+      "loss": 0.8632,
+      "step": 11198
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.234280160377899e-05,
+      "loss": 0.9568,
+      "step": 11199
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.232726877445152e-05,
+      "loss": 0.8765,
+      "step": 11200
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.231173613133714e-05,
+      "loss": 1.3937,
+      "step": 11201
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.229620367481277e-05,
+      "loss": 1.0159,
+      "step": 11202
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.228067140525545e-05,
+      "loss": 1.3876,
+      "step": 11203
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.22651393230421e-05,
+      "loss": 1.1414,
+      "step": 11204
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.224960742854968e-05,
+      "loss": 0.8675,
+      "step": 11205
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.223407572215519e-05,
+      "loss": 0.7573,
+      "step": 11206
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.221854420423548e-05,
+      "loss": 0.8712,
+      "step": 11207
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.220301287516758e-05,
+      "loss": 1.0798,
+      "step": 11208
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.218748173532839e-05,
+      "loss": 0.8492,
+      "step": 11209
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.217195078509483e-05,
+      "loss": 1.4284,
+      "step": 11210
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.215642002484387e-05,
+      "loss": 1.3404,
+      "step": 11211
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.214088945495238e-05,
+      "loss": 1.3585,
+      "step": 11212
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.21253590757973e-05,
+      "loss": 1.0322,
+      "step": 11213
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.210982888775557e-05,
+      "loss": 0.8954,
+      "step": 11214
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.209429889120412e-05,
+      "loss": 0.9561,
+      "step": 11215
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.207876908651975e-05,
+      "loss": 0.9087,
+      "step": 11216
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.206323947407944e-05,
+      "loss": 0.8994,
+      "step": 11217
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.204771005426008e-05,
+      "loss": 0.9939,
+      "step": 11218
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.203218082743853e-05,
+      "loss": 0.8567,
+      "step": 11219
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.201665179399171e-05,
+      "loss": 0.9095,
+      "step": 11220
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.200112295429649e-05,
+      "loss": 0.858,
+      "step": 11221
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.198559430872972e-05,
+      "loss": 0.8439,
+      "step": 11222
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.197006585766832e-05,
+      "loss": 1.0933,
+      "step": 11223
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.19545376014891e-05,
+      "loss": 0.8696,
+      "step": 11224
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.193900954056899e-05,
+      "loss": 0.9998,
+      "step": 11225
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.19234816752848e-05,
+      "loss": 1.0158,
+      "step": 11226
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.19079540060134e-05,
+      "loss": 0.9941,
+      "step": 11227
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.189242653313164e-05,
+      "loss": 1.0015,
+      "step": 11228
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.187689925701638e-05,
+      "loss": 1.015,
+      "step": 11229
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.18613721780444e-05,
+      "loss": 0.9414,
+      "step": 11230
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.18458452965926e-05,
+      "loss": 0.8636,
+      "step": 11231
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.183031861303773e-05,
+      "loss": 0.9441,
+      "step": 11232
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.181479212775668e-05,
+      "loss": 0.8562,
+      "step": 11233
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.179926584112629e-05,
+      "loss": 1.1052,
+      "step": 11234
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.17837397535233e-05,
+      "loss": 0.9554,
+      "step": 11235
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.176821386532459e-05,
+      "loss": 0.9375,
+      "step": 11236
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.17526881769069e-05,
+      "loss": 0.967,
+      "step": 11237
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.173716268864707e-05,
+      "loss": 0.7969,
+      "step": 11238
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.17216374009219e-05,
+      "loss": 0.8273,
+      "step": 11239
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.170611231410815e-05,
+      "loss": 1.035,
+      "step": 11240
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.169058742858265e-05,
+      "loss": 1.0132,
+      "step": 11241
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.167506274472213e-05,
+      "loss": 1.0447,
+      "step": 11242
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.165953826290339e-05,
+      "loss": 0.7348,
+      "step": 11243
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.164401398350326e-05,
+      "loss": 0.8509,
+      "step": 11244
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.162848990689841e-05,
+      "loss": 0.8211,
+      "step": 11245
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.161296603346561e-05,
+      "loss": 0.8873,
+      "step": 11246
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.159744236358169e-05,
+      "loss": 0.9568,
+      "step": 11247
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.158191889762332e-05,
+      "loss": 0.8398,
+      "step": 11248
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.156639563596729e-05,
+      "loss": 0.7067,
+      "step": 11249
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.155087257899035e-05,
+      "loss": 0.9677,
+      "step": 11250
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.153534972706919e-05,
+      "loss": 0.9777,
+      "step": 11251
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.15198270805806e-05,
+      "loss": 0.9566,
+      "step": 11252
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.150430463990125e-05,
+      "loss": 0.9926,
+      "step": 11253
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.14887824054079e-05,
+      "loss": 0.9691,
+      "step": 11254
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.147326037747728e-05,
+      "loss": 0.9557,
+      "step": 11255
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.145773855648604e-05,
+      "loss": 1.4272,
+      "step": 11256
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.144221694281095e-05,
+      "loss": 1.213,
+      "step": 11257
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.142669553682867e-05,
+      "loss": 0.8965,
+      "step": 11258
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.141117433891594e-05,
+      "loss": 0.7572,
+      "step": 11259
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.139565334944939e-05,
+      "loss": 0.8141,
+      "step": 11260
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.138013256880576e-05,
+      "loss": 0.9834,
+      "step": 11261
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.136461199736167e-05,
+      "loss": 0.877,
+      "step": 11262
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.134909163549387e-05,
+      "loss": 0.944,
+      "step": 11263
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.133357148357897e-05,
+      "loss": 0.9087,
+      "step": 11264
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.131805154199366e-05,
+      "loss": 0.9555,
+      "step": 11265
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.130253181111464e-05,
+      "loss": 0.8393,
+      "step": 11266
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.128701229131848e-05,
+      "loss": 0.9049,
+      "step": 11267
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.127149298298191e-05,
+      "loss": 0.9496,
+      "step": 11268
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.125597388648153e-05,
+      "loss": 0.8789,
+      "step": 11269
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.124045500219398e-05,
+      "loss": 0.9651,
+      "step": 11270
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.122493633049593e-05,
+      "loss": 1.0371,
+      "step": 11271
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.120941787176398e-05,
+      "loss": 0.8118,
+      "step": 11272
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.119389962637481e-05,
+      "loss": 0.8522,
+      "step": 11273
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.117838159470496e-05,
+      "loss": 0.7909,
+      "step": 11274
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.116286377713106e-05,
+      "loss": 0.9814,
+      "step": 11275
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.114734617402972e-05,
+      "loss": 1.5164,
+      "step": 11276
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.11318287857776e-05,
+      "loss": 1.2291,
+      "step": 11277
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.111631161275124e-05,
+      "loss": 0.8118,
+      "step": 11278
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.110079465532726e-05,
+      "loss": 1.0071,
+      "step": 11279
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.108527791388223e-05,
+      "loss": 0.97,
+      "step": 11280
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.106976138879274e-05,
+      "loss": 0.9453,
+      "step": 11281
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.105424508043538e-05,
+      "loss": 0.8031,
+      "step": 11282
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.103872898918672e-05,
+      "loss": 0.9307,
+      "step": 11283
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.102321311542332e-05,
+      "loss": 1.0394,
+      "step": 11284
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.100769745952173e-05,
+      "loss": 0.9322,
+      "step": 11285
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.099218202185852e-05,
+      "loss": 1.3684,
+      "step": 11286
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.097666680281026e-05,
+      "loss": 1.0943,
+      "step": 11287
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.09611518027535e-05,
+      "loss": 0.88,
+      "step": 11288
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.094563702206471e-05,
+      "loss": 0.9171,
+      "step": 11289
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.093012246112048e-05,
+      "loss": 1.1072,
+      "step": 11290
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.091460812029735e-05,
+      "loss": 0.9166,
+      "step": 11291
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.089909399997181e-05,
+      "loss": 0.9177,
+      "step": 11292
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.088358010052043e-05,
+      "loss": 0.9536,
+      "step": 11293
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.086806642231966e-05,
+      "loss": 0.876,
+      "step": 11294
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.085255296574604e-05,
+      "loss": 0.9166,
+      "step": 11295
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.08370397311761e-05,
+      "loss": 0.9149,
+      "step": 11296
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.082152671898629e-05,
+      "loss": 1.0555,
+      "step": 11297
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.080601392955313e-05,
+      "loss": 0.9418,
+      "step": 11298
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.07905013632531e-05,
+      "loss": 0.9081,
+      "step": 11299
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.077498902046266e-05,
+      "loss": 1.09,
+      "step": 11300
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.075947690155835e-05,
+      "loss": 0.9315,
+      "step": 11301
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.074396500691657e-05,
+      "loss": 0.898,
+      "step": 11302
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.072845333691388e-05,
+      "loss": 0.8798,
+      "step": 11303
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.071294189192663e-05,
+      "loss": 0.8155,
+      "step": 11304
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.069743067233131e-05,
+      "loss": 0.7951,
+      "step": 11305
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.068191967850437e-05,
+      "loss": 0.8727,
+      "step": 11306
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.066640891082229e-05,
+      "loss": 0.9268,
+      "step": 11307
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.065089836966145e-05,
+      "loss": 0.806,
+      "step": 11308
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.063538805539835e-05,
+      "loss": 1.1678,
+      "step": 11309
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.061987796840935e-05,
+      "loss": 0.7869,
+      "step": 11310
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.06043681090709e-05,
+      "loss": 0.849,
+      "step": 11311
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.058885847775944e-05,
+      "loss": 1.4171,
+      "step": 11312
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.057334907485134e-05,
+      "loss": 1.2231,
+      "step": 11313
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.055783990072305e-05,
+      "loss": 0.8519,
+      "step": 11314
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.054233095575091e-05,
+      "loss": 1.0446,
+      "step": 11315
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.052682224031135e-05,
+      "loss": 0.9387,
+      "step": 11316
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.05113137547808e-05,
+      "loss": 0.8696,
+      "step": 11317
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.049580549953558e-05,
+      "loss": 0.9708,
+      "step": 11318
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.048029747495206e-05,
+      "loss": 0.864,
+      "step": 11319
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.046478968140664e-05,
+      "loss": 1.5474,
+      "step": 11320
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.044928211927568e-05,
+      "loss": 1.0997,
+      "step": 11321
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.043377478893554e-05,
+      "loss": 0.9682,
+      "step": 11322
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.04182676907626e-05,
+      "loss": 0.9484,
+      "step": 11323
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.040276082513315e-05,
+      "loss": 0.8582,
+      "step": 11324
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.03872541924236e-05,
+      "loss": 0.8851,
+      "step": 11325
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.037174779301023e-05,
+      "loss": 0.904,
+      "step": 11326
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.035624162726941e-05,
+      "loss": 0.7923,
+      "step": 11327
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.034073569557746e-05,
+      "loss": 0.8662,
+      "step": 11328
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.03252299983107e-05,
+      "loss": 0.9703,
+      "step": 11329
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.030972453584543e-05,
+      "loss": 0.9525,
+      "step": 11330
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.029421930855797e-05,
+      "loss": 0.8353,
+      "step": 11331
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.027871431682466e-05,
+      "loss": 0.7971,
+      "step": 11332
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.026320956102172e-05,
+      "loss": 0.9527,
+      "step": 11333
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.024770504152549e-05,
+      "loss": 1.0103,
+      "step": 11334
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.023220075871225e-05,
+      "loss": 0.8238,
+      "step": 11335
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.02166967129583e-05,
+      "loss": 1.0809,
+      "step": 11336
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.020119290463987e-05,
+      "loss": 0.8267,
+      "step": 11337
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.018568933413325e-05,
+      "loss": 0.9063,
+      "step": 11338
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.017018600181473e-05,
+      "loss": 0.8525,
+      "step": 11339
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.015468290806052e-05,
+      "loss": 0.8813,
+      "step": 11340
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.013918005324693e-05,
+      "loss": 0.9376,
+      "step": 11341
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.012367743775015e-05,
+      "loss": 1.0429,
+      "step": 11342
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.010817506194644e-05,
+      "loss": 0.9742,
+      "step": 11343
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.009267292621206e-05,
+      "loss": 0.9089,
+      "step": 11344
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.007717103092319e-05,
+      "loss": 0.9639,
+      "step": 11345
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.006166937645611e-05,
+      "loss": 1.2656,
+      "step": 11346
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.004616796318699e-05,
+      "loss": 1.2524,
+      "step": 11347
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.003066679149207e-05,
+      "loss": 0.9517,
+      "step": 11348
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.001516586174752e-05,
+      "loss": 0.9188,
+      "step": 11349
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.999966517432959e-05,
+      "loss": 0.9103,
+      "step": 11350
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.998416472961441e-05,
+      "loss": 0.8295,
+      "step": 11351
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.996866452797821e-05,
+      "loss": 1.03,
+      "step": 11352
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.995316456979719e-05,
+      "loss": 0.8433,
+      "step": 11353
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.993766485544747e-05,
+      "loss": 0.9379,
+      "step": 11354
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.992216538530527e-05,
+      "loss": 0.8476,
+      "step": 11355
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.990666615974673e-05,
+      "loss": 0.9525,
+      "step": 11356
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.989116717914798e-05,
+      "loss": 0.77,
+      "step": 11357
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.987566844388525e-05,
+      "loss": 0.9233,
+      "step": 11358
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98601699543346e-05,
+      "loss": 0.7903,
+      "step": 11359
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.984467171087226e-05,
+      "loss": 0.9489,
+      "step": 11360
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.982917371387426e-05,
+      "loss": 0.8851,
+      "step": 11361
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.981367596371683e-05,
+      "loss": 0.9252,
+      "step": 11362
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.979817846077602e-05,
+      "loss": 1.0527,
+      "step": 11363
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.978268120542798e-05,
+      "loss": 0.8346,
+      "step": 11364
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.976718419804878e-05,
+      "loss": 0.9815,
+      "step": 11365
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.975168743901458e-05,
+      "loss": 0.874,
+      "step": 11366
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.973619092870145e-05,
+      "loss": 0.8082,
+      "step": 11367
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.972069466748546e-05,
+      "loss": 1.0305,
+      "step": 11368
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.970519865574274e-05,
+      "loss": 0.8953,
+      "step": 11369
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.968970289384935e-05,
+      "loss": 1.0374,
+      "step": 11370
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.967420738218137e-05,
+      "loss": 0.976,
+      "step": 11371
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.965871212111486e-05,
+      "loss": 1.0242,
+      "step": 11372
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.964321711102586e-05,
+      "loss": 0.9568,
+      "step": 11373
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.962772235229048e-05,
+      "loss": 0.8903,
+      "step": 11374
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.961222784528471e-05,
+      "loss": 0.9648,
+      "step": 11375
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.959673359038467e-05,
+      "loss": 0.9807,
+      "step": 11376
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.958123958796634e-05,
+      "loss": 1.3897,
+      "step": 11377
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.956574583840573e-05,
+      "loss": 1.1862,
+      "step": 11378
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.95502523420789e-05,
+      "loss": 0.9872,
+      "step": 11379
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.953475909936186e-05,
+      "loss": 0.9785,
+      "step": 11380
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.951926611063062e-05,
+      "loss": 1.5313,
+      "step": 11381
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.950377337626124e-05,
+      "loss": 1.297,
+      "step": 11382
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.948828089662963e-05,
+      "loss": 0.8143,
+      "step": 11383
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.947278867211183e-05,
+      "loss": 0.8948,
+      "step": 11384
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.945729670308385e-05,
+      "loss": 0.8617,
+      "step": 11385
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.944180498992162e-05,
+      "loss": 0.9403,
+      "step": 11386
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.942631353300118e-05,
+      "loss": 0.9113,
+      "step": 11387
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.941082233269842e-05,
+      "loss": 0.9782,
+      "step": 11388
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.939533138938935e-05,
+      "loss": 0.892,
+      "step": 11389
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.937984070344997e-05,
+      "loss": 0.9188,
+      "step": 11390
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.936435027525618e-05,
+      "loss": 0.9026,
+      "step": 11391
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.934886010518389e-05,
+      "loss": 0.8614,
+      "step": 11392
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.933337019360912e-05,
+      "loss": 0.8234,
+      "step": 11393
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.931788054090771e-05,
+      "loss": 0.8001,
+      "step": 11394
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.930239114745565e-05,
+      "loss": 0.9522,
+      "step": 11395
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.928690201362885e-05,
+      "loss": 0.9277,
+      "step": 11396
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92714131398032e-05,
+      "loss": 1.0781,
+      "step": 11397
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.925592452635466e-05,
+      "loss": 0.9633,
+      "step": 11398
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.924043617365905e-05,
+      "loss": 0.9486,
+      "step": 11399
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.922494808209231e-05,
+      "loss": 0.8327,
+      "step": 11400
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.920946025203035e-05,
+      "loss": 1.001,
+      "step": 11401
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.919397268384902e-05,
+      "loss": 1.0773,
+      "step": 11402
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91784853779242e-05,
+      "loss": 0.8534,
+      "step": 11403
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.916299833463176e-05,
+      "loss": 1.4913,
+      "step": 11404
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.914751155434754e-05,
+      "loss": 1.7688,
+      "step": 11405
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.913202503744747e-05,
+      "loss": 1.2385,
+      "step": 11406
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.911653878430732e-05,
+      "loss": 0.9695,
+      "step": 11407
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.910105279530295e-05,
+      "loss": 0.8975,
+      "step": 11408
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.908556707081024e-05,
+      "loss": 0.9413,
+      "step": 11409
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.907008161120495e-05,
+      "loss": 0.9308,
+      "step": 11410
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.905459641686294e-05,
+      "loss": 0.741,
+      "step": 11411
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.903911148816005e-05,
+      "loss": 0.8175,
+      "step": 11412
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.902362682547207e-05,
+      "loss": 0.8177,
+      "step": 11413
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.90081424291748e-05,
+      "loss": 1.0217,
+      "step": 11414
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.899265829964402e-05,
+      "loss": 0.9504,
+      "step": 11415
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.897717443725556e-05,
+      "loss": 1.0034,
+      "step": 11416
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.89616908423852e-05,
+      "loss": 0.8596,
+      "step": 11417
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.894620751540869e-05,
+      "loss": 1.0588,
+      "step": 11418
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.893072445670185e-05,
+      "loss": 0.993,
+      "step": 11419
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.89152416666404e-05,
+      "loss": 0.9788,
+      "step": 11420
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.889975914560013e-05,
+      "loss": 0.848,
+      "step": 11421
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.888427689395674e-05,
+      "loss": 0.8053,
+      "step": 11422
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.886879491208604e-05,
+      "loss": 0.9605,
+      "step": 11423
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.885331320036373e-05,
+      "loss": 0.9567,
+      "step": 11424
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.883783175916555e-05,
+      "loss": 0.8918,
+      "step": 11425
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.882235058886725e-05,
+      "loss": 0.8391,
+      "step": 11426
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.88068696898445e-05,
+      "loss": 0.901,
+      "step": 11427
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.879138906247309e-05,
+      "loss": 0.9075,
+      "step": 11428
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.877590870712864e-05,
+      "loss": 1.0309,
+      "step": 11429
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.876042862418688e-05,
+      "loss": 0.7627,
+      "step": 11430
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.874494881402357e-05,
+      "loss": 0.9449,
+      "step": 11431
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.872946927701429e-05,
+      "loss": 0.9341,
+      "step": 11432
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.87139900135348e-05,
+      "loss": 0.9573,
+      "step": 11433
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.869851102396073e-05,
+      "loss": 0.7963,
+      "step": 11434
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.86830323086678e-05,
+      "loss": 0.9826,
+      "step": 11435
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.86675538680316e-05,
+      "loss": 0.8489,
+      "step": 11436
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.865207570242781e-05,
+      "loss": 1.3717,
+      "step": 11437
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.863659781223209e-05,
+      "loss": 1.1466,
+      "step": 11438
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.862112019782007e-05,
+      "loss": 0.7937,
+      "step": 11439
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.860564285956737e-05,
+      "loss": 0.9511,
+      "step": 11440
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.859016579784964e-05,
+      "loss": 0.7014,
+      "step": 11441
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.857468901304251e-05,
+      "loss": 0.8784,
+      "step": 11442
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.855921250552155e-05,
+      "loss": 0.9916,
+      "step": 11443
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.854373627566241e-05,
+      "loss": 0.8593,
+      "step": 11444
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.852826032384067e-05,
+      "loss": 0.91,
+      "step": 11445
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.85127846504319e-05,
+      "loss": 0.8154,
+      "step": 11446
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.849730925581174e-05,
+      "loss": 0.8193,
+      "step": 11447
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.848183414035574e-05,
+      "loss": 0.9516,
+      "step": 11448
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.846635930443948e-05,
+      "loss": 1.0625,
+      "step": 11449
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.845088474843853e-05,
+      "loss": 1.1561,
+      "step": 11450
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.84354104727284e-05,
+      "loss": 0.8421,
+      "step": 11451
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.841993647768469e-05,
+      "loss": 0.9043,
+      "step": 11452
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.840446276368295e-05,
+      "loss": 0.8925,
+      "step": 11453
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.838898933109869e-05,
+      "loss": 1.025,
+      "step": 11454
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.837351618030747e-05,
+      "loss": 0.8742,
+      "step": 11455
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.835804331168479e-05,
+      "loss": 0.9962,
+      "step": 11456
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.834257072560618e-05,
+      "loss": 0.8658,
+      "step": 11457
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.832709842244718e-05,
+      "loss": 1.4508,
+      "step": 11458
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.831162640258322e-05,
+      "loss": 1.2286,
+      "step": 11459
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.82961546663899e-05,
+      "loss": 0.7537,
+      "step": 11460
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.828068321424262e-05,
+      "loss": 1.0377,
+      "step": 11461
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.826521204651688e-05,
+      "loss": 0.9415,
+      "step": 11462
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.824974116358821e-05,
+      "loss": 0.8983,
+      "step": 11463
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.823427056583208e-05,
+      "loss": 1.4691,
+      "step": 11464
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.821880025362387e-05,
+      "loss": 1.2662,
+      "step": 11465
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.820333022733911e-05,
+      "loss": 0.9771,
+      "step": 11466
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.81878604873532e-05,
+      "loss": 1.4199,
+      "step": 11467
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.81723910340416e-05,
+      "loss": 1.2477,
+      "step": 11468
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.815692186777977e-05,
+      "loss": 0.8843,
+      "step": 11469
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.81414529889431e-05,
+      "loss": 0.8252,
+      "step": 11470
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.812598439790708e-05,
+      "loss": 0.9193,
+      "step": 11471
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.811051609504704e-05,
+      "loss": 1.033,
+      "step": 11472
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.809504808073841e-05,
+      "loss": 0.9454,
+      "step": 11473
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.807958035535664e-05,
+      "loss": 0.9392,
+      "step": 11474
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.806411291927707e-05,
+      "loss": 0.8533,
+      "step": 11475
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.804864577287511e-05,
+      "loss": 0.9899,
+      "step": 11476
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.803317891652613e-05,
+      "loss": 0.8,
+      "step": 11477
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.80177123506055e-05,
+      "loss": 0.9106,
+      "step": 11478
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.800224607548863e-05,
+      "loss": 0.8729,
+      "step": 11479
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.798678009155081e-05,
+      "loss": 1.0546,
+      "step": 11480
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.797131439916741e-05,
+      "loss": 0.9756,
+      "step": 11481
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.795584899871381e-05,
+      "loss": 0.9328,
+      "step": 11482
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.79403838905653e-05,
+      "loss": 0.9875,
+      "step": 11483
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.792491907509722e-05,
+      "loss": 0.878,
+      "step": 11484
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.790945455268495e-05,
+      "loss": 0.9403,
+      "step": 11485
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.789399032370372e-05,
+      "loss": 1.0352,
+      "step": 11486
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.787852638852888e-05,
+      "loss": 0.8049,
+      "step": 11487
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.786306274753575e-05,
+      "loss": 0.9662,
+      "step": 11488
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.784759940109958e-05,
+      "loss": 0.7935,
+      "step": 11489
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.783213634959573e-05,
+      "loss": 0.8191,
+      "step": 11490
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.781667359339938e-05,
+      "loss": 0.9517,
+      "step": 11491
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.780121113288587e-05,
+      "loss": 0.8802,
+      "step": 11492
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.778574896843048e-05,
+      "loss": 0.936,
+      "step": 11493
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.777028710040844e-05,
+      "loss": 0.8215,
+      "step": 11494
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.775482552919496e-05,
+      "loss": 0.8184,
+      "step": 11495
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.773936425516536e-05,
+      "loss": 0.8057,
+      "step": 11496
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.77239032786948e-05,
+      "loss": 0.9851,
+      "step": 11497
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.770844260015857e-05,
+      "loss": 0.9468,
+      "step": 11498
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.769298221993188e-05,
+      "loss": 0.8445,
+      "step": 11499
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.767752213838993e-05,
+      "loss": 0.8442,
+      "step": 11500
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.766206235590794e-05,
+      "loss": 0.8804,
+      "step": 11501
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.764660287286111e-05,
+      "loss": 0.8815,
+      "step": 11502
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.763114368962461e-05,
+      "loss": 1.0215,
+      "step": 11503
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.761568480657368e-05,
+      "loss": 0.802,
+      "step": 11504
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.760022622408344e-05,
+      "loss": 1.0662,
+      "step": 11505
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.75847679425291e-05,
+      "loss": 1.0073,
+      "step": 11506
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.756930996228579e-05,
+      "loss": 0.821,
+      "step": 11507
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.755385228372869e-05,
+      "loss": 0.8365,
+      "step": 11508
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.753839490723301e-05,
+      "loss": 1.3708,
+      "step": 11509
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.752293783317378e-05,
+      "loss": 1.22,
+      "step": 11510
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.750748106192619e-05,
+      "loss": 1.0108,
+      "step": 11511
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.749202459386536e-05,
+      "loss": 0.8577,
+      "step": 11512
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.74765684293664e-05,
+      "loss": 0.9864,
+      "step": 11513
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.746111256880445e-05,
+      "loss": 0.9662,
+      "step": 11514
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.74456570125546e-05,
+      "loss": 0.8111,
+      "step": 11515
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.743020176099194e-05,
+      "loss": 0.968,
+      "step": 11516
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.741474681449158e-05,
+      "loss": 0.9029,
+      "step": 11517
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.739929217342858e-05,
+      "loss": 0.8874,
+      "step": 11518
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.738383783817803e-05,
+      "loss": 0.9099,
+      "step": 11519
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.736838380911502e-05,
+      "loss": 0.8124,
+      "step": 11520
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.735293008661456e-05,
+      "loss": 0.8363,
+      "step": 11521
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.733747667105176e-05,
+      "loss": 0.841,
+      "step": 11522
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.732202356280165e-05,
+      "loss": 0.9407,
+      "step": 11523
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.730657076223922e-05,
+      "loss": 1.0841,
+      "step": 11524
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.729111826973954e-05,
+      "loss": 1.0301,
+      "step": 11525
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.727566608567766e-05,
+      "loss": 1.492,
+      "step": 11526
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.726021421042853e-05,
+      "loss": 1.2073,
+      "step": 11527
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.724476264436723e-05,
+      "loss": 0.8123,
+      "step": 11528
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.72293113878687e-05,
+      "loss": 0.8719,
+      "step": 11529
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.721386044130796e-05,
+      "loss": 0.907,
+      "step": 11530
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.719840980506002e-05,
+      "loss": 0.9035,
+      "step": 11531
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.718295947949983e-05,
+      "loss": 0.9683,
+      "step": 11532
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.716750946500237e-05,
+      "loss": 0.9979,
+      "step": 11533
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.71520597619426e-05,
+      "loss": 0.933,
+      "step": 11534
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.713661037069546e-05,
+      "loss": 1.0819,
+      "step": 11535
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.712116129163594e-05,
+      "loss": 1.6011,
+      "step": 11536
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.710571252513893e-05,
+      "loss": 1.0805,
+      "step": 11537
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.709026407157946e-05,
+      "loss": 0.918,
+      "step": 11538
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.707481593133233e-05,
+      "loss": 0.8634,
+      "step": 11539
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.705936810477252e-05,
+      "loss": 0.9221,
+      "step": 11540
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.704392059227492e-05,
+      "loss": 1.4276,
+      "step": 11541
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.702847339421447e-05,
+      "loss": 1.1007,
+      "step": 11542
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.701302651096602e-05,
+      "loss": 0.9692,
+      "step": 11543
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.699757994290451e-05,
+      "loss": 0.9143,
+      "step": 11544
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.698213369040475e-05,
+      "loss": 0.8978,
+      "step": 11545
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.696668775384166e-05,
+      "loss": 0.8605,
+      "step": 11546
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.695124213359014e-05,
+      "loss": 0.8502,
+      "step": 11547
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.693579683002496e-05,
+      "loss": 0.7883,
+      "step": 11548
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.692035184352105e-05,
+      "loss": 0.8946,
+      "step": 11549
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.690490717445317e-05,
+      "loss": 0.9319,
+      "step": 11550
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.688946282319622e-05,
+      "loss": 0.8279,
+      "step": 11551
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.687401879012502e-05,
+      "loss": 1.0507,
+      "step": 11552
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.68585750756144e-05,
+      "loss": 0.9342,
+      "step": 11553
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.684313168003909e-05,
+      "loss": 0.9061,
+      "step": 11554
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.682768860377395e-05,
+      "loss": 0.8777,
+      "step": 11555
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.68122458471938e-05,
+      "loss": 1.5182,
+      "step": 11556
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.679680341067337e-05,
+      "loss": 1.2784,
+      "step": 11557
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.678136129458749e-05,
+      "loss": 1.172,
+      "step": 11558
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.67659194993109e-05,
+      "loss": 0.8617,
+      "step": 11559
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.675047802521838e-05,
+      "loss": 0.9173,
+      "step": 11560
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.673503687268468e-05,
+      "loss": 0.926,
+      "step": 11561
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.671959604208455e-05,
+      "loss": 0.9484,
+      "step": 11562
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.670415553379274e-05,
+      "loss": 0.8412,
+      "step": 11563
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.668871534818396e-05,
+      "loss": 0.9394,
+      "step": 11564
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.667327548563294e-05,
+      "loss": 0.9232,
+      "step": 11565
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.665783594651444e-05,
+      "loss": 0.8949,
+      "step": 11566
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.664239673120315e-05,
+      "loss": 0.9325,
+      "step": 11567
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.66269578400737e-05,
+      "loss": 0.9196,
+      "step": 11568
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.661151927350088e-05,
+      "loss": 0.9584,
+      "step": 11569
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.659608103185932e-05,
+      "loss": 0.8818,
+      "step": 11570
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.65806431155237e-05,
+      "loss": 0.9322,
+      "step": 11571
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.656520552486873e-05,
+      "loss": 0.9803,
+      "step": 11572
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.654976826026903e-05,
+      "loss": 0.86,
+      "step": 11573
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.65343313220993e-05,
+      "loss": 1.0925,
+      "step": 11574
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.651889471073412e-05,
+      "loss": 0.9041,
+      "step": 11575
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.650345842654815e-05,
+      "loss": 0.9552,
+      "step": 11576
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.648802246991608e-05,
+      "loss": 0.7986,
+      "step": 11577
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.647258684121246e-05,
+      "loss": 0.9971,
+      "step": 11578
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.645715154081197e-05,
+      "loss": 0.7768,
+      "step": 11579
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.644171656908913e-05,
+      "loss": 0.8917,
+      "step": 11580
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.64262819264186e-05,
+      "loss": 0.7975,
+      "step": 11581
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.641084761317501e-05,
+      "loss": 0.87,
+      "step": 11582
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.639541362973286e-05,
+      "loss": 1.0652,
+      "step": 11583
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.637997997646674e-05,
+      "loss": 0.8918,
+      "step": 11584
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.636454665375124e-05,
+      "loss": 0.8107,
+      "step": 11585
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.634911366196089e-05,
+      "loss": 0.8818,
+      "step": 11586
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.633368100147025e-05,
+      "loss": 0.8473,
+      "step": 11587
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.631824867265392e-05,
+      "loss": 0.801,
+      "step": 11588
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.630281667588636e-05,
+      "loss": 0.9136,
+      "step": 11589
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.628738501154215e-05,
+      "loss": 0.8715,
+      "step": 11590
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.627195367999575e-05,
+      "loss": 0.9309,
+      "step": 11591
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.62565226816217e-05,
+      "loss": 0.8955,
+      "step": 11592
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.624109201679452e-05,
+      "loss": 0.9104,
+      "step": 11593
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.622566168588867e-05,
+      "loss": 1.0726,
+      "step": 11594
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.621023168927869e-05,
+      "loss": 0.8411,
+      "step": 11595
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.619480202733899e-05,
+      "loss": 0.9325,
+      "step": 11596
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.61793727004441e-05,
+      "loss": 0.866,
+      "step": 11597
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.616394370896843e-05,
+      "loss": 0.841,
+      "step": 11598
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.614851505328647e-05,
+      "loss": 0.8544,
+      "step": 11599
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.613308673377264e-05,
+      "loss": 1.0364,
+      "step": 11600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.61176587508014e-05,
+      "loss": 0.8976,
+      "step": 11601
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.610223110474715e-05,
+      "loss": 0.9414,
+      "step": 11602
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.608680379598432e-05,
+      "loss": 0.8614,
+      "step": 11603
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.607137682488736e-05,
+      "loss": 0.8461,
+      "step": 11604
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.605595019183063e-05,
+      "loss": 0.9601,
+      "step": 11605
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.604052389718855e-05,
+      "loss": 0.984,
+      "step": 11606
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.602509794133549e-05,
+      "loss": 0.8126,
+      "step": 11607
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.600967232464584e-05,
+      "loss": 0.9852,
+      "step": 11608
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.599424704749397e-05,
+      "loss": 0.8166,
+      "step": 11609
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.597882211025423e-05,
+      "loss": 0.944,
+      "step": 11610
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.596339751330105e-05,
+      "loss": 0.976,
+      "step": 11611
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.594797325700867e-05,
+      "loss": 0.8054,
+      "step": 11612
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.593254934175147e-05,
+      "loss": 0.9539,
+      "step": 11613
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.591712576790378e-05,
+      "loss": 0.9282,
+      "step": 11614
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.590170253583994e-05,
+      "loss": 0.8975,
+      "step": 11615
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.588627964593423e-05,
+      "loss": 0.7939,
+      "step": 11616
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.587085709856099e-05,
+      "loss": 0.9608,
+      "step": 11617
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.58554348940945e-05,
+      "loss": 0.9678,
+      "step": 11618
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.584001303290904e-05,
+      "loss": 0.8821,
+      "step": 11619
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.582459151537892e-05,
+      "loss": 0.9238,
+      "step": 11620
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.580917034187837e-05,
+      "loss": 0.8354,
+      "step": 11621
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.579374951278171e-05,
+      "loss": 0.8815,
+      "step": 11622
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.577832902846313e-05,
+      "loss": 0.9619,
+      "step": 11623
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.576290888929691e-05,
+      "loss": 1.375,
+      "step": 11624
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.57474890956573e-05,
+      "loss": 1.1798,
+      "step": 11625
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.573206964791856e-05,
+      "loss": 0.9576,
+      "step": 11626
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.571665054645482e-05,
+      "loss": 0.8516,
+      "step": 11627
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.570123179164034e-05,
+      "loss": 1.0351,
+      "step": 11628
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.568581338384933e-05,
+      "loss": 0.8724,
+      "step": 11629
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.567039532345598e-05,
+      "loss": 0.8265,
+      "step": 11630
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.565497761083449e-05,
+      "loss": 0.8593,
+      "step": 11631
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.563956024635904e-05,
+      "loss": 0.8492,
+      "step": 11632
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.562414323040378e-05,
+      "loss": 0.8346,
+      "step": 11633
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.56087265633429e-05,
+      "loss": 0.7324,
+      "step": 11634
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.55933102455505e-05,
+      "loss": 0.8762,
+      "step": 11635
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.557789427740081e-05,
+      "loss": 0.9749,
+      "step": 11636
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.55624786592679e-05,
+      "loss": 1.0725,
+      "step": 11637
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.554706339152591e-05,
+      "loss": 1.0754,
+      "step": 11638
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.553164847454902e-05,
+      "loss": 0.9722,
+      "step": 11639
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.551623390871125e-05,
+      "loss": 0.7683,
+      "step": 11640
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.550081969438681e-05,
+      "loss": 0.9399,
+      "step": 11641
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.548540583194968e-05,
+      "loss": 1.0065,
+      "step": 11642
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.546999232177399e-05,
+      "loss": 0.9078,
+      "step": 11643
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.545457916423384e-05,
+      "loss": 0.843,
+      "step": 11644
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.54391663597033e-05,
+      "loss": 0.9202,
+      "step": 11645
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.542375390855638e-05,
+      "loss": 0.8906,
+      "step": 11646
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.54083418111672e-05,
+      "loss": 0.8814,
+      "step": 11647
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.539293006790974e-05,
+      "loss": 0.8429,
+      "step": 11648
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.537751867915809e-05,
+      "loss": 0.8719,
+      "step": 11649
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.536210764528625e-05,
+      "loss": 0.9147,
+      "step": 11650
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.534669696666822e-05,
+      "loss": 0.7977,
+      "step": 11651
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.533128664367806e-05,
+      "loss": 0.9068,
+      "step": 11652
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.53158766766897e-05,
+      "loss": 0.8992,
+      "step": 11653
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.53004670660772e-05,
+      "loss": 1.008,
+      "step": 11654
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.52850578122145e-05,
+      "loss": 1.4207,
+      "step": 11655
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.526964891547563e-05,
+      "loss": 1.2453,
+      "step": 11656
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.525424037623447e-05,
+      "loss": 1.03,
+      "step": 11657
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.523883219486503e-05,
+      "loss": 1.023,
+      "step": 11658
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.522342437174123e-05,
+      "loss": 0.9274,
+      "step": 11659
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.520801690723703e-05,
+      "loss": 1.2025,
+      "step": 11660
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.519260980172639e-05,
+      "loss": 0.7959,
+      "step": 11661
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.517720305558317e-05,
+      "loss": 0.9729,
+      "step": 11662
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.516179666918134e-05,
+      "loss": 0.9485,
+      "step": 11663
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.514639064289476e-05,
+      "loss": 0.9141,
+      "step": 11664
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.513098497709734e-05,
+      "loss": 1.3444,
+      "step": 11665
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.5115579672163e-05,
+      "loss": 1.3333,
+      "step": 11666
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.510017472846556e-05,
+      "loss": 0.8671,
+      "step": 11667
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.508477014637896e-05,
+      "loss": 0.8149,
+      "step": 11668
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.506936592627699e-05,
+      "loss": 0.9084,
+      "step": 11669
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.505396206853358e-05,
+      "loss": 0.8837,
+      "step": 11670
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.503855857352249e-05,
+      "loss": 0.9263,
+      "step": 11671
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.50231554416176e-05,
+      "loss": 0.9178,
+      "step": 11672
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.500775267319271e-05,
+      "loss": 0.9959,
+      "step": 11673
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.499235026862169e-05,
+      "loss": 0.879,
+      "step": 11674
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.497694822827828e-05,
+      "loss": 0.7079,
+      "step": 11675
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.496154655253632e-05,
+      "loss": 0.8527,
+      "step": 11676
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.49461452417696e-05,
+      "loss": 0.8545,
+      "step": 11677
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.493074429635189e-05,
+      "loss": 0.8148,
+      "step": 11678
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.491534371665698e-05,
+      "loss": 0.8347,
+      "step": 11679
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.48999435030586e-05,
+      "loss": 0.7557,
+      "step": 11680
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.488454365593053e-05,
+      "loss": 0.9768,
+      "step": 11681
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.486914417564654e-05,
+      "loss": 0.9817,
+      "step": 11682
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.48537450625803e-05,
+      "loss": 0.9488,
+      "step": 11683
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.483834631710561e-05,
+      "loss": 0.9064,
+      "step": 11684
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.482294793959616e-05,
+      "loss": 0.9349,
+      "step": 11685
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.480754993042565e-05,
+      "loss": 0.9241,
+      "step": 11686
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.479215228996777e-05,
+      "loss": 0.952,
+      "step": 11687
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.477675501859626e-05,
+      "loss": 0.85,
+      "step": 11688
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.476135811668473e-05,
+      "loss": 0.8095,
+      "step": 11689
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.474596158460692e-05,
+      "loss": 0.8749,
+      "step": 11690
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.47305654227365e-05,
+      "loss": 0.8621,
+      "step": 11691
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.471516963144706e-05,
+      "loss": 0.8917,
+      "step": 11692
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.469977421111233e-05,
+      "loss": 0.8582,
+      "step": 11693
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.468437916210588e-05,
+      "loss": 0.9848,
+      "step": 11694
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.466898448480138e-05,
+      "loss": 1.0819,
+      "step": 11695
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.465359017957244e-05,
+      "loss": 1.0314,
+      "step": 11696
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.463819624679266e-05,
+      "loss": 0.7689,
+      "step": 11697
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.462280268683566e-05,
+      "loss": 0.8774,
+      "step": 11698
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.460740950007501e-05,
+      "loss": 0.9398,
+      "step": 11699
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.459201668688437e-05,
+      "loss": 0.8773,
+      "step": 11700
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.45766242476372e-05,
+      "loss": 0.9161,
+      "step": 11701
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.456123218270713e-05,
+      "loss": 0.8883,
+      "step": 11702
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.454584049246771e-05,
+      "loss": 0.9894,
+      "step": 11703
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.45304491772925e-05,
+      "loss": 1.4518,
+      "step": 11704
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.4515058237555e-05,
+      "loss": 1.2715,
+      "step": 11705
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.449966767362877e-05,
+      "loss": 0.876,
+      "step": 11706
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.448427748588734e-05,
+      "loss": 1.0663,
+      "step": 11707
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.44688876747042e-05,
+      "loss": 0.9031,
+      "step": 11708
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.445349824045288e-05,
+      "loss": 1.0261,
+      "step": 11709
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.443810918350683e-05,
+      "loss": 0.7831,
+      "step": 11710
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.442272050423956e-05,
+      "loss": 1.0862,
+      "step": 11711
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.440733220302456e-05,
+      "loss": 0.9001,
+      "step": 11712
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.439194428023526e-05,
+      "loss": 0.8462,
+      "step": 11713
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.437655673624518e-05,
+      "loss": 1.3998,
+      "step": 11714
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.43611695714277e-05,
+      "loss": 1.2614,
+      "step": 11715
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.434578278615625e-05,
+      "loss": 0.9701,
+      "step": 11716
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.43303963808043e-05,
+      "loss": 1.063,
+      "step": 11717
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.431501035574531e-05,
+      "loss": 0.9251,
+      "step": 11718
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.429962471135258e-05,
+      "loss": 0.8779,
+      "step": 11719
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.428423944799962e-05,
+      "loss": 0.9595,
+      "step": 11720
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.426885456605973e-05,
+      "loss": 0.8781,
+      "step": 11721
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.425347006590636e-05,
+      "loss": 0.8043,
+      "step": 11722
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.423808594791287e-05,
+      "loss": 0.8929,
+      "step": 11723
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.42227022124526e-05,
+      "loss": 0.8453,
+      "step": 11724
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.420731885989895e-05,
+      "loss": 1.0246,
+      "step": 11725
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.419193589062519e-05,
+      "loss": 1.0197,
+      "step": 11726
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.417655330500471e-05,
+      "loss": 0.8109,
+      "step": 11727
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.416117110341087e-05,
+      "loss": 0.9538,
+      "step": 11728
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.414578928621695e-05,
+      "loss": 0.9274,
+      "step": 11729
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.413040785379622e-05,
+      "loss": 0.8813,
+      "step": 11730
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.411502680652204e-05,
+      "loss": 1.4269,
+      "step": 11731
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.409964614476765e-05,
+      "loss": 1.1134,
+      "step": 11732
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.408426586890635e-05,
+      "loss": 0.9146,
+      "step": 11733
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.406888597931145e-05,
+      "loss": 0.9442,
+      "step": 11734
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.405350647635615e-05,
+      "loss": 0.9124,
+      "step": 11735
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.403812736041376e-05,
+      "loss": 0.9007,
+      "step": 11736
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.402274863185746e-05,
+      "loss": 0.8953,
+      "step": 11737
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.400737029106052e-05,
+      "loss": 0.8651,
+      "step": 11738
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.399199233839619e-05,
+      "loss": 0.8001,
+      "step": 11739
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.397661477423763e-05,
+      "loss": 0.8637,
+      "step": 11740
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.396123759895809e-05,
+      "loss": 0.9942,
+      "step": 11741
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.394586081293073e-05,
+      "loss": 0.9673,
+      "step": 11742
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.393048441652876e-05,
+      "loss": 0.8777,
+      "step": 11743
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.391510841012537e-05,
+      "loss": 0.8759,
+      "step": 11744
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.38997327940937e-05,
+      "loss": 0.8265,
+      "step": 11745
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.38843575688069e-05,
+      "loss": 0.9156,
+      "step": 11746
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.386898273463816e-05,
+      "loss": 0.9372,
+      "step": 11747
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.385360829196056e-05,
+      "loss": 1.0103,
+      "step": 11748
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.383823424114728e-05,
+      "loss": 0.7169,
+      "step": 11749
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.382286058257143e-05,
+      "loss": 0.9654,
+      "step": 11750
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.380748731660609e-05,
+      "loss": 1.0416,
+      "step": 11751
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.37921144436244e-05,
+      "loss": 0.8661,
+      "step": 11752
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.377674196399943e-05,
+      "loss": 0.9093,
+      "step": 11753
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.376136987810426e-05,
+      "loss": 1.0398,
+      "step": 11754
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.374599818631199e-05,
+      "loss": 1.0063,
+      "step": 11755
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.373062688899564e-05,
+      "loss": 1.0492,
+      "step": 11756
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.371525598652831e-05,
+      "loss": 1.0298,
+      "step": 11757
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.369988547928302e-05,
+      "loss": 1.0374,
+      "step": 11758
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.368451536763281e-05,
+      "loss": 1.2411,
+      "step": 11759
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.366914565195067e-05,
+      "loss": 1.3747,
+      "step": 11760
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.365377633260967e-05,
+      "loss": 0.9444,
+      "step": 11761
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.363840740998276e-05,
+      "loss": 0.8508,
+      "step": 11762
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.362303888444296e-05,
+      "loss": 0.9868,
+      "step": 11763
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.360767075636328e-05,
+      "loss": 0.9307,
+      "step": 11764
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.359230302611666e-05,
+      "loss": 0.7943,
+      "step": 11765
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.35769356940761e-05,
+      "loss": 0.7116,
+      "step": 11766
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.356156876061451e-05,
+      "loss": 0.8784,
+      "step": 11767
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.354620222610488e-05,
+      "loss": 0.8978,
+      "step": 11768
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.353083609092017e-05,
+      "loss": 0.9406,
+      "step": 11769
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.351547035543322e-05,
+      "loss": 0.9336,
+      "step": 11770
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.350010502001704e-05,
+      "loss": 0.9719,
+      "step": 11771
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.348474008504447e-05,
+      "loss": 0.8727,
+      "step": 11772
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.346937555088848e-05,
+      "loss": 1.4329,
+      "step": 11773
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.345401141792187e-05,
+      "loss": 1.2127,
+      "step": 11774
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.343864768651759e-05,
+      "loss": 0.9917,
+      "step": 11775
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.342328435704846e-05,
+      "loss": 1.5269,
+      "step": 11776
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.34079214298874e-05,
+      "loss": 1.2798,
+      "step": 11777
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.339255890540719e-05,
+      "loss": 0.8214,
+      "step": 11778
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.337719678398071e-05,
+      "loss": 0.8631,
+      "step": 11779
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.33618350659808e-05,
+      "loss": 0.8897,
+      "step": 11780
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.334647375178026e-05,
+      "loss": 0.9024,
+      "step": 11781
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.333111284175193e-05,
+      "loss": 0.9049,
+      "step": 11782
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.331575233626854e-05,
+      "loss": 0.9808,
+      "step": 11783
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.330039223570296e-05,
+      "loss": 1.0894,
+      "step": 11784
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.328503254042794e-05,
+      "loss": 0.9275,
+      "step": 11785
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.326967325081625e-05,
+      "loss": 0.9477,
+      "step": 11786
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.325431436724067e-05,
+      "loss": 0.7762,
+      "step": 11787
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.323895589007395e-05,
+      "loss": 0.9221,
+      "step": 11788
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.322359781968878e-05,
+      "loss": 0.9678,
+      "step": 11789
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.320824015645795e-05,
+      "loss": 0.9001,
+      "step": 11790
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.319288290075416e-05,
+      "loss": 0.8763,
+      "step": 11791
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.31775260529501e-05,
+      "loss": 0.9857,
+      "step": 11792
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.316216961341853e-05,
+      "loss": 0.8814,
+      "step": 11793
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.314681358253208e-05,
+      "loss": 0.9807,
+      "step": 11794
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.313145796066348e-05,
+      "loss": 0.9137,
+      "step": 11795
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.311610274818539e-05,
+      "loss": 0.9217,
+      "step": 11796
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.310074794547047e-05,
+      "loss": 0.9848,
+      "step": 11797
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.308539355289136e-05,
+      "loss": 0.8184,
+      "step": 11798
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.307003957082072e-05,
+      "loss": 0.8652,
+      "step": 11799
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.305468599963116e-05,
+      "loss": 0.9734,
+      "step": 11800
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.303933283969536e-05,
+      "loss": 0.8625,
+      "step": 11801
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.302398009138585e-05,
+      "loss": 1.0292,
+      "step": 11802
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.300862775507533e-05,
+      "loss": 0.9324,
+      "step": 11803
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.299327583113632e-05,
+      "loss": 0.87,
+      "step": 11804
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.29779243199414e-05,
+      "loss": 0.9732,
+      "step": 11805
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.296257322186316e-05,
+      "loss": 0.9648,
+      "step": 11806
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.294722253727419e-05,
+      "loss": 0.7844,
+      "step": 11807
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.2931872266547e-05,
+      "loss": 0.9441,
+      "step": 11808
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.291652241005418e-05,
+      "loss": 0.8819,
+      "step": 11809
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.290117296816822e-05,
+      "loss": 0.91,
+      "step": 11810
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.288582394126166e-05,
+      "loss": 0.9977,
+      "step": 11811
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.287047532970702e-05,
+      "loss": 0.738,
+      "step": 11812
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.285512713387677e-05,
+      "loss": 0.9376,
+      "step": 11813
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.283977935414346e-05,
+      "loss": 0.9112,
+      "step": 11814
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.282443199087953e-05,
+      "loss": 0.8772,
+      "step": 11815
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.280908504445745e-05,
+      "loss": 0.9317,
+      "step": 11816
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.279373851524973e-05,
+      "loss": 0.8359,
+      "step": 11817
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.277839240362877e-05,
+      "loss": 0.8299,
+      "step": 11818
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.276304670996701e-05,
+      "loss": 0.757,
+      "step": 11819
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.274770143463691e-05,
+      "loss": 0.8009,
+      "step": 11820
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.273235657801089e-05,
+      "loss": 0.9317,
+      "step": 11821
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.271701214046133e-05,
+      "loss": 0.8893,
+      "step": 11822
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.270166812236067e-05,
+      "loss": 0.9349,
+      "step": 11823
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.268632452408127e-05,
+      "loss": 0.7683,
+      "step": 11824
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.267098134599553e-05,
+      "loss": 0.9617,
+      "step": 11825
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.265563858847584e-05,
+      "loss": 0.7452,
+      "step": 11826
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.264029625189451e-05,
+      "loss": 0.9501,
+      "step": 11827
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.262495433662395e-05,
+      "loss": 0.877,
+      "step": 11828
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.260961284303643e-05,
+      "loss": 0.7301,
+      "step": 11829
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.259427177150434e-05,
+      "loss": 1.0654,
+      "step": 11830
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.257893112239999e-05,
+      "loss": 0.9367,
+      "step": 11831
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.256359089609569e-05,
+      "loss": 0.9123,
+      "step": 11832
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.254825109296369e-05,
+      "loss": 0.9627,
+      "step": 11833
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.253291171337633e-05,
+      "loss": 0.7628,
+      "step": 11834
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.251757275770587e-05,
+      "loss": 0.9246,
+      "step": 11835
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.250223422632456e-05,
+      "loss": 0.7168,
+      "step": 11836
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.248689611960472e-05,
+      "loss": 0.9257,
+      "step": 11837
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.247155843791851e-05,
+      "loss": 0.8467,
+      "step": 11838
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.245622118163826e-05,
+      "loss": 0.9554,
+      "step": 11839
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.244088435113612e-05,
+      "loss": 0.8679,
+      "step": 11840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.242554794678434e-05,
+      "loss": 0.9671,
+      "step": 11841
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.241021196895516e-05,
+      "loss": 0.8047,
+      "step": 11842
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.23948764180207e-05,
+      "loss": 0.929,
+      "step": 11843
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.237954129435322e-05,
+      "loss": 0.8406,
+      "step": 11844
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.236420659832482e-05,
+      "loss": 0.8359,
+      "step": 11845
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.234887233030773e-05,
+      "loss": 1.0795,
+      "step": 11846
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.233353849067412e-05,
+      "loss": 0.9302,
+      "step": 11847
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.231820507979606e-05,
+      "loss": 0.8979,
+      "step": 11848
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.23028720980457e-05,
+      "loss": 0.8794,
+      "step": 11849
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.228753954579521e-05,
+      "loss": 0.8192,
+      "step": 11850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.227220742341666e-05,
+      "loss": 1.4259,
+      "step": 11851
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.225687573128216e-05,
+      "loss": 1.3273,
+      "step": 11852
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.224154446976383e-05,
+      "loss": 0.8666,
+      "step": 11853
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.222621363923371e-05,
+      "loss": 0.7998,
+      "step": 11854
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.221088324006392e-05,
+      "loss": 1.478,
+      "step": 11855
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.219555327262645e-05,
+      "loss": 1.2011,
+      "step": 11856
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.218022373729341e-05,
+      "loss": 0.9115,
+      "step": 11857
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.216489463443684e-05,
+      "loss": 1.0295,
+      "step": 11858
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.214956596442873e-05,
+      "loss": 0.8964,
+      "step": 11859
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.213423772764114e-05,
+      "loss": 0.9158,
+      "step": 11860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.211890992444608e-05,
+      "loss": 0.9999,
+      "step": 11861
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.21035825552155e-05,
+      "loss": 1.4135,
+      "step": 11862
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.20882556203214e-05,
+      "loss": 1.1251,
+      "step": 11863
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.207292912013578e-05,
+      "loss": 0.8163,
+      "step": 11864
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.205760305503059e-05,
+      "loss": 0.7755,
+      "step": 11865
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.20422774253778e-05,
+      "loss": 0.734,
+      "step": 11866
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.202695223154933e-05,
+      "loss": 1.1102,
+      "step": 11867
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.201162747391713e-05,
+      "loss": 1.0495,
+      "step": 11868
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.199630315285314e-05,
+      "loss": 0.7801,
+      "step": 11869
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.198097926872924e-05,
+      "loss": 1.0372,
+      "step": 11870
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.196565582191738e-05,
+      "loss": 0.9206,
+      "step": 11871
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.195033281278937e-05,
+      "loss": 1.0396,
+      "step": 11872
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.193501024171716e-05,
+      "loss": 0.9647,
+      "step": 11873
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.191968810907264e-05,
+      "loss": 0.9988,
+      "step": 11874
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.19043664152276e-05,
+      "loss": 0.8238,
+      "step": 11875
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.188904516055395e-05,
+      "loss": 0.8423,
+      "step": 11876
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.187372434542349e-05,
+      "loss": 0.9294,
+      "step": 11877
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.185840397020805e-05,
+      "loss": 0.8758,
+      "step": 11878
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.184308403527944e-05,
+      "loss": 0.9536,
+      "step": 11879
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.182776454100951e-05,
+      "loss": 0.8632,
+      "step": 11880
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.181244548777e-05,
+      "loss": 1.0149,
+      "step": 11881
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.179712687593275e-05,
+      "loss": 1.0705,
+      "step": 11882
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.178180870586949e-05,
+      "loss": 0.8496,
+      "step": 11883
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.176649097795201e-05,
+      "loss": 1.0482,
+      "step": 11884
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.175117369255206e-05,
+      "loss": 1.0584,
+      "step": 11885
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.173585685004138e-05,
+      "loss": 0.9044,
+      "step": 11886
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.172054045079169e-05,
+      "loss": 0.9114,
+      "step": 11887
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.170522449517472e-05,
+      "loss": 0.8675,
+      "step": 11888
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.168990898356216e-05,
+      "loss": 0.9677,
+      "step": 11889
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.167459391632576e-05,
+      "loss": 0.9518,
+      "step": 11890
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.165927929383722e-05,
+      "loss": 0.8619,
+      "step": 11891
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.164396511646811e-05,
+      "loss": 0.9068,
+      "step": 11892
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.162865138459017e-05,
+      "loss": 0.9653,
+      "step": 11893
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.161333809857508e-05,
+      "loss": 0.8882,
+      "step": 11894
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.15980252587944e-05,
+      "loss": 1.0034,
+      "step": 11895
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.158271286561987e-05,
+      "loss": 0.9341,
+      "step": 11896
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.156740091942304e-05,
+      "loss": 0.8313,
+      "step": 11897
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.155208942057554e-05,
+      "loss": 0.9227,
+      "step": 11898
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.1536778369449e-05,
+      "loss": 0.847,
+      "step": 11899
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.152146776641495e-05,
+      "loss": 0.8428,
+      "step": 11900
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.150615761184505e-05,
+      "loss": 0.8952,
+      "step": 11901
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.14908479061108e-05,
+      "loss": 1.0316,
+      "step": 11902
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.147553864958377e-05,
+      "loss": 0.6812,
+      "step": 11903
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.146022984263556e-05,
+      "loss": 0.9911,
+      "step": 11904
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.144492148563769e-05,
+      "loss": 0.8804,
+      "step": 11905
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.142961357896161e-05,
+      "loss": 0.873,
+      "step": 11906
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.141430612297893e-05,
+      "loss": 0.8726,
+      "step": 11907
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.139899911806108e-05,
+      "loss": 0.8654,
+      "step": 11908
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.138369256457958e-05,
+      "loss": 1.0401,
+      "step": 11909
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.136838646290596e-05,
+      "loss": 0.7718,
+      "step": 11910
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.135308081341162e-05,
+      "loss": 0.9765,
+      "step": 11911
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.133777561646808e-05,
+      "loss": 1.0182,
+      "step": 11912
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.132247087244671e-05,
+      "loss": 1.4908,
+      "step": 11913
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.130716658171901e-05,
+      "loss": 1.235,
+      "step": 11914
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.129186274465641e-05,
+      "loss": 0.9845,
+      "step": 11915
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.12765593616303e-05,
+      "loss": 1.0035,
+      "step": 11916
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.12612564330121e-05,
+      "loss": 1.3543,
+      "step": 11917
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.12459539591732e-05,
+      "loss": 1.2482,
+      "step": 11918
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.123065194048496e-05,
+      "loss": 0.9717,
+      "step": 11919
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.121535037731882e-05,
+      "loss": 0.8448,
+      "step": 11920
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.120004927004607e-05,
+      "loss": 0.9675,
+      "step": 11921
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.118474861903806e-05,
+      "loss": 0.9597,
+      "step": 11922
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.116944842466618e-05,
+      "loss": 0.9318,
+      "step": 11923
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.115414868730172e-05,
+      "loss": 0.9834,
+      "step": 11924
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.113884940731599e-05,
+      "loss": 0.8348,
+      "step": 11925
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.112355058508034e-05,
+      "loss": 1.3945,
+      "step": 11926
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.110825222096599e-05,
+      "loss": 1.2398,
+      "step": 11927
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.109295431534432e-05,
+      "loss": 0.9379,
+      "step": 11928
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.107765686858651e-05,
+      "loss": 0.909,
+      "step": 11929
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.106235988106388e-05,
+      "loss": 0.8664,
+      "step": 11930
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.104706335314765e-05,
+      "loss": 0.8146,
+      "step": 11931
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.103176728520909e-05,
+      "loss": 0.9485,
+      "step": 11932
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.10164716776194e-05,
+      "loss": 0.6752,
+      "step": 11933
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.10011765307498e-05,
+      "loss": 0.9178,
+      "step": 11934
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.098588184497151e-05,
+      "loss": 0.8165,
+      "step": 11935
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.097058762065569e-05,
+      "loss": 0.8298,
+      "step": 11936
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.095529385817358e-05,
+      "loss": 0.9884,
+      "step": 11937
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.094000055789627e-05,
+      "loss": 0.931,
+      "step": 11938
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.0924707720195e-05,
+      "loss": 1.3275,
+      "step": 11939
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.090941534544087e-05,
+      "loss": 1.2439,
+      "step": 11940
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.089412343400502e-05,
+      "loss": 0.8459,
+      "step": 11941
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.087883198625861e-05,
+      "loss": 1.0469,
+      "step": 11942
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.086354100257271e-05,
+      "loss": 0.8297,
+      "step": 11943
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.08482504833185e-05,
+      "loss": 1.4492,
+      "step": 11944
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.083296042886697e-05,
+      "loss": 0.9841,
+      "step": 11945
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.081767083958926e-05,
+      "loss": 0.9,
+      "step": 11946
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.080238171585646e-05,
+      "loss": 0.9771,
+      "step": 11947
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.078709305803958e-05,
+      "loss": 0.8485,
+      "step": 11948
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.077180486650972e-05,
+      "loss": 0.9799,
+      "step": 11949
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.07565171416379e-05,
+      "loss": 0.8406,
+      "step": 11950
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.074122988379508e-05,
+      "loss": 0.8095,
+      "step": 11951
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.072594309335232e-05,
+      "loss": 0.8129,
+      "step": 11952
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.071065677068066e-05,
+      "loss": 0.8895,
+      "step": 11953
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.069537091615105e-05,
+      "loss": 0.7905,
+      "step": 11954
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.068008553013448e-05,
+      "loss": 0.9979,
+      "step": 11955
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.06648006130019e-05,
+      "loss": 1.5147,
+      "step": 11956
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.064951616512428e-05,
+      "loss": 1.2644,
+      "step": 11957
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.063423218687258e-05,
+      "loss": 0.8352,
+      "step": 11958
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.061894867861771e-05,
+      "loss": 0.8777,
+      "step": 11959
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.060366564073063e-05,
+      "loss": 0.7824,
+      "step": 11960
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.05883830735822e-05,
+      "loss": 0.8306,
+      "step": 11961
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.057310097754333e-05,
+      "loss": 1.5857,
+      "step": 11962
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.055781935298496e-05,
+      "loss": 1.3309,
+      "step": 11963
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.054253820027797e-05,
+      "loss": 0.778,
+      "step": 11964
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.052725751979311e-05,
+      "loss": 0.9898,
+      "step": 11965
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.051197731190132e-05,
+      "loss": 0.8006,
+      "step": 11966
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.049669757697346e-05,
+      "loss": 0.791,
+      "step": 11967
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.048141831538029e-05,
+      "loss": 0.9641,
+      "step": 11968
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.046613952749272e-05,
+      "loss": 1.0022,
+      "step": 11969
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.045086121368148e-05,
+      "loss": 0.9267,
+      "step": 11970
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.04355833743174e-05,
+      "loss": 0.8081,
+      "step": 11971
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.042030600977128e-05,
+      "loss": 0.7835,
+      "step": 11972
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.040502912041386e-05,
+      "loss": 0.9623,
+      "step": 11973
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.038975270661593e-05,
+      "loss": 0.8644,
+      "step": 11974
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.03744767687482e-05,
+      "loss": 0.9872,
+      "step": 11975
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.035920130718145e-05,
+      "loss": 0.8664,
+      "step": 11976
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.034392632228641e-05,
+      "loss": 1.0244,
+      "step": 11977
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.032865181443376e-05,
+      "loss": 0.9904,
+      "step": 11978
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.031337778399427e-05,
+      "loss": 1.0824,
+      "step": 11979
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.029810423133856e-05,
+      "loss": 0.9694,
+      "step": 11980
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.028283115683733e-05,
+      "loss": 0.8991,
+      "step": 11981
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.026755856086123e-05,
+      "loss": 0.801,
+      "step": 11982
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.025228644378098e-05,
+      "loss": 0.833,
+      "step": 11983
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.023701480596719e-05,
+      "loss": 1.4183,
+      "step": 11984
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.022174364779049e-05,
+      "loss": 1.6543,
+      "step": 11985
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.020647296962149e-05,
+      "loss": 1.2375,
+      "step": 11986
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.019120277183083e-05,
+      "loss": 0.8764,
+      "step": 11987
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.01759330547891e-05,
+      "loss": 0.8872,
+      "step": 11988
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.016066381886688e-05,
+      "loss": 0.8527,
+      "step": 11989
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.014539506443477e-05,
+      "loss": 0.7975,
+      "step": 11990
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.01301267918633e-05,
+      "loss": 0.9286,
+      "step": 11991
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.011485900152303e-05,
+      "loss": 0.8446,
+      "step": 11992
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.009959169378455e-05,
+      "loss": 0.8019,
+      "step": 11993
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.008432486901835e-05,
+      "loss": 0.998,
+      "step": 11994
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.006905852759493e-05,
+      "loss": 1.0642,
+      "step": 11995
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.005379266988482e-05,
+      "loss": 0.9294,
+      "step": 11996
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.003852729625849e-05,
+      "loss": 0.9606,
+      "step": 11997
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.002326240708645e-05,
+      "loss": 0.8217,
+      "step": 11998
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.000799800273917e-05,
+      "loss": 1.4457,
+      "step": 11999
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.999273408358708e-05,
+      "loss": 1.7253,
+      "step": 12000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.997747065000069e-05,
+      "loss": 1.1683,
+      "step": 12001
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.996220770235035e-05,
+      "loss": 0.8824,
+      "step": 12002
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.994694524100653e-05,
+      "loss": 0.753,
+      "step": 12003
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.993168326633966e-05,
+      "loss": 0.9671,
+      "step": 12004
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.99164217787201e-05,
+      "loss": 0.9414,
+      "step": 12005
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.990116077851829e-05,
+      "loss": 0.8735,
+      "step": 12006
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.988590026610453e-05,
+      "loss": 0.921,
+      "step": 12007
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.987064024184928e-05,
+      "loss": 1.0588,
+      "step": 12008
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.98553807061228e-05,
+      "loss": 0.9497,
+      "step": 12009
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.984012165929548e-05,
+      "loss": 0.9498,
+      "step": 12010
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.982486310173764e-05,
+      "loss": 1.29,
+      "step": 12011
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.98096050338196e-05,
+      "loss": 1.2814,
+      "step": 12012
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.979434745591165e-05,
+      "loss": 1.0573,
+      "step": 12013
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.977909036838409e-05,
+      "loss": 0.8577,
+      "step": 12014
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.976383377160723e-05,
+      "loss": 0.8407,
+      "step": 12015
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.97485776659513e-05,
+      "loss": 0.9494,
+      "step": 12016
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.973332205178658e-05,
+      "loss": 0.9225,
+      "step": 12017
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.97180669294833e-05,
+      "loss": 0.8509,
+      "step": 12018
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.97028122994117e-05,
+      "loss": 0.954,
+      "step": 12019
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.968755816194204e-05,
+      "loss": 0.8931,
+      "step": 12020
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.967230451744446e-05,
+      "loss": 0.8836,
+      "step": 12021
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.96570513662892e-05,
+      "loss": 0.6665,
+      "step": 12022
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.964179870884645e-05,
+      "loss": 0.9303,
+      "step": 12023
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.962654654548639e-05,
+      "loss": 0.9739,
+      "step": 12024
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.961129487657911e-05,
+      "loss": 0.934,
+      "step": 12025
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.959604370249484e-05,
+      "loss": 0.9072,
+      "step": 12026
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.95807930236037e-05,
+      "loss": 0.9555,
+      "step": 12027
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.956554284027578e-05,
+      "loss": 1.0321,
+      "step": 12028
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.955029315288125e-05,
+      "loss": 0.9207,
+      "step": 12029
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.953504396179017e-05,
+      "loss": 0.8722,
+      "step": 12030
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.951979526737265e-05,
+      "loss": 0.8955,
+      "step": 12031
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.950454706999875e-05,
+      "loss": 0.787,
+      "step": 12032
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.948929937003854e-05,
+      "loss": 0.8602,
+      "step": 12033
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.94740521678621e-05,
+      "loss": 0.9093,
+      "step": 12034
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.945880546383944e-05,
+      "loss": 0.7381,
+      "step": 12035
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.944355925834061e-05,
+      "loss": 0.9474,
+      "step": 12036
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.94283135517356e-05,
+      "loss": 0.9352,
+      "step": 12037
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.941306834439447e-05,
+      "loss": 1.3736,
+      "step": 12038
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.939782363668714e-05,
+      "loss": 1.1959,
+      "step": 12039
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.938257942898364e-05,
+      "loss": 0.8727,
+      "step": 12040
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.936733572165391e-05,
+      "loss": 0.8784,
+      "step": 12041
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.935209251506795e-05,
+      "loss": 0.8162,
+      "step": 12042
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.933684980959565e-05,
+      "loss": 0.8848,
+      "step": 12043
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.932160760560696e-05,
+      "loss": 0.8353,
+      "step": 12044
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.930636590347184e-05,
+      "loss": 0.8219,
+      "step": 12045
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.929112470356015e-05,
+      "loss": 0.8741,
+      "step": 12046
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.927588400624181e-05,
+      "loss": 1.4608,
+      "step": 12047
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.926064381188668e-05,
+      "loss": 1.1321,
+      "step": 12048
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.924540412086465e-05,
+      "loss": 0.9246,
+      "step": 12049
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.92301649335456e-05,
+      "loss": 0.8402,
+      "step": 12050
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.921492625029934e-05,
+      "loss": 0.9869,
+      "step": 12051
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.919968807149575e-05,
+      "loss": 0.884,
+      "step": 12052
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.918445039750459e-05,
+      "loss": 0.9723,
+      "step": 12053
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.916921322869569e-05,
+      "loss": 0.8405,
+      "step": 12054
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.915397656543887e-05,
+      "loss": 0.8274,
+      "step": 12055
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.913874040810391e-05,
+      "loss": 1.0017,
+      "step": 12056
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.912350475706057e-05,
+      "loss": 0.8717,
+      "step": 12057
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.910826961267863e-05,
+      "loss": 0.9008,
+      "step": 12058
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.90930349753278e-05,
+      "loss": 0.9786,
+      "step": 12059
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.907780084537788e-05,
+      "loss": 0.8607,
+      "step": 12060
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.906256722319854e-05,
+      "loss": 0.9411,
+      "step": 12061
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.90473341091595e-05,
+      "loss": 0.9599,
+      "step": 12062
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.90321015036305e-05,
+      "loss": 0.8376,
+      "step": 12063
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.901686940698117e-05,
+      "loss": 0.8918,
+      "step": 12064
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.900163781958122e-05,
+      "loss": 0.894,
+      "step": 12065
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.898640674180032e-05,
+      "loss": 0.8879,
+      "step": 12066
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.897117617400813e-05,
+      "loss": 0.8159,
+      "step": 12067
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.895594611657422e-05,
+      "loss": 0.7658,
+      "step": 12068
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.894071656986827e-05,
+      "loss": 0.9964,
+      "step": 12069
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.892548753425986e-05,
+      "loss": 0.806,
+      "step": 12070
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.891025901011863e-05,
+      "loss": 0.9146,
+      "step": 12071
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.889503099781416e-05,
+      "loss": 0.9045,
+      "step": 12072
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.887980349771598e-05,
+      "loss": 0.9015,
+      "step": 12073
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.886457651019373e-05,
+      "loss": 0.9314,
+      "step": 12074
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.88493500356169e-05,
+      "loss": 0.9832,
+      "step": 12075
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.883412407435504e-05,
+      "loss": 0.9137,
+      "step": 12076
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.881889862677769e-05,
+      "loss": 0.7018,
+      "step": 12077
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.880367369325436e-05,
+      "loss": 0.9277,
+      "step": 12078
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.878844927415456e-05,
+      "loss": 1.0626,
+      "step": 12079
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.877322536984774e-05,
+      "loss": 1.5578,
+      "step": 12080
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.875800198070341e-05,
+      "loss": 1.2736,
+      "step": 12081
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.874277910709107e-05,
+      "loss": 0.847,
+      "step": 12082
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.87275567493801e-05,
+      "loss": 0.9086,
+      "step": 12083
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.871233490793995e-05,
+      "loss": 0.9197,
+      "step": 12084
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.86971135831401e-05,
+      "loss": 0.8776,
+      "step": 12085
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.868189277534988e-05,
+      "loss": 1.0533,
+      "step": 12086
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.866667248493873e-05,
+      "loss": 0.8718,
+      "step": 12087
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.865145271227609e-05,
+      "loss": 0.9497,
+      "step": 12088
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.863623345773125e-05,
+      "loss": 0.9924,
+      "step": 12089
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.862101472167365e-05,
+      "loss": 1.4883,
+      "step": 12090
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.860579650447258e-05,
+      "loss": 1.1042,
+      "step": 12091
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.859057880649738e-05,
+      "loss": 0.8771,
+      "step": 12092
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.857536162811744e-05,
+      "loss": 0.8471,
+      "step": 12093
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.856014496970201e-05,
+      "loss": 0.9,
+      "step": 12094
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.854492883162042e-05,
+      "loss": 1.0665,
+      "step": 12095
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.852971321424196e-05,
+      "loss": 0.8313,
+      "step": 12096
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.851449811793588e-05,
+      "loss": 1.3287,
+      "step": 12097
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.849928354307145e-05,
+      "loss": 1.2371,
+      "step": 12098
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.848406949001794e-05,
+      "loss": 1.0006,
+      "step": 12099
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.846885595914454e-05,
+      "loss": 1.0047,
+      "step": 12100
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.845364295082052e-05,
+      "loss": 0.8678,
+      "step": 12101
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.843843046541509e-05,
+      "loss": 0.9088,
+      "step": 12102
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.842321850329741e-05,
+      "loss": 0.9472,
+      "step": 12103
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.840800706483672e-05,
+      "loss": 0.9228,
+      "step": 12104
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.839279615040216e-05,
+      "loss": 0.8311,
+      "step": 12105
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.837758576036289e-05,
+      "loss": 0.8862,
+      "step": 12106
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.836237589508807e-05,
+      "loss": 0.986,
+      "step": 12107
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.834716655494681e-05,
+      "loss": 0.8882,
+      "step": 12108
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.833195774030829e-05,
+      "loss": 0.9483,
+      "step": 12109
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.831674945154154e-05,
+      "loss": 0.9128,
+      "step": 12110
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.830154168901577e-05,
+      "loss": 0.8807,
+      "step": 12111
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.828633445309994e-05,
+      "loss": 0.8801,
+      "step": 12112
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.82711277441632e-05,
+      "loss": 1.0192,
+      "step": 12113
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.825592156257456e-05,
+      "loss": 0.9381,
+      "step": 12114
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.82407159087031e-05,
+      "loss": 0.8199,
+      "step": 12115
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.822551078291784e-05,
+      "loss": 0.8843,
+      "step": 12116
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.821030618558777e-05,
+      "loss": 0.9579,
+      "step": 12117
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.819510211708198e-05,
+      "loss": 0.8361,
+      "step": 12118
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.81798985777694e-05,
+      "loss": 0.9113,
+      "step": 12119
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.816469556801904e-05,
+      "loss": 0.8234,
+      "step": 12120
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.814949308819983e-05,
+      "loss": 0.962,
+      "step": 12121
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.813429113868076e-05,
+      "loss": 0.8321,
+      "step": 12122
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.811908971983078e-05,
+      "loss": 0.8002,
+      "step": 12123
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.810388883201879e-05,
+      "loss": 0.8622,
+      "step": 12124
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.808868847561375e-05,
+      "loss": 0.7434,
+      "step": 12125
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.807348865098455e-05,
+      "loss": 0.8743,
+      "step": 12126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.805828935850003e-05,
+      "loss": 0.9269,
+      "step": 12127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.80430905985291e-05,
+      "loss": 0.955,
+      "step": 12128
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.802789237144066e-05,
+      "loss": 0.8351,
+      "step": 12129
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.801269467760352e-05,
+      "loss": 1.0532,
+      "step": 12130
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.799749751738655e-05,
+      "loss": 0.9458,
+      "step": 12131
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.798230089115856e-05,
+      "loss": 1.1251,
+      "step": 12132
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.796710479928835e-05,
+      "loss": 0.8464,
+      "step": 12133
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.795190924214476e-05,
+      "loss": 0.7722,
+      "step": 12134
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.793671422009654e-05,
+      "loss": 0.9017,
+      "step": 12135
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.792151973351251e-05,
+      "loss": 0.8968,
+      "step": 12136
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.790632578276137e-05,
+      "loss": 0.9341,
+      "step": 12137
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.789113236821191e-05,
+      "loss": 0.9666,
+      "step": 12138
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.787593949023288e-05,
+      "loss": 0.7871,
+      "step": 12139
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.786074714919295e-05,
+      "loss": 0.9675,
+      "step": 12140
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.784555534546091e-05,
+      "loss": 0.8614,
+      "step": 12141
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.783036407940541e-05,
+      "loss": 1.4335,
+      "step": 12142
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.781517335139509e-05,
+      "loss": 1.1325,
+      "step": 12143
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.779998316179866e-05,
+      "loss": 0.9297,
+      "step": 12144
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.778479351098482e-05,
+      "loss": 0.8767,
+      "step": 12145
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.776960439932214e-05,
+      "loss": 0.9264,
+      "step": 12146
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.775441582717932e-05,
+      "loss": 0.9543,
+      "step": 12147
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.773922779492491e-05,
+      "loss": 0.9805,
+      "step": 12148
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.772404030292757e-05,
+      "loss": 0.8217,
+      "step": 12149
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.770885335155589e-05,
+      "loss": 0.9408,
+      "step": 12150
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.76936669411784e-05,
+      "loss": 0.8768,
+      "step": 12151
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.767848107216374e-05,
+      "loss": 0.8722,
+      "step": 12152
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.766329574488039e-05,
+      "loss": 0.8727,
+      "step": 12153
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.764811095969691e-05,
+      "loss": 0.7962,
+      "step": 12154
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.76329267169819e-05,
+      "loss": 0.9911,
+      "step": 12155
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.761774301710379e-05,
+      "loss": 1.0702,
+      "step": 12156
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.760255986043107e-05,
+      "loss": 0.6913,
+      "step": 12157
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.758737724733229e-05,
+      "loss": 0.8733,
+      "step": 12158
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.757219517817586e-05,
+      "loss": 0.95,
+      "step": 12159
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.755701365333029e-05,
+      "loss": 0.8912,
+      "step": 12160
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.754183267316403e-05,
+      "loss": 0.8294,
+      "step": 12161
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.752665223804548e-05,
+      "loss": 1.0238,
+      "step": 12162
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.751147234834307e-05,
+      "loss": 0.9413,
+      "step": 12163
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.749629300442522e-05,
+      "loss": 0.9364,
+      "step": 12164
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.748111420666031e-05,
+      "loss": 0.761,
+      "step": 12165
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.746593595541676e-05,
+      "loss": 1.0098,
+      "step": 12166
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.745075825106289e-05,
+      "loss": 0.9502,
+      "step": 12167
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.743558109396706e-05,
+      "loss": 0.9507,
+      "step": 12168
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.742040448449765e-05,
+      "loss": 0.8671,
+      "step": 12169
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.740522842302298e-05,
+      "loss": 1.0087,
+      "step": 12170
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.739005290991132e-05,
+      "loss": 1.3961,
+      "step": 12171
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.737487794553101e-05,
+      "loss": 1.1875,
+      "step": 12172
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.735970353025032e-05,
+      "loss": 0.9743,
+      "step": 12173
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.734452966443751e-05,
+      "loss": 0.9546,
+      "step": 12174
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.73293563484609e-05,
+      "loss": 1.1738,
+      "step": 12175
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.73141835826887e-05,
+      "loss": 0.9571,
+      "step": 12176
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.729901136748914e-05,
+      "loss": 0.9811,
+      "step": 12177
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.728383970323044e-05,
+      "loss": 0.948,
+      "step": 12178
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.726866859028082e-05,
+      "loss": 1.0708,
+      "step": 12179
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.725349802900847e-05,
+      "loss": 0.932,
+      "step": 12180
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.723832801978158e-05,
+      "loss": 1.0108,
+      "step": 12181
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.722315856296833e-05,
+      "loss": 0.8717,
+      "step": 12182
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.720798965893683e-05,
+      "loss": 0.8889,
+      "step": 12183
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.719282130805525e-05,
+      "loss": 0.8971,
+      "step": 12184
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.717765351069175e-05,
+      "loss": 0.9492,
+      "step": 12185
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.716248626721439e-05,
+      "loss": 1.0028,
+      "step": 12186
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.714731957799128e-05,
+      "loss": 1.3719,
+      "step": 12187
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.713215344339053e-05,
+      "loss": 1.0887,
+      "step": 12188
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.711698786378018e-05,
+      "loss": 1.0132,
+      "step": 12189
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.710182283952833e-05,
+      "loss": 0.9904,
+      "step": 12190
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.708665837100302e-05,
+      "loss": 0.8991,
+      "step": 12191
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.707149445857225e-05,
+      "loss": 0.964,
+      "step": 12192
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.705633110260408e-05,
+      "loss": 0.9105,
+      "step": 12193
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.704116830346648e-05,
+      "loss": 0.8033,
+      "step": 12194
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.702600606152746e-05,
+      "loss": 1.4068,
+      "step": 12195
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.701084437715504e-05,
+      "loss": 1.1052,
+      "step": 12196
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.699568325071712e-05,
+      "loss": 0.7468,
+      "step": 12197
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.698052268258169e-05,
+      "loss": 0.8598,
+      "step": 12198
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.696536267311671e-05,
+      "loss": 1.0634,
+      "step": 12199
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.695020322269003e-05,
+      "loss": 0.8472,
+      "step": 12200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.693504433166961e-05,
+      "loss": 1.3066,
+      "step": 12201
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.691988600042336e-05,
+      "loss": 1.2222,
+      "step": 12202
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.690472822931914e-05,
+      "loss": 0.9236,
+      "step": 12203
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.688957101872483e-05,
+      "loss": 0.8673,
+      "step": 12204
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.68744143690083e-05,
+      "loss": 0.9381,
+      "step": 12205
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.685925828053736e-05,
+      "loss": 0.854,
+      "step": 12206
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.684410275367988e-05,
+      "loss": 0.9463,
+      "step": 12207
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.682894778880365e-05,
+      "loss": 0.9659,
+      "step": 12208
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.68137933862765e-05,
+      "loss": 0.8505,
+      "step": 12209
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.679863954646617e-05,
+      "loss": 0.8736,
+      "step": 12210
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.678348626974047e-05,
+      "loss": 0.8378,
+      "step": 12211
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.676833355646719e-05,
+      "loss": 0.9868,
+      "step": 12212
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.6753181407014e-05,
+      "loss": 0.9328,
+      "step": 12213
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.673802982174876e-05,
+      "loss": 1.4218,
+      "step": 12214
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.672287880103906e-05,
+      "loss": 1.0594,
+      "step": 12215
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.670772834525265e-05,
+      "loss": 0.9505,
+      "step": 12216
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.669257845475724e-05,
+      "loss": 0.9192,
+      "step": 12217
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.667742912992051e-05,
+      "loss": 0.8852,
+      "step": 12218
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.66622803711101e-05,
+      "loss": 0.7667,
+      "step": 12219
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.66471321786937e-05,
+      "loss": 0.9588,
+      "step": 12220
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.663198455303891e-05,
+      "loss": 0.9718,
+      "step": 12221
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.661683749451338e-05,
+      "loss": 0.9625,
+      "step": 12222
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.660169100348473e-05,
+      "loss": 1.4767,
+      "step": 12223
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.658654508032052e-05,
+      "loss": 1.1386,
+      "step": 12224
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.657139972538838e-05,
+      "loss": 0.9792,
+      "step": 12225
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.655625493905583e-05,
+      "loss": 0.8602,
+      "step": 12226
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.654111072169047e-05,
+      "loss": 1.5222,
+      "step": 12227
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.652596707365982e-05,
+      "loss": 0.9885,
+      "step": 12228
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.651082399533145e-05,
+      "loss": 0.903,
+      "step": 12229
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.649568148707278e-05,
+      "loss": 0.8651,
+      "step": 12230
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.648053954925137e-05,
+      "loss": 0.9525,
+      "step": 12231
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.646539818223474e-05,
+      "loss": 0.8358,
+      "step": 12232
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.645025738639029e-05,
+      "loss": 1.0921,
+      "step": 12233
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.643511716208553e-05,
+      "loss": 0.8458,
+      "step": 12234
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.641997750968787e-05,
+      "loss": 0.724,
+      "step": 12235
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.640483842956478e-05,
+      "loss": 0.9385,
+      "step": 12236
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.638969992208367e-05,
+      "loss": 0.8291,
+      "step": 12237
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.637456198761192e-05,
+      "loss": 1.321,
+      "step": 12238
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.635942462651692e-05,
+      "loss": 1.2879,
+      "step": 12239
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.634428783916608e-05,
+      "loss": 0.7684,
+      "step": 12240
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.63291516259267e-05,
+      "loss": 0.867,
+      "step": 12241
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.631401598716622e-05,
+      "loss": 0.9927,
+      "step": 12242
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.629888092325194e-05,
+      "loss": 0.8083,
+      "step": 12243
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.62837464345511e-05,
+      "loss": 0.8665,
+      "step": 12244
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.62686125214311e-05,
+      "loss": 1.2159,
+      "step": 12245
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.625347918425918e-05,
+      "loss": 1.1221,
+      "step": 12246
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.623834642340263e-05,
+      "loss": 0.882,
+      "step": 12247
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.622321423922875e-05,
+      "loss": 0.9429,
+      "step": 12248
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.620808263210473e-05,
+      "loss": 0.7964,
+      "step": 12249
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.619295160239786e-05,
+      "loss": 0.7521,
+      "step": 12250
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.617782115047532e-05,
+      "loss": 0.9205,
+      "step": 12251
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.616269127670434e-05,
+      "loss": 1.0267,
+      "step": 12252
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.614756198145214e-05,
+      "loss": 0.8473,
+      "step": 12253
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.613243326508583e-05,
+      "loss": 0.8733,
+      "step": 12254
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.611730512797266e-05,
+      "loss": 0.9518,
+      "step": 12255
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.610217757047971e-05,
+      "loss": 0.8971,
+      "step": 12256
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.608705059297414e-05,
+      "loss": 0.8874,
+      "step": 12257
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.607192419582311e-05,
+      "loss": 0.878,
+      "step": 12258
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.60567983793937e-05,
+      "loss": 0.9191,
+      "step": 12259
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.604167314405299e-05,
+      "loss": 0.9517,
+      "step": 12260
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.602654849016809e-05,
+      "loss": 0.9945,
+      "step": 12261
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.601142441810601e-05,
+      "loss": 1.0176,
+      "step": 12262
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.599630092823388e-05,
+      "loss": 0.9051,
+      "step": 12263
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.59811780209187e-05,
+      "loss": 0.9643,
+      "step": 12264
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.596605569652749e-05,
+      "loss": 0.9433,
+      "step": 12265
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.595093395542731e-05,
+      "loss": 0.9329,
+      "step": 12266
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.593581279798507e-05,
+      "loss": 1.0522,
+      "step": 12267
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.592069222456782e-05,
+      "loss": 0.8472,
+      "step": 12268
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.590557223554252e-05,
+      "loss": 0.9131,
+      "step": 12269
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.589045283127609e-05,
+      "loss": 0.8226,
+      "step": 12270
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.587533401213552e-05,
+      "loss": 0.8561,
+      "step": 12271
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.586021577848768e-05,
+      "loss": 0.8791,
+      "step": 12272
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.584509813069956e-05,
+      "loss": 0.9433,
+      "step": 12273
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.582998106913797e-05,
+      "loss": 0.7971,
+      "step": 12274
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.581486459416985e-05,
+      "loss": 0.7922,
+      "step": 12275
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.5799748706162e-05,
+      "loss": 0.9306,
+      "step": 12276
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.578463340548138e-05,
+      "loss": 0.9172,
+      "step": 12277
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.576951869249473e-05,
+      "loss": 0.8988,
+      "step": 12278
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.575440456756893e-05,
+      "loss": 0.9803,
+      "step": 12279
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.57392910310708e-05,
+      "loss": 0.856,
+      "step": 12280
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.57241780833671e-05,
+      "loss": 1.0224,
+      "step": 12281
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.570906572482466e-05,
+      "loss": 0.892,
+      "step": 12282
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.569395395581019e-05,
+      "loss": 0.9133,
+      "step": 12283
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.567884277669048e-05,
+      "loss": 0.8435,
+      "step": 12284
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.566373218783228e-05,
+      "loss": 0.9002,
+      "step": 12285
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.564862218960228e-05,
+      "loss": 0.8938,
+      "step": 12286
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.563351278236724e-05,
+      "loss": 0.9711,
+      "step": 12287
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.561840396649386e-05,
+      "loss": 0.9265,
+      "step": 12288
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.560329574234873e-05,
+      "loss": 0.8729,
+      "step": 12289
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.55881881102986e-05,
+      "loss": 0.8841,
+      "step": 12290
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.557308107071011e-05,
+      "loss": 1.403,
+      "step": 12291
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.555797462394989e-05,
+      "loss": 1.2002,
+      "step": 12292
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.554286877038459e-05,
+      "loss": 0.9912,
+      "step": 12293
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.552776351038078e-05,
+      "loss": 1.0515,
+      "step": 12294
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.551265884430509e-05,
+      "loss": 0.8518,
+      "step": 12295
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.549755477252409e-05,
+      "loss": 1.0097,
+      "step": 12296
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.548245129540435e-05,
+      "loss": 0.9142,
+      "step": 12297
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.546734841331242e-05,
+      "loss": 0.8749,
+      "step": 12298
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.545224612661484e-05,
+      "loss": 0.7724,
+      "step": 12299
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.543714443567812e-05,
+      "loss": 0.8783,
+      "step": 12300
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.542204334086881e-05,
+      "loss": 0.8874,
+      "step": 12301
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.540694284255341e-05,
+      "loss": 0.8165,
+      "step": 12302
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.539184294109832e-05,
+      "loss": 0.7486,
+      "step": 12303
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.537674363687006e-05,
+      "loss": 0.966,
+      "step": 12304
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.536164493023511e-05,
+      "loss": 0.7661,
+      "step": 12305
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.534654682155985e-05,
+      "loss": 0.9796,
+      "step": 12306
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.533144931121076e-05,
+      "loss": 0.9518,
+      "step": 12307
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.53163523995542e-05,
+      "loss": 0.884,
+      "step": 12308
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.530125608695656e-05,
+      "loss": 0.7472,
+      "step": 12309
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.528616037378429e-05,
+      "loss": 0.8533,
+      "step": 12310
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.527106526040367e-05,
+      "loss": 0.8397,
+      "step": 12311
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.525597074718111e-05,
+      "loss": 0.9603,
+      "step": 12312
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.524087683448291e-05,
+      "loss": 1.0128,
+      "step": 12313
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.522578352267542e-05,
+      "loss": 1.0294,
+      "step": 12314
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.521069081212493e-05,
+      "loss": 0.869,
+      "step": 12315
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.519559870319772e-05,
+      "loss": 1.0522,
+      "step": 12316
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.518050719626012e-05,
+      "loss": 0.7837,
+      "step": 12317
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.516541629167835e-05,
+      "loss": 0.9254,
+      "step": 12318
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.515032598981863e-05,
+      "loss": 0.8907,
+      "step": 12319
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.513523629104721e-05,
+      "loss": 1.5166,
+      "step": 12320
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.512014719573038e-05,
+      "loss": 1.3018,
+      "step": 12321
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.510505870423423e-05,
+      "loss": 0.8899,
+      "step": 12322
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.508997081692506e-05,
+      "loss": 0.9363,
+      "step": 12323
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.507488353416896e-05,
+      "loss": 0.9425,
+      "step": 12324
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.505979685633213e-05,
+      "loss": 0.8344,
+      "step": 12325
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.504471078378072e-05,
+      "loss": 1.0699,
+      "step": 12326
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.502962531688084e-05,
+      "loss": 0.9171,
+      "step": 12327
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.501454045599862e-05,
+      "loss": 0.8176,
+      "step": 12328
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.499945620150015e-05,
+      "loss": 1.0164,
+      "step": 12329
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.498437255375152e-05,
+      "loss": 0.6848,
+      "step": 12330
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.496928951311885e-05,
+      "loss": 0.8241,
+      "step": 12331
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.495420707996815e-05,
+      "loss": 1.0101,
+      "step": 12332
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.493912525466544e-05,
+      "loss": 0.9415,
+      "step": 12333
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.49240440375768e-05,
+      "loss": 1.4464,
+      "step": 12334
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.490896342906817e-05,
+      "loss": 1.084,
+      "step": 12335
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.489388342950562e-05,
+      "loss": 0.9693,
+      "step": 12336
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.487880403925512e-05,
+      "loss": 0.8906,
+      "step": 12337
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.486372525868261e-05,
+      "loss": 0.8801,
+      "step": 12338
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.48486470881541e-05,
+      "loss": 0.9347,
+      "step": 12339
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.483356952803544e-05,
+      "loss": 0.8974,
+      "step": 12340
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.481849257869263e-05,
+      "loss": 0.8651,
+      "step": 12341
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.480341624049156e-05,
+      "loss": 0.9211,
+      "step": 12342
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.478834051379811e-05,
+      "loss": 0.8097,
+      "step": 12343
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.47732653989782e-05,
+      "loss": 0.8271,
+      "step": 12344
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.475819089639763e-05,
+      "loss": 1.0039,
+      "step": 12345
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.474311700642233e-05,
+      "loss": 0.9425,
+      "step": 12346
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.472804372941804e-05,
+      "loss": 0.9945,
+      "step": 12347
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.471297106575066e-05,
+      "loss": 0.7549,
+      "step": 12348
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.469789901578595e-05,
+      "loss": 0.9897,
+      "step": 12349
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.468282757988974e-05,
+      "loss": 0.8159,
+      "step": 12350
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.466775675842775e-05,
+      "loss": 0.9343,
+      "step": 12351
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.465268655176579e-05,
+      "loss": 0.9067,
+      "step": 12352
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.463761696026959e-05,
+      "loss": 0.9267,
+      "step": 12353
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.462254798430487e-05,
+      "loss": 1.4967,
+      "step": 12354
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.460747962423738e-05,
+      "loss": 1.2836,
+      "step": 12355
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.459241188043276e-05,
+      "loss": 0.9039,
+      "step": 12356
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.457734475325675e-05,
+      "loss": 0.8281,
+      "step": 12357
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.456227824307502e-05,
+      "loss": 1.4028,
+      "step": 12358
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.454721235025319e-05,
+      "loss": 1.2592,
+      "step": 12359
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.453214707515694e-05,
+      "loss": 0.9441,
+      "step": 12360
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.451708241815188e-05,
+      "loss": 0.8989,
+      "step": 12361
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.450201837960362e-05,
+      "loss": 1.3036,
+      "step": 12362
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.448695495987774e-05,
+      "loss": 1.6359,
+      "step": 12363
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.447189215933985e-05,
+      "loss": 1.1794,
+      "step": 12364
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.445682997835549e-05,
+      "loss": 0.9125,
+      "step": 12365
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.444176841729022e-05,
+      "loss": 0.871,
+      "step": 12366
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.44267074765096e-05,
+      "loss": 0.9402,
+      "step": 12367
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.44116471563791e-05,
+      "loss": 0.8635,
+      "step": 12368
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.439658745726428e-05,
+      "loss": 0.8588,
+      "step": 12369
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.438152837953059e-05,
+      "loss": 0.8138,
+      "step": 12370
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.436646992354353e-05,
+      "loss": 0.9729,
+      "step": 12371
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.435141208966858e-05,
+      "loss": 1.4504,
+      "step": 12372
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.433635487827113e-05,
+      "loss": 1.1684,
+      "step": 12373
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.432129828971665e-05,
+      "loss": 0.7068,
+      "step": 12374
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.430624232437053e-05,
+      "loss": 0.804,
+      "step": 12375
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.429118698259824e-05,
+      "loss": 0.8584,
+      "step": 12376
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.427613226476506e-05,
+      "loss": 1.3571,
+      "step": 12377
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.426107817123644e-05,
+      "loss": 1.1221,
+      "step": 12378
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.424602470237767e-05,
+      "loss": 0.8652,
+      "step": 12379
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.423097185855415e-05,
+      "loss": 0.8914,
+      "step": 12380
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.421591964013117e-05,
+      "loss": 1.0068,
+      "step": 12381
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.420086804747403e-05,
+      "loss": 0.8064,
+      "step": 12382
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.418581708094808e-05,
+      "loss": 0.8126,
+      "step": 12383
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.417076674091854e-05,
+      "loss": 0.8792,
+      "step": 12384
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.41557170277507e-05,
+      "loss": 0.9166,
+      "step": 12385
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.41406679418098e-05,
+      "loss": 0.7696,
+      "step": 12386
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.412561948346107e-05,
+      "loss": 0.949,
+      "step": 12387
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.411057165306976e-05,
+      "loss": 0.9382,
+      "step": 12388
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.409552445100103e-05,
+      "loss": 0.9527,
+      "step": 12389
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.408047787762012e-05,
+      "loss": 0.8447,
+      "step": 12390
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.406543193329216e-05,
+      "loss": 0.8861,
+      "step": 12391
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.405038661838228e-05,
+      "loss": 0.8152,
+      "step": 12392
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.403534193325568e-05,
+      "loss": 0.8836,
+      "step": 12393
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.402029787827745e-05,
+      "loss": 0.7325,
+      "step": 12394
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.400525445381272e-05,
+      "loss": 0.9243,
+      "step": 12395
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.399021166022659e-05,
+      "loss": 0.9875,
+      "step": 12396
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.397516949788411e-05,
+      "loss": 1.0751,
+      "step": 12397
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.396012796715038e-05,
+      "loss": 0.9487,
+      "step": 12398
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.394508706839043e-05,
+      "loss": 0.9185,
+      "step": 12399
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.39300468019693e-05,
+      "loss": 1.0729,
+      "step": 12400
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.391500716825203e-05,
+      "loss": 0.8319,
+      "step": 12401
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.389996816760359e-05,
+      "loss": 1.4102,
+      "step": 12402
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.388492980038897e-05,
+      "loss": 1.2623,
+      "step": 12403
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.386989206697317e-05,
+      "loss": 0.9882,
+      "step": 12404
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.385485496772116e-05,
+      "loss": 0.9299,
+      "step": 12405
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.383981850299781e-05,
+      "loss": 0.8208,
+      "step": 12406
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.382478267316813e-05,
+      "loss": 0.9235,
+      "step": 12407
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.380974747859697e-05,
+      "loss": 0.8654,
+      "step": 12408
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.379471291964924e-05,
+      "loss": 0.9529,
+      "step": 12409
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.377967899668985e-05,
+      "loss": 0.9129,
+      "step": 12410
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.376464571008364e-05,
+      "loss": 1.0125,
+      "step": 12411
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.374961306019548e-05,
+      "loss": 0.9199,
+      "step": 12412
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.373458104739017e-05,
+      "loss": 1.0028,
+      "step": 12413
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.371954967203254e-05,
+      "loss": 1.0206,
+      "step": 12414
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.370451893448743e-05,
+      "loss": 0.88,
+      "step": 12415
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.368948883511959e-05,
+      "loss": 0.9167,
+      "step": 12416
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.367445937429381e-05,
+      "loss": 0.871,
+      "step": 12417
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.365943055237483e-05,
+      "loss": 0.9414,
+      "step": 12418
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.36444023697274e-05,
+      "loss": 0.8841,
+      "step": 12419
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.362937482671628e-05,
+      "loss": 0.8889,
+      "step": 12420
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.361434792370614e-05,
+      "loss": 0.8058,
+      "step": 12421
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.359932166106165e-05,
+      "loss": 0.8394,
+      "step": 12422
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.358429603914756e-05,
+      "loss": 0.8604,
+      "step": 12423
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.356927105832847e-05,
+      "loss": 0.879,
+      "step": 12424
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.355424671896905e-05,
+      "loss": 0.9321,
+      "step": 12425
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.353922302143395e-05,
+      "loss": 0.8962,
+      "step": 12426
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.352419996608777e-05,
+      "loss": 0.8818,
+      "step": 12427
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.350917755329512e-05,
+      "loss": 0.8468,
+      "step": 12428
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.349415578342057e-05,
+      "loss": 0.8745,
+      "step": 12429
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.34791346568287e-05,
+      "loss": 1.0239,
+      "step": 12430
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.346411417388409e-05,
+      "loss": 0.9056,
+      "step": 12431
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.344909433495124e-05,
+      "loss": 0.7576,
+      "step": 12432
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.343407514039471e-05,
+      "loss": 0.8016,
+      "step": 12433
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.341905659057895e-05,
+      "loss": 0.8815,
+      "step": 12434
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.340403868586855e-05,
+      "loss": 0.7371,
+      "step": 12435
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.338902142662789e-05,
+      "loss": 0.871,
+      "step": 12436
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.337400481322147e-05,
+      "loss": 0.9284,
+      "step": 12437
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.335898884601371e-05,
+      "loss": 0.8175,
+      "step": 12438
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.334397352536908e-05,
+      "loss": 0.7578,
+      "step": 12439
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.332895885165197e-05,
+      "loss": 1.0883,
+      "step": 12440
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.331394482522677e-05,
+      "loss": 0.9627,
+      "step": 12441
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.329893144645792e-05,
+      "loss": 0.9982,
+      "step": 12442
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.32839187157097e-05,
+      "loss": 0.8145,
+      "step": 12443
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.326890663334651e-05,
+      "loss": 0.8423,
+      "step": 12444
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.325389519973269e-05,
+      "loss": 0.8813,
+      "step": 12445
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.323888441523254e-05,
+      "loss": 1.0688,
+      "step": 12446
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.32238742802104e-05,
+      "loss": 1.3414,
+      "step": 12447
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.32088647950305e-05,
+      "loss": 1.3232,
+      "step": 12448
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.319385596005718e-05,
+      "loss": 0.9103,
+      "step": 12449
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.317884777565462e-05,
+      "loss": 0.9524,
+      "step": 12450
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.316384024218712e-05,
+      "loss": 0.9262,
+      "step": 12451
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.314883336001885e-05,
+      "loss": 0.8985,
+      "step": 12452
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.31338271295141e-05,
+      "loss": 0.9247,
+      "step": 12453
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.311882155103697e-05,
+      "loss": 0.9413,
+      "step": 12454
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.310381662495171e-05,
+      "loss": 0.7457,
+      "step": 12455
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.308881235162246e-05,
+      "loss": 0.9737,
+      "step": 12456
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.307380873141334e-05,
+      "loss": 0.8513,
+      "step": 12457
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.305880576468852e-05,
+      "loss": 0.9178,
+      "step": 12458
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.304380345181208e-05,
+      "loss": 0.8618,
+      "step": 12459
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.302880179314813e-05,
+      "loss": 0.8544,
+      "step": 12460
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.301380078906077e-05,
+      "loss": 0.7791,
+      "step": 12461
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.299880043991404e-05,
+      "loss": 1.4621,
+      "step": 12462
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.298380074607202e-05,
+      "loss": 1.2042,
+      "step": 12463
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.296880170789875e-05,
+      "loss": 0.8648,
+      "step": 12464
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.295380332575817e-05,
+      "loss": 0.9166,
+      "step": 12465
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.293880560001435e-05,
+      "loss": 0.9239,
+      "step": 12466
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.292380853103128e-05,
+      "loss": 0.7545,
+      "step": 12467
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.29088121191729e-05,
+      "loss": 0.8805,
+      "step": 12468
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.289381636480319e-05,
+      "loss": 0.909,
+      "step": 12469
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.287882126828607e-05,
+      "loss": 0.9868,
+      "step": 12470
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.286382682998547e-05,
+      "loss": 0.9106,
+      "step": 12471
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.284883305026533e-05,
+      "loss": 0.8649,
+      "step": 12472
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.283383992948947e-05,
+      "loss": 0.7176,
+      "step": 12473
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.281884746802183e-05,
+      "loss": 1.0401,
+      "step": 12474
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.280385566622623e-05,
+      "loss": 0.8363,
+      "step": 12475
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.278886452446654e-05,
+      "loss": 0.9283,
+      "step": 12476
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.277387404310657e-05,
+      "loss": 0.8477,
+      "step": 12477
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.275888422251012e-05,
+      "loss": 0.91,
+      "step": 12478
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.274389506304105e-05,
+      "loss": 0.9253,
+      "step": 12479
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.272890656506308e-05,
+      "loss": 0.9654,
+      "step": 12480
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.271391872893994e-05,
+      "loss": 0.7691,
+      "step": 12481
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.269893155503542e-05,
+      "loss": 0.8895,
+      "step": 12482
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.268394504371327e-05,
+      "loss": 0.8383,
+      "step": 12483
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.266895919533719e-05,
+      "loss": 0.9632,
+      "step": 12484
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.265397401027086e-05,
+      "loss": 0.9222,
+      "step": 12485
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.263898948887798e-05,
+      "loss": 0.899,
+      "step": 12486
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.262400563152219e-05,
+      "loss": 0.8432,
+      "step": 12487
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.26090224385672e-05,
+      "loss": 0.9422,
+      "step": 12488
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.259403991037656e-05,
+      "loss": 0.7791,
+      "step": 12489
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.257905804731397e-05,
+      "loss": 0.9052,
+      "step": 12490
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.256407684974298e-05,
+      "loss": 0.8055,
+      "step": 12491
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.25490963180272e-05,
+      "loss": 1.0818,
+      "step": 12492
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.253411645253022e-05,
+      "loss": 0.8666,
+      "step": 12493
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.251913725361554e-05,
+      "loss": 0.7947,
+      "step": 12494
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.25041587216467e-05,
+      "loss": 0.9477,
+      "step": 12495
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.248918085698725e-05,
+      "loss": 0.8246,
+      "step": 12496
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.24742036600007e-05,
+      "loss": 0.9299,
+      "step": 12497
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.24592271310505e-05,
+      "loss": 0.8883,
+      "step": 12498
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.244425127050019e-05,
+      "loss": 1.0494,
+      "step": 12499
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.242927607871313e-05,
+      "loss": 0.8959,
+      "step": 12500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.241430155605285e-05,
+      "loss": 0.8913,
+      "step": 12501
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.239932770288273e-05,
+      "loss": 0.8343,
+      "step": 12502
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.238435451956617e-05,
+      "loss": 0.8994,
+      "step": 12503
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.236938200646662e-05,
+      "loss": 0.8712,
+      "step": 12504
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.235441016394737e-05,
+      "loss": 0.7985,
+      "step": 12505
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.23394389923718e-05,
+      "loss": 1.0508,
+      "step": 12506
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.232446849210333e-05,
+      "loss": 0.881,
+      "step": 12507
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.230949866350524e-05,
+      "loss": 0.8869,
+      "step": 12508
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.229452950694077e-05,
+      "loss": 0.9163,
+      "step": 12509
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.227956102277331e-05,
+      "loss": 0.9675,
+      "step": 12510
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.226459321136607e-05,
+      "loss": 0.8971,
+      "step": 12511
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.224962607308234e-05,
+      "loss": 0.9469,
+      "step": 12512
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.223465960828539e-05,
+      "loss": 1.465,
+      "step": 12513
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.22196938173384e-05,
+      "loss": 1.2061,
+      "step": 12514
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.220472870060462e-05,
+      "loss": 0.8627,
+      "step": 12515
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.218976425844719e-05,
+      "loss": 0.9016,
+      "step": 12516
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.217480049122935e-05,
+      "loss": 0.8939,
+      "step": 12517
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.215983739931425e-05,
+      "loss": 0.9322,
+      "step": 12518
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.214487498306501e-05,
+      "loss": 0.8063,
+      "step": 12519
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.212991324284482e-05,
+      "loss": 0.8185,
+      "step": 12520
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.21149521790167e-05,
+      "loss": 0.8463,
+      "step": 12521
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.20999917919438e-05,
+      "loss": 0.9651,
+      "step": 12522
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.208503208198924e-05,
+      "loss": 0.8678,
+      "step": 12523
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.207007304951604e-05,
+      "loss": 1.0321,
+      "step": 12524
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.205511469488721e-05,
+      "loss": 0.8802,
+      "step": 12525
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.204015701846585e-05,
+      "loss": 0.8614,
+      "step": 12526
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.202520002061493e-05,
+      "loss": 0.8333,
+      "step": 12527
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.201024370169745e-05,
+      "loss": 0.9997,
+      "step": 12528
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.199528806207642e-05,
+      "loss": 0.9214,
+      "step": 12529
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.198033310211479e-05,
+      "loss": 0.9204,
+      "step": 12530
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.196537882217554e-05,
+      "loss": 0.8377,
+      "step": 12531
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.195042522262152e-05,
+      "loss": 0.9056,
+      "step": 12532
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.193547230381572e-05,
+      "loss": 0.9543,
+      "step": 12533
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.192052006612102e-05,
+      "loss": 0.9134,
+      "step": 12534
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.19055685099003e-05,
+      "loss": 1.5309,
+      "step": 12535
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.189061763551644e-05,
+      "loss": 1.1552,
+      "step": 12536
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.187566744333229e-05,
+      "loss": 0.6958,
+      "step": 12537
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.186071793371066e-05,
+      "loss": 0.8335,
+      "step": 12538
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.184576910701434e-05,
+      "loss": 0.7887,
+      "step": 12539
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.18308209636062e-05,
+      "loss": 0.9478,
+      "step": 12540
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.181587350384897e-05,
+      "loss": 0.9445,
+      "step": 12541
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.180092672810545e-05,
+      "loss": 0.9635,
+      "step": 12542
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.178598063673836e-05,
+      "loss": 0.9274,
+      "step": 12543
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.177103523011046e-05,
+      "loss": 0.8047,
+      "step": 12544
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.175609050858447e-05,
+      "loss": 0.7522,
+      "step": 12545
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.174114647252305e-05,
+      "loss": 0.7595,
+      "step": 12546
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.172620312228895e-05,
+      "loss": 0.8433,
+      "step": 12547
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.171126045824477e-05,
+      "loss": 0.8131,
+      "step": 12548
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.16963184807532e-05,
+      "loss": 0.8549,
+      "step": 12549
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.168137719017687e-05,
+      "loss": 0.7629,
+      "step": 12550
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.166643658687837e-05,
+      "loss": 0.827,
+      "step": 12551
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.165149667122037e-05,
+      "loss": 0.9077,
+      "step": 12552
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.163655744356538e-05,
+      "loss": 0.8794,
+      "step": 12553
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.162161890427597e-05,
+      "loss": 0.8996,
+      "step": 12554
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.16066810537147e-05,
+      "loss": 0.9142,
+      "step": 12555
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.159174389224416e-05,
+      "loss": 0.7381,
+      "step": 12556
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.157680742022678e-05,
+      "loss": 0.8369,
+      "step": 12557
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.156187163802515e-05,
+      "loss": 0.988,
+      "step": 12558
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.154693654600166e-05,
+      "loss": 0.8369,
+      "step": 12559
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.153200214451885e-05,
+      "loss": 0.9674,
+      "step": 12560
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.151706843393914e-05,
+      "loss": 0.9568,
+      "step": 12561
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.150213541462496e-05,
+      "loss": 0.933,
+      "step": 12562
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.148720308693875e-05,
+      "loss": 0.8777,
+      "step": 12563
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.147227145124287e-05,
+      "loss": 1.5214,
+      "step": 12564
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.145734050789971e-05,
+      "loss": 1.1935,
+      "step": 12565
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.14424102572717e-05,
+      "loss": 1.0358,
+      "step": 12566
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.142748069972114e-05,
+      "loss": 0.9787,
+      "step": 12567
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.141255183561033e-05,
+      "loss": 0.9923,
+      "step": 12568
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.139762366530161e-05,
+      "loss": 0.7952,
+      "step": 12569
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.138269618915731e-05,
+      "loss": 0.8091,
+      "step": 12570
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.136776940753967e-05,
+      "loss": 1.2696,
+      "step": 12571
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.1352843320811e-05,
+      "loss": 1.0222,
+      "step": 12572
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.13379179293335e-05,
+      "loss": 1.1573,
+      "step": 12573
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.132299323346942e-05,
+      "loss": 0.9202,
+      "step": 12574
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.1308069233581e-05,
+      "loss": 0.8012,
+      "step": 12575
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.129314593003039e-05,
+      "loss": 0.8264,
+      "step": 12576
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.127822332317983e-05,
+      "loss": 0.8668,
+      "step": 12577
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.126330141339141e-05,
+      "loss": 0.9456,
+      "step": 12578
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.124838020102734e-05,
+      "loss": 0.9351,
+      "step": 12579
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.123345968644973e-05,
+      "loss": 1.0238,
+      "step": 12580
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.121853987002066e-05,
+      "loss": 0.9317,
+      "step": 12581
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.120362075210232e-05,
+      "loss": 1.4092,
+      "step": 12582
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.118870233305667e-05,
+      "loss": 1.1073,
+      "step": 12583
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.117378461324584e-05,
+      "loss": 1.0756,
+      "step": 12584
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.115886759303185e-05,
+      "loss": 0.93,
+      "step": 12585
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.114395127277676e-05,
+      "loss": 0.916,
+      "step": 12586
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.112903565284252e-05,
+      "loss": 0.8851,
+      "step": 12587
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.11141207335912e-05,
+      "loss": 0.8961,
+      "step": 12588
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.109920651538473e-05,
+      "loss": 0.8289,
+      "step": 12589
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.108429299858508e-05,
+      "loss": 0.9458,
+      "step": 12590
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.106938018355421e-05,
+      "loss": 0.8767,
+      "step": 12591
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.105446807065401e-05,
+      "loss": 0.9174,
+      "step": 12592
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.103955666024644e-05,
+      "loss": 0.9742,
+      "step": 12593
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.102464595269333e-05,
+      "loss": 0.8506,
+      "step": 12594
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.100973594835659e-05,
+      "loss": 0.8414,
+      "step": 12595
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.099482664759812e-05,
+      "loss": 1.3365,
+      "step": 12596
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.097991805077968e-05,
+      "loss": 1.7438,
+      "step": 12597
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.096501015826313e-05,
+      "loss": 1.1661,
+      "step": 12598
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.095010297041028e-05,
+      "loss": 0.8881,
+      "step": 12599
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.09351964875829e-05,
+      "loss": 0.9743,
+      "step": 12600
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.092029071014277e-05,
+      "loss": 0.8422,
+      "step": 12601
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.090538563845169e-05,
+      "loss": 0.7749,
+      "step": 12602
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.089048127287132e-05,
+      "loss": 0.8477,
+      "step": 12603
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.087557761376346e-05,
+      "loss": 1.275,
+      "step": 12604
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.086067466148976e-05,
+      "loss": 1.1758,
+      "step": 12605
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.084577241641191e-05,
+      "loss": 0.7811,
+      "step": 12606
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.083087087889161e-05,
+      "loss": 1.3363,
+      "step": 12607
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.081597004929048e-05,
+      "loss": 1.2175,
+      "step": 12608
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.080106992797018e-05,
+      "loss": 0.8675,
+      "step": 12609
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.078617051529231e-05,
+      "loss": 1.191,
+      "step": 12610
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.077127181161852e-05,
+      "loss": 0.9545,
+      "step": 12611
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.07563738173103e-05,
+      "loss": 0.8027,
+      "step": 12612
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.07414765327293e-05,
+      "loss": 0.862,
+      "step": 12613
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.072657995823702e-05,
+      "loss": 0.7166,
+      "step": 12614
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.071168409419502e-05,
+      "loss": 0.8039,
+      "step": 12615
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.069678894096479e-05,
+      "loss": 1.2144,
+      "step": 12616
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.068189449890783e-05,
+      "loss": 1.2657,
+      "step": 12617
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.066700076838566e-05,
+      "loss": 0.8181,
+      "step": 12618
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.065210774975969e-05,
+      "loss": 0.8138,
+      "step": 12619
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.06372154433914e-05,
+      "loss": 0.9525,
+      "step": 12620
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.062232384964221e-05,
+      "loss": 0.8517,
+      "step": 12621
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.060743296887352e-05,
+      "loss": 0.8573,
+      "step": 12622
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.059254280144677e-05,
+      "loss": 0.9327,
+      "step": 12623
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.057765334772325e-05,
+      "loss": 0.9966,
+      "step": 12624
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.05627646080644e-05,
+      "loss": 1.0032,
+      "step": 12625
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.054787658283155e-05,
+      "loss": 0.7349,
+      "step": 12626
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.053298927238597e-05,
+      "loss": 0.9559,
+      "step": 12627
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.0518102677089e-05,
+      "loss": 0.7734,
+      "step": 12628
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.050321679730195e-05,
+      "loss": 0.8739,
+      "step": 12629
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.048833163338606e-05,
+      "loss": 0.9047,
+      "step": 12630
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.047344718570261e-05,
+      "loss": 0.875,
+      "step": 12631
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.045856345461281e-05,
+      "loss": 0.9624,
+      "step": 12632
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.044368044047789e-05,
+      "loss": 0.8466,
+      "step": 12633
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.042879814365909e-05,
+      "loss": 0.9048,
+      "step": 12634
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.041391656451754e-05,
+      "loss": 0.9578,
+      "step": 12635
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.039903570341446e-05,
+      "loss": 0.8064,
+      "step": 12636
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.038415556071094e-05,
+      "loss": 0.8743,
+      "step": 12637
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.036927613676816e-05,
+      "loss": 0.9278,
+      "step": 12638
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.035439743194724e-05,
+      "loss": 0.886,
+      "step": 12639
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.033951944660928e-05,
+      "loss": 0.9162,
+      "step": 12640
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.032464218111529e-05,
+      "loss": 1.035,
+      "step": 12641
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.030976563582639e-05,
+      "loss": 0.9375,
+      "step": 12642
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.029488981110365e-05,
+      "loss": 0.7931,
+      "step": 12643
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.028001470730804e-05,
+      "loss": 0.8744,
+      "step": 12644
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.026514032480063e-05,
+      "loss": 1.2946,
+      "step": 12645
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.025026666394234e-05,
+      "loss": 1.8,
+      "step": 12646
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.023539372509422e-05,
+      "loss": 1.0663,
+      "step": 12647
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.022052150861719e-05,
+      "loss": 0.9021,
+      "step": 12648
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.020565001487218e-05,
+      "loss": 0.916,
+      "step": 12649
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.019077924422017e-05,
+      "loss": 0.9838,
+      "step": 12650
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.017590919702199e-05,
+      "loss": 0.8653,
+      "step": 12651
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.016103987363857e-05,
+      "loss": 1.3097,
+      "step": 12652
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.01461712744308e-05,
+      "loss": 1.2788,
+      "step": 12653
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.013130339975947e-05,
+      "loss": 0.8425,
+      "step": 12654
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.011643624998552e-05,
+      "loss": 0.7895,
+      "step": 12655
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.010156982546966e-05,
+      "loss": 0.8546,
+      "step": 12656
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.008670412657273e-05,
+      "loss": 0.9569,
+      "step": 12657
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.00718391536555e-05,
+      "loss": 0.8974,
+      "step": 12658
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.005697490707877e-05,
+      "loss": 0.7923,
+      "step": 12659
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.004211138720324e-05,
+      "loss": 0.9463,
+      "step": 12660
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.002724859438969e-05,
+      "loss": 1.0228,
+      "step": 12661
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.00123865289988e-05,
+      "loss": 0.9949,
+      "step": 12662
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.999752519139127e-05,
+      "loss": 0.8497,
+      "step": 12663
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.998266458192779e-05,
+      "loss": 0.9948,
+      "step": 12664
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.996780470096899e-05,
+      "loss": 0.9099,
+      "step": 12665
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.995294554887555e-05,
+      "loss": 0.9055,
+      "step": 12666
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.993808712600805e-05,
+      "loss": 0.968,
+      "step": 12667
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.992322943272712e-05,
+      "loss": 0.8506,
+      "step": 12668
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.990837246939336e-05,
+      "loss": 0.8568,
+      "step": 12669
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.989351623636738e-05,
+      "loss": 0.8424,
+      "step": 12670
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.987866073400962e-05,
+      "loss": 0.9219,
+      "step": 12671
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.98638059626807e-05,
+      "loss": 0.9303,
+      "step": 12672
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.984895192274108e-05,
+      "loss": 0.7816,
+      "step": 12673
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.983409861455131e-05,
+      "loss": 0.9772,
+      "step": 12674
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.981924603847187e-05,
+      "loss": 0.7798,
+      "step": 12675
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.980439419486318e-05,
+      "loss": 0.8909,
+      "step": 12676
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.978954308408574e-05,
+      "loss": 1.3739,
+      "step": 12677
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.977469270649995e-05,
+      "loss": 1.0916,
+      "step": 12678
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.975984306246619e-05,
+      "loss": 0.9534,
+      "step": 12679
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.974499415234493e-05,
+      "loss": 0.909,
+      "step": 12680
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.973014597649649e-05,
+      "loss": 1.3553,
+      "step": 12681
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.971529853528125e-05,
+      "loss": 1.2286,
+      "step": 12682
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.97004518290595e-05,
+      "loss": 0.9471,
+      "step": 12683
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.968560585819168e-05,
+      "loss": 0.8856,
+      "step": 12684
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.967076062303795e-05,
+      "loss": 0.9262,
+      "step": 12685
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.965591612395869e-05,
+      "loss": 1.02,
+      "step": 12686
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.964107236131412e-05,
+      "loss": 0.959,
+      "step": 12687
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.962622933546452e-05,
+      "loss": 0.9172,
+      "step": 12688
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.961138704677009e-05,
+      "loss": 1.1208,
+      "step": 12689
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.959654549559107e-05,
+      "loss": 0.9124,
+      "step": 12690
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.958170468228767e-05,
+      "loss": 0.801,
+      "step": 12691
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.956686460722004e-05,
+      "loss": 0.8752,
+      "step": 12692
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.955202527074836e-05,
+      "loss": 0.8436,
+      "step": 12693
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.953718667323275e-05,
+      "loss": 0.8728,
+      "step": 12694
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.952234881503334e-05,
+      "loss": 0.8164,
+      "step": 12695
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.950751169651029e-05,
+      "loss": 0.8912,
+      "step": 12696
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.949267531802361e-05,
+      "loss": 0.8487,
+      "step": 12697
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.947783967993343e-05,
+      "loss": 0.8522,
+      "step": 12698
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.946300478259977e-05,
+      "loss": 0.941,
+      "step": 12699
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.944817062638267e-05,
+      "loss": 0.9155,
+      "step": 12700
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.943333721164214e-05,
+      "loss": 0.8105,
+      "step": 12701
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.941850453873821e-05,
+      "loss": 0.8126,
+      "step": 12702
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.940367260803082e-05,
+      "loss": 1.0178,
+      "step": 12703
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.938884141987996e-05,
+      "loss": 0.8859,
+      "step": 12704
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.937401097464558e-05,
+      "loss": 0.8926,
+      "step": 12705
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.935918127268759e-05,
+      "loss": 0.8195,
+      "step": 12706
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.93443523143659e-05,
+      "loss": 0.9437,
+      "step": 12707
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.932952410004041e-05,
+      "loss": 1.3814,
+      "step": 12708
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.931469663007098e-05,
+      "loss": 1.1799,
+      "step": 12709
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.929986990481749e-05,
+      "loss": 0.8936,
+      "step": 12710
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.928504392463975e-05,
+      "loss": 0.9043,
+      "step": 12711
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.92702186898976e-05,
+      "loss": 0.8899,
+      "step": 12712
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.925539420095081e-05,
+      "loss": 0.8557,
+      "step": 12713
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.924057045815922e-05,
+      "loss": 0.8639,
+      "step": 12714
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.922574746188252e-05,
+      "loss": 0.9719,
+      "step": 12715
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.921092521248052e-05,
+      "loss": 0.7771,
+      "step": 12716
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.919610371031289e-05,
+      "loss": 0.8963,
+      "step": 12717
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.91812829557394e-05,
+      "loss": 0.8007,
+      "step": 12718
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.916646294911968e-05,
+      "loss": 0.8606,
+      "step": 12719
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.915164369081344e-05,
+      "loss": 0.8318,
+      "step": 12720
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.913682518118037e-05,
+      "loss": 1.3784,
+      "step": 12721
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.912200742058001e-05,
+      "loss": 1.2445,
+      "step": 12722
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.91071904093721e-05,
+      "loss": 0.9481,
+      "step": 12723
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.909237414791614e-05,
+      "loss": 0.8333,
+      "step": 12724
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.907755863657175e-05,
+      "loss": 0.9722,
+      "step": 12725
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.90627438756985e-05,
+      "loss": 0.8281,
+      "step": 12726
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.904792986565594e-05,
+      "loss": 0.8219,
+      "step": 12727
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.90331166068036e-05,
+      "loss": 0.9216,
+      "step": 12728
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.9018304099501e-05,
+      "loss": 0.907,
+      "step": 12729
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.900349234410757e-05,
+      "loss": 0.807,
+      "step": 12730
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.898868134098282e-05,
+      "loss": 0.8926,
+      "step": 12731
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.897387109048625e-05,
+      "loss": 0.8338,
+      "step": 12732
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.895906159297722e-05,
+      "loss": 1.3786,
+      "step": 12733
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 6.89442528488152e-05,
+      "loss": 1.2084,
+      "step": 12734
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.892944485835957e-05,
+      "loss": 0.925,
+      "step": 12735
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.891463762196972e-05,
+      "loss": 0.7953,
+      "step": 12736
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.889983114000504e-05,
+      "loss": 0.8428,
+      "step": 12737
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.88850254128248e-05,
+      "loss": 0.9528,
+      "step": 12738
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.887022044078841e-05,
+      "loss": 0.917,
+      "step": 12739
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.885541622425513e-05,
+      "loss": 0.9699,
+      "step": 12740
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.884061276358426e-05,
+      "loss": 0.9682,
+      "step": 12741
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.882581005913509e-05,
+      "loss": 0.9757,
+      "step": 12742
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.881100811126688e-05,
+      "loss": 0.9954,
+      "step": 12743
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.879620692033881e-05,
+      "loss": 0.9732,
+      "step": 12744
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.878140648671015e-05,
+      "loss": 0.8331,
+      "step": 12745
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.876660681074007e-05,
+      "loss": 0.8681,
+      "step": 12746
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.875180789278775e-05,
+      "loss": 0.8023,
+      "step": 12747
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.873700973321238e-05,
+      "loss": 0.6686,
+      "step": 12748
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.872221233237306e-05,
+      "loss": 0.8175,
+      "step": 12749
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.870741569062898e-05,
+      "loss": 0.8362,
+      "step": 12750
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.869261980833917e-05,
+      "loss": 0.88,
+      "step": 12751
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.867782468586276e-05,
+      "loss": 0.9781,
+      "step": 12752
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.866303032355885e-05,
+      "loss": 0.7631,
+      "step": 12753
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.864823672178642e-05,
+      "loss": 0.9041,
+      "step": 12754
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.863344388090459e-05,
+      "loss": 0.8819,
+      "step": 12755
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.861865180127226e-05,
+      "loss": 0.9368,
+      "step": 12756
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.860386048324852e-05,
+      "loss": 0.8044,
+      "step": 12757
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.858906992719234e-05,
+      "loss": 0.9055,
+      "step": 12758
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.857428013346265e-05,
+      "loss": 0.882,
+      "step": 12759
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.855949110241837e-05,
+      "loss": 0.9801,
+      "step": 12760
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.854470283441846e-05,
+      "loss": 0.8416,
+      "step": 12761
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.85299153298218e-05,
+      "loss": 0.9182,
+      "step": 12762
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.851512858898728e-05,
+      "loss": 0.8318,
+      "step": 12763
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.850034261227381e-05,
+      "loss": 1.0095,
+      "step": 12764
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.848555740004017e-05,
+      "loss": 0.9233,
+      "step": 12765
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.847077295264524e-05,
+      "loss": 1.0277,
+      "step": 12766
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.845598927044778e-05,
+      "loss": 1.337,
+      "step": 12767
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.844120635380665e-05,
+      "loss": 1.2527,
+      "step": 12768
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.842642420308057e-05,
+      "loss": 0.7871,
+      "step": 12769
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.84116428186283e-05,
+      "loss": 0.8751,
+      "step": 12770
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.83968622008086e-05,
+      "loss": 0.946,
+      "step": 12771
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.838208234998016e-05,
+      "loss": 0.9328,
+      "step": 12772
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.836730326650175e-05,
+      "loss": 0.9629,
+      "step": 12773
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.835252495073194e-05,
+      "loss": 1.0045,
+      "step": 12774
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.833774740302947e-05,
+      "loss": 0.8456,
+      "step": 12775
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.832297062375292e-05,
+      "loss": 0.882,
+      "step": 12776
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.8308194613261e-05,
+      "loss": 0.829,
+      "step": 12777
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.829341937191224e-05,
+      "loss": 0.7492,
+      "step": 12778
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.827864490006526e-05,
+      "loss": 0.9445,
+      "step": 12779
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.826387119807864e-05,
+      "loss": 0.9106,
+      "step": 12780
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.82490982663109e-05,
+      "loss": 0.8259,
+      "step": 12781
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.823432610512059e-05,
+      "loss": 0.8936,
+      "step": 12782
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.821955471486624e-05,
+      "loss": 0.8126,
+      "step": 12783
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.82047840959063e-05,
+      "loss": 0.8659,
+      "step": 12784
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.819001424859929e-05,
+      "loss": 0.8741,
+      "step": 12785
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.817524517330362e-05,
+      "loss": 1.4426,
+      "step": 12786
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.816047687037779e-05,
+      "loss": 1.1842,
+      "step": 12787
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.814570934018015e-05,
+      "loss": 1.0051,
+      "step": 12788
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.813094258306915e-05,
+      "loss": 0.8567,
+      "step": 12789
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.811617659940312e-05,
+      "loss": 1.0138,
+      "step": 12790
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.810141138954049e-05,
+      "loss": 0.9171,
+      "step": 12791
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.808664695383952e-05,
+      "loss": 0.9405,
+      "step": 12792
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.807188329265861e-05,
+      "loss": 0.9284,
+      "step": 12793
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.805712040635606e-05,
+      "loss": 0.92,
+      "step": 12794
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.804235829529011e-05,
+      "loss": 0.8574,
+      "step": 12795
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.802759695981908e-05,
+      "loss": 0.907,
+      "step": 12796
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.801283640030116e-05,
+      "loss": 0.7718,
+      "step": 12797
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.799807661709464e-05,
+      "loss": 0.8353,
+      "step": 12798
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.79833176105577e-05,
+      "loss": 0.9478,
+      "step": 12799
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.796855938104855e-05,
+      "loss": 0.9264,
+      "step": 12800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.795380192892536e-05,
+      "loss": 0.8684,
+      "step": 12801
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.793904525454631e-05,
+      "loss": 0.9422,
+      "step": 12802
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.792428935826945e-05,
+      "loss": 0.9515,
+      "step": 12803
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.790953424045296e-05,
+      "loss": 1.0103,
+      "step": 12804
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.789477990145496e-05,
+      "loss": 0.8851,
+      "step": 12805
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.78800263416335e-05,
+      "loss": 0.7941,
+      "step": 12806
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.786527356134665e-05,
+      "loss": 0.9029,
+      "step": 12807
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.785052156095242e-05,
+      "loss": 0.9698,
+      "step": 12808
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.783577034080887e-05,
+      "loss": 1.031,
+      "step": 12809
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.7821019901274e-05,
+      "loss": 0.753,
+      "step": 12810
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.78062702427058e-05,
+      "loss": 0.9052,
+      "step": 12811
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.779152136546223e-05,
+      "loss": 0.9747,
+      "step": 12812
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.77767732699012e-05,
+      "loss": 0.9435,
+      "step": 12813
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.776202595638069e-05,
+      "loss": 1.0015,
+      "step": 12814
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.774727942525861e-05,
+      "loss": 0.9366,
+      "step": 12815
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.77325336768928e-05,
+      "loss": 0.9478,
+      "step": 12816
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.771778871164123e-05,
+      "loss": 0.9832,
+      "step": 12817
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.770304452986164e-05,
+      "loss": 0.8547,
+      "step": 12818
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.768830113191189e-05,
+      "loss": 0.6364,
+      "step": 12819
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.767355851814982e-05,
+      "loss": 0.7936,
+      "step": 12820
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.765881668893325e-05,
+      "loss": 0.9536,
+      "step": 12821
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.76440756446199e-05,
+      "loss": 0.7856,
+      "step": 12822
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.762933538556756e-05,
+      "loss": 0.8376,
+      "step": 12823
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.761459591213395e-05,
+      "loss": 1.0391,
+      "step": 12824
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.75998572246768e-05,
+      "loss": 0.8222,
+      "step": 12825
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.758511932355383e-05,
+      "loss": 1.3924,
+      "step": 12826
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.757038220912269e-05,
+      "loss": 1.1698,
+      "step": 12827
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.755564588174106e-05,
+      "loss": 0.9321,
+      "step": 12828
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.754091034176655e-05,
+      "loss": 0.917,
+      "step": 12829
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.752617558955681e-05,
+      "loss": 0.9441,
+      "step": 12830
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.75114416254695e-05,
+      "loss": 1.0383,
+      "step": 12831
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.749670844986213e-05,
+      "loss": 0.899,
+      "step": 12832
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.748197606309226e-05,
+      "loss": 0.799,
+      "step": 12833
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.746724446551747e-05,
+      "loss": 0.9758,
+      "step": 12834
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.745251365749527e-05,
+      "loss": 0.7092,
+      "step": 12835
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.743778363938317e-05,
+      "loss": 1.0128,
+      "step": 12836
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.74230544115387e-05,
+      "loss": 0.8525,
+      "step": 12837
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.74083259743193e-05,
+      "loss": 0.9169,
+      "step": 12838
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.739359832808242e-05,
+      "loss": 0.9106,
+      "step": 12839
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.737887147318548e-05,
+      "loss": 0.8741,
+      "step": 12840
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.73641454099859e-05,
+      "loss": 0.9655,
+      "step": 12841
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.73494201388411e-05,
+      "loss": 0.9258,
+      "step": 12842
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.733469566010842e-05,
+      "loss": 0.8467,
+      "step": 12843
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.731997197414521e-05,
+      "loss": 0.8546,
+      "step": 12844
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.730524908130888e-05,
+      "loss": 0.773,
+      "step": 12845
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.729052698195668e-05,
+      "loss": 1.3529,
+      "step": 12846
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.72758056764459e-05,
+      "loss": 1.2846,
+      "step": 12847
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.726108516513385e-05,
+      "loss": 0.9839,
+      "step": 12848
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.724636544837776e-05,
+      "loss": 0.7757,
+      "step": 12849
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.723164652653488e-05,
+      "loss": 0.9596,
+      "step": 12850
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.721692839996247e-05,
+      "loss": 0.8598,
+      "step": 12851
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.720221106901768e-05,
+      "loss": 0.9323,
+      "step": 12852
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.71874945340577e-05,
+      "loss": 0.9137,
+      "step": 12853
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.71727787954397e-05,
+      "loss": 0.9323,
+      "step": 12854
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.715806385352084e-05,
+      "loss": 1.009,
+      "step": 12855
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.714334970865824e-05,
+      "loss": 0.9374,
+      "step": 12856
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.712863636120897e-05,
+      "loss": 0.9276,
+      "step": 12857
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.711392381153017e-05,
+      "loss": 0.8888,
+      "step": 12858
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.709921205997883e-05,
+      "loss": 1.0191,
+      "step": 12859
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.708450110691205e-05,
+      "loss": 0.8225,
+      "step": 12860
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.706979095268691e-05,
+      "loss": 0.8975,
+      "step": 12861
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.705508159766032e-05,
+      "loss": 0.9181,
+      "step": 12862
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.704037304218927e-05,
+      "loss": 1.423,
+      "step": 12863
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.70256652866308e-05,
+      "loss": 1.655,
+      "step": 12864
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.701095833134178e-05,
+      "loss": 1.2182,
+      "step": 12865
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.699625217667922e-05,
+      "loss": 1.0117,
+      "step": 12866
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.698154682299997e-05,
+      "loss": 0.7321,
+      "step": 12867
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.696684227066095e-05,
+      "loss": 0.98,
+      "step": 12868
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.695213852001904e-05,
+      "loss": 0.9609,
+      "step": 12869
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.693743557143107e-05,
+      "loss": 0.9497,
+      "step": 12870
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.692273342525386e-05,
+      "loss": 0.8344,
+      "step": 12871
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.690803208184427e-05,
+      "loss": 0.9387,
+      "step": 12872
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.689333154155903e-05,
+      "loss": 0.8844,
+      "step": 12873
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.6878631804755e-05,
+      "loss": 0.8434,
+      "step": 12874
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.686393287178889e-05,
+      "loss": 0.8952,
+      "step": 12875
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.684923474301739e-05,
+      "loss": 0.877,
+      "step": 12876
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.683453741879727e-05,
+      "loss": 0.8062,
+      "step": 12877
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.681984089948521e-05,
+      "loss": 0.6983,
+      "step": 12878
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.680514518543787e-05,
+      "loss": 0.8134,
+      "step": 12879
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.679045027701198e-05,
+      "loss": 0.9451,
+      "step": 12880
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.677575617456408e-05,
+      "loss": 0.7854,
+      "step": 12881
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.676106287845083e-05,
+      "loss": 0.8043,
+      "step": 12882
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.674637038902887e-05,
+      "loss": 0.8942,
+      "step": 12883
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.67316787066547e-05,
+      "loss": 0.982,
+      "step": 12884
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.671698783168494e-05,
+      "loss": 0.7867,
+      "step": 12885
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.67022977644761e-05,
+      "loss": 0.8814,
+      "step": 12886
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.66876085053847e-05,
+      "loss": 0.8968,
+      "step": 12887
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.667292005476729e-05,
+      "loss": 0.8663,
+      "step": 12888
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.665823241298026e-05,
+      "loss": 0.9593,
+      "step": 12889
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.664354558038017e-05,
+      "loss": 0.8514,
+      "step": 12890
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.66288595573234e-05,
+      "loss": 0.7547,
+      "step": 12891
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.661417434416636e-05,
+      "loss": 0.8385,
+      "step": 12892
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.659948994126549e-05,
+      "loss": 1.428,
+      "step": 12893
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.658480634897717e-05,
+      "loss": 1.2375,
+      "step": 12894
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.657012356765773e-05,
+      "loss": 0.8753,
+      "step": 12895
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.655544159766356e-05,
+      "loss": 0.8949,
+      "step": 12896
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.654076043935093e-05,
+      "loss": 0.9491,
+      "step": 12897
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.652608009307618e-05,
+      "loss": 0.7757,
+      "step": 12898
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.651140055919561e-05,
+      "loss": 0.982,
+      "step": 12899
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.649672183806543e-05,
+      "loss": 0.8474,
+      "step": 12900
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.648204393004195e-05,
+      "loss": 0.9494,
+      "step": 12901
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.646736683548134e-05,
+      "loss": 0.8359,
+      "step": 12902
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.645269055473982e-05,
+      "loss": 0.7834,
+      "step": 12903
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.643801508817362e-05,
+      "loss": 0.7278,
+      "step": 12904
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.642334043613887e-05,
+      "loss": 0.9115,
+      "step": 12905
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.640866659899168e-05,
+      "loss": 0.9546,
+      "step": 12906
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.639399357708823e-05,
+      "loss": 0.907,
+      "step": 12907
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.63793213707846e-05,
+      "loss": 0.7911,
+      "step": 12908
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.636464998043687e-05,
+      "loss": 0.8328,
+      "step": 12909
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.634997940640113e-05,
+      "loss": 0.999,
+      "step": 12910
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.63353096490334e-05,
+      "loss": 0.9189,
+      "step": 12911
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.632064070868975e-05,
+      "loss": 0.8025,
+      "step": 12912
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.630597258572616e-05,
+      "loss": 0.9564,
+      "step": 12913
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.62913052804986e-05,
+      "loss": 0.799,
+      "step": 12914
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.627663879336308e-05,
+      "loss": 0.8516,
+      "step": 12915
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.626197312467551e-05,
+      "loss": 0.9334,
+      "step": 12916
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.624730827479182e-05,
+      "loss": 0.8148,
+      "step": 12917
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.623264424406796e-05,
+      "loss": 0.9747,
+      "step": 12918
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.621798103285977e-05,
+      "loss": 1.38,
+      "step": 12919
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.620331864152318e-05,
+      "loss": 1.5979,
+      "step": 12920
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.618865707041397e-05,
+      "loss": 1.153,
+      "step": 12921
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.617399631988797e-05,
+      "loss": 0.9848,
+      "step": 12922
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.615933639030102e-05,
+      "loss": 0.859,
+      "step": 12923
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.614467728200893e-05,
+      "loss": 0.989,
+      "step": 12924
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.613001899536739e-05,
+      "loss": 0.7946,
+      "step": 12925
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.611536153073224e-05,
+      "loss": 0.8552,
+      "step": 12926
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.610070488845915e-05,
+      "loss": 0.7321,
+      "step": 12927
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.608604906890384e-05,
+      "loss": 0.926,
+      "step": 12928
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.607139407242202e-05,
+      "loss": 0.8326,
+      "step": 12929
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.605673989936933e-05,
+      "loss": 1.433,
+      "step": 12930
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.604208655010147e-05,
+      "loss": 1.7035,
+      "step": 12931
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.602743402497398e-05,
+      "loss": 1.3005,
+      "step": 12932
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.601278232434255e-05,
+      "loss": 0.8117,
+      "step": 12933
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.599813144856277e-05,
+      "loss": 0.9576,
+      "step": 12934
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.598348139799015e-05,
+      "loss": 0.9175,
+      "step": 12935
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.596883217298027e-05,
+      "loss": 1.0002,
+      "step": 12936
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.595418377388866e-05,
+      "loss": 0.9528,
+      "step": 12937
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.59395362010708e-05,
+      "loss": 0.8773,
+      "step": 12938
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.592488945488223e-05,
+      "loss": 0.8923,
+      "step": 12939
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.591024353567838e-05,
+      "loss": 0.9504,
+      "step": 12940
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.589559844381472e-05,
+      "loss": 0.8276,
+      "step": 12941
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 6.588095417964668e-05,
+      "loss": 1.0461,
+      "step": 12942
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.586631074352962e-05,
+      "loss": 0.8579,
+      "step": 12943
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.5851668135819e-05,
+      "loss": 0.9515,
+      "step": 12944
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.583702635687016e-05,
+      "loss": 0.9106,
+      "step": 12945
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.582238540703842e-05,
+      "loss": 0.9452,
+      "step": 12946
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.580774528667916e-05,
+      "loss": 1.0146,
+      "step": 12947
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.579310599614763e-05,
+      "loss": 0.9858,
+      "step": 12948
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.57784675357992e-05,
+      "loss": 0.7706,
+      "step": 12949
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.576382990598904e-05,
+      "loss": 1.0316,
+      "step": 12950
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.574919310707246e-05,
+      "loss": 0.9682,
+      "step": 12951
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.573455713940464e-05,
+      "loss": 0.8205,
+      "step": 12952
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.571992200334085e-05,
+      "loss": 0.9527,
+      "step": 12953
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.570528769923623e-05,
+      "loss": 1.4723,
+      "step": 12954
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.569065422744594e-05,
+      "loss": 1.2089,
+      "step": 12955
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.567602158832519e-05,
+      "loss": 0.8427,
+      "step": 12956
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.566138978222903e-05,
+      "loss": 1.0661,
+      "step": 12957
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.564675880951262e-05,
+      "loss": 0.8434,
+      "step": 12958
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.563212867053101e-05,
+      "loss": 0.9213,
+      "step": 12959
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.561749936563929e-05,
+      "loss": 0.9257,
+      "step": 12960
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.56028708951925e-05,
+      "loss": 0.9165,
+      "step": 12961
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.558824325954564e-05,
+      "loss": 1.3744,
+      "step": 12962
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.557361645905377e-05,
+      "loss": 1.1847,
+      "step": 12963
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.555899049407185e-05,
+      "loss": 0.8805,
+      "step": 12964
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.554436536495481e-05,
+      "loss": 0.7659,
+      "step": 12965
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.55297410720576e-05,
+      "loss": 0.8723,
+      "step": 12966
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.551511761573518e-05,
+      "loss": 0.8096,
+      "step": 12967
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.550049499634244e-05,
+      "loss": 1.2785,
+      "step": 12968
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.548587321423428e-05,
+      "loss": 1.0439,
+      "step": 12969
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.547125226976551e-05,
+      "loss": 1.3815,
+      "step": 12970
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.545663216329101e-05,
+      "loss": 1.0548,
+      "step": 12971
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.544201289516562e-05,
+      "loss": 1.2748,
+      "step": 12972
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.542739446574408e-05,
+      "loss": 1.2077,
+      "step": 12973
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.541277687538125e-05,
+      "loss": 0.8545,
+      "step": 12974
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.539816012443183e-05,
+      "loss": 0.9146,
+      "step": 12975
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.538354421325059e-05,
+      "loss": 0.8322,
+      "step": 12976
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.536892914219226e-05,
+      "loss": 0.8533,
+      "step": 12977
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.535431491161153e-05,
+      "loss": 1.3802,
+      "step": 12978
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.533970152186304e-05,
+      "loss": 1.0968,
+      "step": 12979
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.532508897330148e-05,
+      "loss": 0.9761,
+      "step": 12980
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.531047726628152e-05,
+      "loss": 0.9526,
+      "step": 12981
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.529586640115772e-05,
+      "loss": 0.8053,
+      "step": 12982
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.528125637828473e-05,
+      "loss": 0.7027,
+      "step": 12983
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.526664719801708e-05,
+      "loss": 0.9159,
+      "step": 12984
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.525203886070936e-05,
+      "loss": 0.938,
+      "step": 12985
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.523743136671611e-05,
+      "loss": 0.8785,
+      "step": 12986
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.522282471639181e-05,
+      "loss": 0.9707,
+      "step": 12987
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.520821891009101e-05,
+      "loss": 1.0546,
+      "step": 12988
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.519361394816813e-05,
+      "loss": 0.8307,
+      "step": 12989
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.517900983097765e-05,
+      "loss": 0.9424,
+      "step": 12990
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.516440655887402e-05,
+      "loss": 0.7543,
+      "step": 12991
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.514980413221162e-05,
+      "loss": 0.9316,
+      "step": 12992
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.513520255134491e-05,
+      "loss": 0.8398,
+      "step": 12993
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.512060181662818e-05,
+      "loss": 0.9951,
+      "step": 12994
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.51060019284158e-05,
+      "loss": 0.9385,
+      "step": 12995
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.509140288706211e-05,
+      "loss": 0.9281,
+      "step": 12996
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.507680469292144e-05,
+      "loss": 1.4418,
+      "step": 12997
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.506220734634807e-05,
+      "loss": 1.1674,
+      "step": 12998
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.504761084769627e-05,
+      "loss": 0.8109,
+      "step": 12999
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.503301519732026e-05,
+      "loss": 0.8947,
+      "step": 13000
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.501842039557432e-05,
+      "loss": 0.7088,
+      "step": 13001
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.500382644281264e-05,
+      "loss": 1.3595,
+      "step": 13002
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.498923333938938e-05,
+      "loss": 1.1465,
+      "step": 13003
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.497464108565874e-05,
+      "loss": 0.7647,
+      "step": 13004
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.496004968197484e-05,
+      "loss": 0.9824,
+      "step": 13005
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.494545912869183e-05,
+      "loss": 1.0385,
+      "step": 13006
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.493086942616382e-05,
+      "loss": 0.9026,
+      "step": 13007
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.491628057474487e-05,
+      "loss": 0.9059,
+      "step": 13008
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.490169257478906e-05,
+      "loss": 0.8036,
+      "step": 13009
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.488710542665042e-05,
+      "loss": 0.8095,
+      "step": 13010
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.487251913068295e-05,
+      "loss": 0.8026,
+      "step": 13011
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.485793368724069e-05,
+      "loss": 0.807,
+      "step": 13012
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.484334909667762e-05,
+      "loss": 1.4478,
+      "step": 13013
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.482876535934767e-05,
+      "loss": 1.2634,
+      "step": 13014
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.481418247560485e-05,
+      "loss": 0.8561,
+      "step": 13015
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.479960044580298e-05,
+      "loss": 0.8731,
+      "step": 13016
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.478501927029601e-05,
+      "loss": 0.8869,
+      "step": 13017
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.477043894943782e-05,
+      "loss": 0.9083,
+      "step": 13018
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.475585948358227e-05,
+      "loss": 0.995,
+      "step": 13019
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.474128087308318e-05,
+      "loss": 1.0461,
+      "step": 13020
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.472670311829436e-05,
+      "loss": 0.8362,
+      "step": 13021
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.471212621956967e-05,
+      "loss": 0.8589,
+      "step": 13022
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.469755017726278e-05,
+      "loss": 0.7109,
+      "step": 13023
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.46829749917275e-05,
+      "loss": 1.0823,
+      "step": 13024
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.466840066331756e-05,
+      "loss": 0.7021,
+      "step": 13025
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.465382719238666e-05,
+      "loss": 0.9363,
+      "step": 13026
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.463925457928849e-05,
+      "loss": 0.8612,
+      "step": 13027
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.462468282437671e-05,
+      "loss": 0.8749,
+      "step": 13028
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.461011192800501e-05,
+      "loss": 0.833,
+      "step": 13029
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.459554189052699e-05,
+      "loss": 0.8414,
+      "step": 13030
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.458097271229626e-05,
+      "loss": 0.8563,
+      "step": 13031
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.456640439366639e-05,
+      "loss": 0.9672,
+      "step": 13032
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.455183693499096e-05,
+      "loss": 0.8208,
+      "step": 13033
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.453727033662354e-05,
+      "loss": 0.8319,
+      "step": 13034
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.452270459891759e-05,
+      "loss": 0.7927,
+      "step": 13035
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.45081397222267e-05,
+      "loss": 0.9897,
+      "step": 13036
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.449357570690426e-05,
+      "loss": 0.9131,
+      "step": 13037
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.44790125533038e-05,
+      "loss": 0.8558,
+      "step": 13038
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.446445026177872e-05,
+      "loss": 0.8142,
+      "step": 13039
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.444988883268246e-05,
+      "loss": 0.9764,
+      "step": 13040
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.44353282663684e-05,
+      "loss": 0.8798,
+      "step": 13041
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.442076856318992e-05,
+      "loss": 0.8575,
+      "step": 13042
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.44062097235004e-05,
+      "loss": 1.0293,
+      "step": 13043
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.439165174765315e-05,
+      "loss": 0.8546,
+      "step": 13044
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.43770946360015e-05,
+      "loss": 0.9331,
+      "step": 13045
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.436253838889873e-05,
+      "loss": 0.9108,
+      "step": 13046
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.434798300669812e-05,
+      "loss": 0.9204,
+      "step": 13047
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.433342848975294e-05,
+      "loss": 0.7484,
+      "step": 13048
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.431887483841637e-05,
+      "loss": 0.9435,
+      "step": 13049
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.430432205304168e-05,
+      "loss": 0.8842,
+      "step": 13050
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.428977013398201e-05,
+      "loss": 0.7986,
+      "step": 13051
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.427521908159057e-05,
+      "loss": 0.8061,
+      "step": 13052
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.426066889622046e-05,
+      "loss": 0.8405,
+      "step": 13053
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.424611957822484e-05,
+      "loss": 0.9184,
+      "step": 13054
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.423157112795678e-05,
+      "loss": 0.8518,
+      "step": 13055
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.42170235457694e-05,
+      "loss": 1.0091,
+      "step": 13056
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.420247683201575e-05,
+      "loss": 0.969,
+      "step": 13057
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.418793098704885e-05,
+      "loss": 0.8751,
+      "step": 13058
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.417338601122176e-05,
+      "loss": 1.3367,
+      "step": 13059
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.415884190488742e-05,
+      "loss": 1.1993,
+      "step": 13060
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.414429866839888e-05,
+      "loss": 0.9076,
+      "step": 13061
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.412975630210904e-05,
+      "loss": 0.8983,
+      "step": 13062
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.411521480637086e-05,
+      "loss": 0.9656,
+      "step": 13063
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.410067418153727e-05,
+      "loss": 0.9083,
+      "step": 13064
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.408613442796112e-05,
+      "loss": 0.9317,
+      "step": 13065
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.407159554599532e-05,
+      "loss": 0.9166,
+      "step": 13066
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.405705753599273e-05,
+      "loss": 0.8762,
+      "step": 13067
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.404252039830612e-05,
+      "loss": 0.919,
+      "step": 13068
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.402798413328833e-05,
+      "loss": 0.7908,
+      "step": 13069
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.401344874129218e-05,
+      "loss": 0.9313,
+      "step": 13070
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.399891422267039e-05,
+      "loss": 0.8186,
+      "step": 13071
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.398438057777574e-05,
+      "loss": 0.9468,
+      "step": 13072
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.396984780696092e-05,
+      "loss": 0.8699,
+      "step": 13073
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.395531591057863e-05,
+      "loss": 0.7817,
+      "step": 13074
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.394078488898161e-05,
+      "loss": 0.8971,
+      "step": 13075
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.392625474252246e-05,
+      "loss": 0.8904,
+      "step": 13076
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.391172547155385e-05,
+      "loss": 0.8348,
+      "step": 13077
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.389719707642838e-05,
+      "loss": 1.3963,
+      "step": 13078
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.388266955749865e-05,
+      "loss": 1.0688,
+      "step": 13079
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.386814291511726e-05,
+      "loss": 0.9075,
+      "step": 13080
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.385361714963676e-05,
+      "loss": 0.959,
+      "step": 13081
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.383909226140962e-05,
+      "loss": 0.9652,
+      "step": 13082
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.382456825078843e-05,
+      "loss": 0.7957,
+      "step": 13083
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.381004511812563e-05,
+      "loss": 0.9218,
+      "step": 13084
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.379552286377369e-05,
+      "loss": 0.9222,
+      "step": 13085
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.37810014880851e-05,
+      "loss": 0.7563,
+      "step": 13086
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.376648099141225e-05,
+      "loss": 0.9628,
+      "step": 13087
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.375196137410755e-05,
+      "loss": 0.8182,
+      "step": 13088
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.37374426365234e-05,
+      "loss": 0.9354,
+      "step": 13089
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.372292477901211e-05,
+      "loss": 0.8764,
+      "step": 13090
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.37084078019261e-05,
+      "loss": 0.9505,
+      "step": 13091
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.369389170561763e-05,
+      "loss": 0.9574,
+      "step": 13092
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.367937649043904e-05,
+      "loss": 0.7429,
+      "step": 13093
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.366486215674257e-05,
+      "loss": 0.7889,
+      "step": 13094
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.365034870488047e-05,
+      "loss": 0.7636,
+      "step": 13095
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.363583613520506e-05,
+      "loss": 0.9347,
+      "step": 13096
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.362132444806843e-05,
+      "loss": 0.9042,
+      "step": 13097
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.360681364382282e-05,
+      "loss": 1.0069,
+      "step": 13098
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.359230372282044e-05,
+      "loss": 0.8389,
+      "step": 13099
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.357779468541336e-05,
+      "loss": 0.794,
+      "step": 13100
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.356328653195376e-05,
+      "loss": 0.8593,
+      "step": 13101
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.354877926279375e-05,
+      "loss": 0.7976,
+      "step": 13102
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.353427287828539e-05,
+      "loss": 0.9662,
+      "step": 13103
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.351976737878076e-05,
+      "loss": 0.8174,
+      "step": 13104
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.350526276463187e-05,
+      "loss": 0.9855,
+      "step": 13105
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.349075903619077e-05,
+      "loss": 0.904,
+      "step": 13106
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.347625619380945e-05,
+      "loss": 0.8527,
+      "step": 13107
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.346175423783989e-05,
+      "loss": 0.9316,
+      "step": 13108
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.344725316863405e-05,
+      "loss": 0.9075,
+      "step": 13109
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.343275298654383e-05,
+      "loss": 0.8874,
+      "step": 13110
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.34182536919212e-05,
+      "loss": 0.8483,
+      "step": 13111
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.340375528511799e-05,
+      "loss": 0.882,
+      "step": 13112
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.338925776648609e-05,
+      "loss": 0.8057,
+      "step": 13113
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.337476113637735e-05,
+      "loss": 0.8223,
+      "step": 13114
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.33602653951436e-05,
+      "loss": 0.8683,
+      "step": 13115
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.334577054313665e-05,
+      "loss": 0.881,
+      "step": 13116
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.333127658070823e-05,
+      "loss": 0.8553,
+      "step": 13117
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.33167835082102e-05,
+      "loss": 0.901,
+      "step": 13118
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.330229132599418e-05,
+      "loss": 0.8931,
+      "step": 13119
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.328780003441199e-05,
+      "loss": 0.8425,
+      "step": 13120
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.327330963381528e-05,
+      "loss": 0.8337,
+      "step": 13121
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.325882012455573e-05,
+      "loss": 0.8342,
+      "step": 13122
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.324433150698498e-05,
+      "loss": 0.7857,
+      "step": 13123
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.322984378145467e-05,
+      "loss": 1.0389,
+      "step": 13124
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.321535694831646e-05,
+      "loss": 0.7843,
+      "step": 13125
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.320087100792185e-05,
+      "loss": 0.9274,
+      "step": 13126
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.318638596062246e-05,
+      "loss": 0.7998,
+      "step": 13127
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.31719018067698e-05,
+      "loss": 0.9197,
+      "step": 13128
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.315741854671544e-05,
+      "loss": 0.7894,
+      "step": 13129
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.314293618081082e-05,
+      "loss": 0.621,
+      "step": 13130
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.312845470940747e-05,
+      "loss": 0.838,
+      "step": 13131
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.311397413285683e-05,
+      "loss": 0.848,
+      "step": 13132
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.309949445151033e-05,
+      "loss": 1.0369,
+      "step": 13133
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.308501566571941e-05,
+      "loss": 1.0139,
+      "step": 13134
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.307053777583542e-05,
+      "loss": 0.94,
+      "step": 13135
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.305606078220976e-05,
+      "loss": 0.885,
+      "step": 13136
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.304158468519378e-05,
+      "loss": 0.8443,
+      "step": 13137
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.30271094851388e-05,
+      "loss": 1.0276,
+      "step": 13138
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.301263518239614e-05,
+      "loss": 0.8653,
+      "step": 13139
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.299816177731708e-05,
+      "loss": 1.2635,
+      "step": 13140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.298368927025284e-05,
+      "loss": 1.1734,
+      "step": 13141
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.296921766155469e-05,
+      "loss": 0.8082,
+      "step": 13142
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.295474695157386e-05,
+      "loss": 0.8771,
+      "step": 13143
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.294027714066153e-05,
+      "loss": 1.0135,
+      "step": 13144
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.292580822916888e-05,
+      "loss": 1.0379,
+      "step": 13145
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.291134021744705e-05,
+      "loss": 0.9418,
+      "step": 13146
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.28968731058472e-05,
+      "loss": 0.7279,
+      "step": 13147
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.288240689472043e-05,
+      "loss": 0.8109,
+      "step": 13148
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.286794158441778e-05,
+      "loss": 1.0866,
+      "step": 13149
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 6.28534771752904e-05,
+      "loss": 0.8217,
+      "step": 13150
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.283901366768927e-05,
+      "loss": 0.8661,
+      "step": 13151
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.282455106196543e-05,
+      "loss": 0.9131,
+      "step": 13152
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.281008935846989e-05,
+      "loss": 0.8799,
+      "step": 13153
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.27956285575536e-05,
+      "loss": 0.7559,
+      "step": 13154
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.278116865956758e-05,
+      "loss": 0.8809,
+      "step": 13155
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.276670966486268e-05,
+      "loss": 1.4047,
+      "step": 13156
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.275225157378986e-05,
+      "loss": 1.0205,
+      "step": 13157
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.273779438669998e-05,
+      "loss": 0.9445,
+      "step": 13158
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.272333810394395e-05,
+      "loss": 0.937,
+      "step": 13159
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.270888272587259e-05,
+      "loss": 0.8514,
+      "step": 13160
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.269442825283674e-05,
+      "loss": 0.8774,
+      "step": 13161
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.267997468518716e-05,
+      "loss": 0.8618,
+      "step": 13162
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.266552202327467e-05,
+      "loss": 1.5025,
+      "step": 13163
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.265107026745003e-05,
+      "loss": 1.1913,
+      "step": 13164
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.263661941806396e-05,
+      "loss": 0.851,
+      "step": 13165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.26221694754672e-05,
+      "loss": 0.8469,
+      "step": 13166
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.26077204400104e-05,
+      "loss": 0.8717,
+      "step": 13167
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.259327231204425e-05,
+      "loss": 0.883,
+      "step": 13168
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.257882509191944e-05,
+      "loss": 0.9543,
+      "step": 13169
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.256437877998655e-05,
+      "loss": 0.9411,
+      "step": 13170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.254993337659616e-05,
+      "loss": 0.9432,
+      "step": 13171
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.25354888820989e-05,
+      "loss": 0.8515,
+      "step": 13172
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.252104529684532e-05,
+      "loss": 0.9304,
+      "step": 13173
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.250660262118593e-05,
+      "loss": 0.8515,
+      "step": 13174
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.249216085547129e-05,
+      "loss": 0.9281,
+      "step": 13175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.247772000005188e-05,
+      "loss": 0.8512,
+      "step": 13176
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.246328005527815e-05,
+      "loss": 0.8737,
+      "step": 13177
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.244884102150057e-05,
+      "loss": 1.1433,
+      "step": 13178
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.243440289906954e-05,
+      "loss": 0.8298,
+      "step": 13179
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.241996568833553e-05,
+      "loss": 1.446,
+      "step": 13180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.240552938964884e-05,
+      "loss": 1.1769,
+      "step": 13181
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.239109400335988e-05,
+      "loss": 0.8491,
+      "step": 13182
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.237665952981901e-05,
+      "loss": 0.9385,
+      "step": 13183
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.236222596937654e-05,
+      "loss": 0.9391,
+      "step": 13184
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.234779332238268e-05,
+      "loss": 0.9559,
+      "step": 13185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.23333615891878e-05,
+      "loss": 0.8552,
+      "step": 13186
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.231893077014211e-05,
+      "loss": 0.8825,
+      "step": 13187
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.230450086559583e-05,
+      "loss": 0.9067,
+      "step": 13188
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.229007187589921e-05,
+      "loss": 0.7745,
+      "step": 13189
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.227564380140238e-05,
+      "loss": 0.8885,
+      "step": 13190
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.226121664245553e-05,
+      "loss": 0.8325,
+      "step": 13191
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.224679039940879e-05,
+      "loss": 0.9009,
+      "step": 13192
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.223236507261229e-05,
+      "loss": 1.4092,
+      "step": 13193
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.221794066241613e-05,
+      "loss": 1.2299,
+      "step": 13194
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.220351716917035e-05,
+      "loss": 0.9743,
+      "step": 13195
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.218909459322504e-05,
+      "loss": 0.7365,
+      "step": 13196
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.217467293493018e-05,
+      "loss": 0.9333,
+      "step": 13197
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.216025219463581e-05,
+      "loss": 0.8697,
+      "step": 13198
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.214583237269197e-05,
+      "loss": 0.829,
+      "step": 13199
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.213141346944853e-05,
+      "loss": 0.8369,
+      "step": 13200
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.211699548525543e-05,
+      "loss": 0.8442,
+      "step": 13201
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.210257842046263e-05,
+      "loss": 0.7818,
+      "step": 13202
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.208816227542001e-05,
+      "loss": 0.8857,
+      "step": 13203
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.207374705047744e-05,
+      "loss": 1.0233,
+      "step": 13204
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.205933274598477e-05,
+      "loss": 0.7901,
+      "step": 13205
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.204491936229184e-05,
+      "loss": 0.7804,
+      "step": 13206
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.203050689974845e-05,
+      "loss": 0.9877,
+      "step": 13207
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.201609535870437e-05,
+      "loss": 1.3669,
+      "step": 13208
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.200168473950933e-05,
+      "loss": 1.7527,
+      "step": 13209
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.198727504251317e-05,
+      "loss": 1.0352,
+      "step": 13210
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.19728662680655e-05,
+      "loss": 0.9057,
+      "step": 13211
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.195845841651608e-05,
+      "loss": 0.8248,
+      "step": 13212
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.194405148821453e-05,
+      "loss": 0.8312,
+      "step": 13213
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.192964548351058e-05,
+      "loss": 0.8386,
+      "step": 13214
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.191524040275375e-05,
+      "loss": 0.9145,
+      "step": 13215
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.190083624629373e-05,
+      "loss": 0.7536,
+      "step": 13216
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.188643301448002e-05,
+      "loss": 0.9007,
+      "step": 13217
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.187203070766225e-05,
+      "loss": 0.8473,
+      "step": 13218
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.185762932618993e-05,
+      "loss": 1.0018,
+      "step": 13219
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.184322887041256e-05,
+      "loss": 0.8987,
+      "step": 13220
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.182882934067966e-05,
+      "loss": 1.043,
+      "step": 13221
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.181443073734067e-05,
+      "loss": 0.8008,
+      "step": 13222
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.180003306074507e-05,
+      "loss": 1.3854,
+      "step": 13223
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.178563631124224e-05,
+      "loss": 1.2625,
+      "step": 13224
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.177124048918162e-05,
+      "loss": 0.9401,
+      "step": 13225
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.17568455949126e-05,
+      "loss": 0.9468,
+      "step": 13226
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.174245162878447e-05,
+      "loss": 0.781,
+      "step": 13227
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.172805859114665e-05,
+      "loss": 1.0056,
+      "step": 13228
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.171366648234839e-05,
+      "loss": 0.727,
+      "step": 13229
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.169927530273897e-05,
+      "loss": 0.875,
+      "step": 13230
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.16848850526677e-05,
+      "loss": 0.8919,
+      "step": 13231
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.167049573248381e-05,
+      "loss": 0.8138,
+      "step": 13232
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.16561073425365e-05,
+      "loss": 0.9039,
+      "step": 13233
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.1641719883175e-05,
+      "loss": 0.8568,
+      "step": 13234
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.162733335474847e-05,
+      "loss": 0.9579,
+      "step": 13235
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.161294775760604e-05,
+      "loss": 1.4547,
+      "step": 13236
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.15985630920969e-05,
+      "loss": 1.148,
+      "step": 13237
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.158417935857011e-05,
+      "loss": 0.7213,
+      "step": 13238
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.156979655737477e-05,
+      "loss": 1.312,
+      "step": 13239
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.155541468885994e-05,
+      "loss": 1.1284,
+      "step": 13240
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.154103375337465e-05,
+      "loss": 1.0005,
+      "step": 13241
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.152665375126794e-05,
+      "loss": 1.3757,
+      "step": 13242
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.151227468288884e-05,
+      "loss": 1.2781,
+      "step": 13243
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.149789654858622e-05,
+      "loss": 0.939,
+      "step": 13244
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.148351934870908e-05,
+      "loss": 0.9468,
+      "step": 13245
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.14691430836064e-05,
+      "loss": 0.9239,
+      "step": 13246
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.1454767753627e-05,
+      "loss": 0.8905,
+      "step": 13247
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.144039335911982e-05,
+      "loss": 0.9332,
+      "step": 13248
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.142601990043368e-05,
+      "loss": 0.9182,
+      "step": 13249
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.141164737791744e-05,
+      "loss": 0.8919,
+      "step": 13250
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.139727579191992e-05,
+      "loss": 0.8898,
+      "step": 13251
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.138290514278988e-05,
+      "loss": 0.8081,
+      "step": 13252
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.136853543087612e-05,
+      "loss": 0.8278,
+      "step": 13253
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 6.135416665652737e-05,
+      "loss": 0.8843,
+      "step": 13254
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.133979882009233e-05,
+      "loss": 0.8054,
+      "step": 13255
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.132543192191978e-05,
+      "loss": 0.9704,
+      "step": 13256
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.131106596235829e-05,
+      "loss": 0.8454,
+      "step": 13257
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.129670094175661e-05,
+      "loss": 1.0014,
+      "step": 13258
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.128233686046333e-05,
+      "loss": 0.9014,
+      "step": 13259
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.1267973718827e-05,
+      "loss": 0.8853,
+      "step": 13260
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.125361151719627e-05,
+      "loss": 1.3629,
+      "step": 13261
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.123925025591972e-05,
+      "loss": 1.171,
+      "step": 13262
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.122488993534586e-05,
+      "loss": 0.8086,
+      "step": 13263
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.12105305558232e-05,
+      "loss": 0.968,
+      "step": 13264
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.119617211770025e-05,
+      "loss": 1.035,
+      "step": 13265
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.118181462132546e-05,
+      "loss": 0.7958,
+      "step": 13266
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.116745806704734e-05,
+      "loss": 0.8848,
+      "step": 13267
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.115310245521422e-05,
+      "loss": 0.8659,
+      "step": 13268
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.11387477861746e-05,
+      "loss": 0.7865,
+      "step": 13269
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.112439406027679e-05,
+      "loss": 0.86,
+      "step": 13270
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.111004127786917e-05,
+      "loss": 0.753,
+      "step": 13271
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.109568943930011e-05,
+      "loss": 0.915,
+      "step": 13272
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.108133854491788e-05,
+      "loss": 0.6927,
+      "step": 13273
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.106698859507076e-05,
+      "loss": 0.8577,
+      "step": 13274
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.105263959010703e-05,
+      "loss": 0.769,
+      "step": 13275
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.103829153037493e-05,
+      "loss": 0.7919,
+      "step": 13276
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.102394441622269e-05,
+      "loss": 0.734,
+      "step": 13277
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.100959824799852e-05,
+      "loss": 0.9813,
+      "step": 13278
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.099525302605056e-05,
+      "loss": 0.8783,
+      "step": 13279
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0980908750726974e-05,
+      "loss": 0.8438,
+      "step": 13280
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.096656542237589e-05,
+      "loss": 0.8549,
+      "step": 13281
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.095222304134541e-05,
+      "loss": 0.9864,
+      "step": 13282
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.093788160798365e-05,
+      "loss": 0.7477,
+      "step": 13283
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0923541122638606e-05,
+      "loss": 0.8782,
+      "step": 13284
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.090920158565838e-05,
+      "loss": 0.8106,
+      "step": 13285
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.089486299739093e-05,
+      "loss": 0.9758,
+      "step": 13286
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.088052535818432e-05,
+      "loss": 0.6962,
+      "step": 13287
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0866188668386405e-05,
+      "loss": 0.8342,
+      "step": 13288
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.085185292834521e-05,
+      "loss": 0.8754,
+      "step": 13289
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.083751813840863e-05,
+      "loss": 1.0402,
+      "step": 13290
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0823184298924575e-05,
+      "loss": 0.7845,
+      "step": 13291
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.08088514102409e-05,
+      "loss": 0.8687,
+      "step": 13292
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0794519472705467e-05,
+      "loss": 0.8254,
+      "step": 13293
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.078018848666613e-05,
+      "loss": 0.938,
+      "step": 13294
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.076585845247065e-05,
+      "loss": 0.8679,
+      "step": 13295
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.075152937046684e-05,
+      "loss": 0.8242,
+      "step": 13296
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.073720124100244e-05,
+      "loss": 1.5224,
+      "step": 13297
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.072287406442519e-05,
+      "loss": 1.2871,
+      "step": 13298
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0708547841082844e-05,
+      "loss": 0.9348,
+      "step": 13299
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.069422257132301e-05,
+      "loss": 0.9004,
+      "step": 13300
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.067989825549344e-05,
+      "loss": 0.8942,
+      "step": 13301
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0665574893941753e-05,
+      "loss": 0.8849,
+      "step": 13302
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0651252487015507e-05,
+      "loss": 0.8067,
+      "step": 13303
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.063693103506236e-05,
+      "loss": 1.3322,
+      "step": 13304
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.062261053842988e-05,
+      "loss": 1.1509,
+      "step": 13305
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.06082909974656e-05,
+      "loss": 1.0175,
+      "step": 13306
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.059397241251707e-05,
+      "loss": 0.9101,
+      "step": 13307
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.057965478393177e-05,
+      "loss": 0.9295,
+      "step": 13308
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.05653381120572e-05,
+      "loss": 0.8875,
+      "step": 13309
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.055102239724082e-05,
+      "loss": 0.7857,
+      "step": 13310
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.053670763983004e-05,
+      "loss": 1.4036,
+      "step": 13311
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.052239384017231e-05,
+      "loss": 1.0321,
+      "step": 13312
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.050808099861498e-05,
+      "loss": 0.8474,
+      "step": 13313
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.049376911550542e-05,
+      "loss": 0.8453,
+      "step": 13314
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.047945819119101e-05,
+      "loss": 0.8966,
+      "step": 13315
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0465148226019056e-05,
+      "loss": 0.915,
+      "step": 13316
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.045083922033681e-05,
+      "loss": 0.8815,
+      "step": 13317
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.043653117449156e-05,
+      "loss": 0.8709,
+      "step": 13318
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0422224088830606e-05,
+      "loss": 0.8785,
+      "step": 13319
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.040791796370109e-05,
+      "loss": 1.4293,
+      "step": 13320
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.039361279945029e-05,
+      "loss": 1.1826,
+      "step": 13321
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.037930859642533e-05,
+      "loss": 0.8699,
+      "step": 13322
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0365005354973405e-05,
+      "loss": 0.9262,
+      "step": 13323
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.035070307544164e-05,
+      "loss": 0.8694,
+      "step": 13324
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.033640175817711e-05,
+      "loss": 0.9592,
+      "step": 13325
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.032210140352694e-05,
+      "loss": 0.9878,
+      "step": 13326
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0307802011838164e-05,
+      "loss": 0.8292,
+      "step": 13327
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0293503583457845e-05,
+      "loss": 0.76,
+      "step": 13328
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0279206118732986e-05,
+      "loss": 0.8647,
+      "step": 13329
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.026490961801058e-05,
+      "loss": 0.8844,
+      "step": 13330
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.025061408163761e-05,
+      "loss": 0.9249,
+      "step": 13331
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0236319509961e-05,
+      "loss": 0.9863,
+      "step": 13332
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.022202590332766e-05,
+      "loss": 0.9144,
+      "step": 13333
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.020773326208451e-05,
+      "loss": 0.834,
+      "step": 13334
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.019344158657844e-05,
+      "loss": 0.8753,
+      "step": 13335
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0179150877156256e-05,
+      "loss": 0.8265,
+      "step": 13336
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.016486113416483e-05,
+      "loss": 0.9294,
+      "step": 13337
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.015057235795093e-05,
+      "loss": 0.8015,
+      "step": 13338
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0136284548861356e-05,
+      "loss": 0.9518,
+      "step": 13339
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0121997707242875e-05,
+      "loss": 0.8711,
+      "step": 13340
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0107711833442206e-05,
+      "loss": 0.8817,
+      "step": 13341
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.009342692780607e-05,
+      "loss": 0.927,
+      "step": 13342
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0079142990681136e-05,
+      "loss": 0.8157,
+      "step": 13343
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.006486002241408e-05,
+      "loss": 0.7718,
+      "step": 13344
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0050578023351564e-05,
+      "loss": 1.3002,
+      "step": 13345
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0036296993840205e-05,
+      "loss": 1.1535,
+      "step": 13346
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.002201693422652e-05,
+      "loss": 0.9981,
+      "step": 13347
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 6.0007737844857156e-05,
+      "loss": 0.8691,
+      "step": 13348
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.9993459726078616e-05,
+      "loss": 0.8765,
+      "step": 13349
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.997918257823745e-05,
+      "loss": 0.9005,
+      "step": 13350
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.9964906401680156e-05,
+      "loss": 0.9371,
+      "step": 13351
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.995063119675319e-05,
+      "loss": 0.8388,
+      "step": 13352
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.993635696380302e-05,
+      "loss": 0.7914,
+      "step": 13353
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.992208370317607e-05,
+      "loss": 0.8534,
+      "step": 13354
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.9907811415218715e-05,
+      "loss": 0.9214,
+      "step": 13355
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.98935401002774e-05,
+      "loss": 0.9555,
+      "step": 13356
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.987926975869842e-05,
+      "loss": 0.8575,
+      "step": 13357
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.986500039082816e-05,
+      "loss": 0.7869,
+      "step": 13358
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.985073199701289e-05,
+      "loss": 0.7723,
+      "step": 13359
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9836464577598906e-05,
+      "loss": 1.0755,
+      "step": 13360
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.98221981329325e-05,
+      "loss": 0.9896,
+      "step": 13361
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.980793266335989e-05,
+      "loss": 1.1374,
+      "step": 13362
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9793668169227244e-05,
+      "loss": 0.7112,
+      "step": 13363
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9779404650880835e-05,
+      "loss": 0.8347,
+      "step": 13364
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9765142108666775e-05,
+      "loss": 0.9125,
+      "step": 13365
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.975088054293121e-05,
+      "loss": 0.7614,
+      "step": 13366
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9736619954020315e-05,
+      "loss": 0.9362,
+      "step": 13367
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.972236034228011e-05,
+      "loss": 0.9068,
+      "step": 13368
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.970810170805674e-05,
+      "loss": 1.3929,
+      "step": 13369
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9693844051696204e-05,
+      "loss": 1.1876,
+      "step": 13370
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.967958737354453e-05,
+      "loss": 0.8553,
+      "step": 13371
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9665331673947766e-05,
+      "loss": 0.8101,
+      "step": 13372
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.965107695325183e-05,
+      "loss": 0.8892,
+      "step": 13373
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.963682321180274e-05,
+      "loss": 0.7737,
+      "step": 13374
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.96225704499464e-05,
+      "loss": 1.4347,
+      "step": 13375
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9608318668028674e-05,
+      "loss": 1.2455,
+      "step": 13376
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9594067866395486e-05,
+      "loss": 1.2766,
+      "step": 13377
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9579818045392696e-05,
+      "loss": 1.2535,
+      "step": 13378
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.956556920536613e-05,
+      "loss": 0.9446,
+      "step": 13379
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9551321346661594e-05,
+      "loss": 0.8087,
+      "step": 13380
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.953707446962491e-05,
+      "loss": 1.4308,
+      "step": 13381
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.95228285746018e-05,
+      "loss": 1.1417,
+      "step": 13382
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9508583661938035e-05,
+      "loss": 0.8881,
+      "step": 13383
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.94943397319793e-05,
+      "loss": 0.8681,
+      "step": 13384
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.948009678507129e-05,
+      "loss": 1.521,
+      "step": 13385
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.946585482155973e-05,
+      "loss": 1.3138,
+      "step": 13386
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.94516138417902e-05,
+      "loss": 0.9629,
+      "step": 13387
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.943737384610837e-05,
+      "loss": 0.9571,
+      "step": 13388
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.942313483485977e-05,
+      "loss": 0.8105,
+      "step": 13389
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.940889680839008e-05,
+      "loss": 0.9191,
+      "step": 13390
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.939465976704472e-05,
+      "loss": 0.8378,
+      "step": 13391
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.938042371116932e-05,
+      "loss": 0.8847,
+      "step": 13392
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.936618864110932e-05,
+      "loss": 0.8177,
+      "step": 13393
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9351954557210234e-05,
+      "loss": 0.7072,
+      "step": 13394
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9337721459817484e-05,
+      "loss": 0.8423,
+      "step": 13395
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9323489349276515e-05,
+      "loss": 1.0112,
+      "step": 13396
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9309258225932754e-05,
+      "loss": 0.9986,
+      "step": 13397
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9295028090131547e-05,
+      "loss": 0.9185,
+      "step": 13398
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.928079894221827e-05,
+      "loss": 0.7192,
+      "step": 13399
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.926657078253826e-05,
+      "loss": 0.9031,
+      "step": 13400
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.925234361143679e-05,
+      "loss": 0.8482,
+      "step": 13401
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.923811742925922e-05,
+      "loss": 0.9064,
+      "step": 13402
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.922389223635074e-05,
+      "loss": 0.7691,
+      "step": 13403
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.920966803305664e-05,
+      "loss": 1.4358,
+      "step": 13404
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.919544481972213e-05,
+      "loss": 1.1683,
+      "step": 13405
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.918122259669234e-05,
+      "loss": 1.0065,
+      "step": 13406
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.916700136431248e-05,
+      "loss": 0.8371,
+      "step": 13407
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9152781122927705e-05,
+      "loss": 0.8658,
+      "step": 13408
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.913856187288309e-05,
+      "loss": 0.9365,
+      "step": 13409
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.912434361452379e-05,
+      "loss": 0.8159,
+      "step": 13410
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9110126348194793e-05,
+      "loss": 0.8759,
+      "step": 13411
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.909591007424121e-05,
+      "loss": 0.7589,
+      "step": 13412
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.908169479300807e-05,
+      "loss": 0.7832,
+      "step": 13413
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.90674805048403e-05,
+      "loss": 0.8515,
+      "step": 13414
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.905326721008295e-05,
+      "loss": 0.8437,
+      "step": 13415
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.9039054909080905e-05,
+      "loss": 1.3401,
+      "step": 13416
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.902484360217912e-05,
+      "loss": 1.2109,
+      "step": 13417
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.901063328972253e-05,
+      "loss": 0.8205,
+      "step": 13418
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.899642397205599e-05,
+      "loss": 0.8283,
+      "step": 13419
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.89822156495243e-05,
+      "loss": 0.9268,
+      "step": 13420
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8968008322472345e-05,
+      "loss": 0.8485,
+      "step": 13421
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8953801991244895e-05,
+      "loss": 0.8586,
+      "step": 13422
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8939596656186744e-05,
+      "loss": 0.8962,
+      "step": 13423
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.892539231764268e-05,
+      "loss": 0.9404,
+      "step": 13424
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.891118897595738e-05,
+      "loss": 0.9915,
+      "step": 13425
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.88969866314756e-05,
+      "loss": 1.2369,
+      "step": 13426
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8882785284541986e-05,
+      "loss": 1.0755,
+      "step": 13427
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8868584935501206e-05,
+      "loss": 0.9471,
+      "step": 13428
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8854385584697915e-05,
+      "loss": 0.7831,
+      "step": 13429
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.884018723247671e-05,
+      "loss": 0.9879,
+      "step": 13430
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8825989879182176e-05,
+      "loss": 0.8726,
+      "step": 13431
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.881179352515887e-05,
+      "loss": 0.9,
+      "step": 13432
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.879759817075132e-05,
+      "loss": 0.8524,
+      "step": 13433
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8783403816304115e-05,
+      "loss": 0.8651,
+      "step": 13434
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.876921046216166e-05,
+      "loss": 0.7934,
+      "step": 13435
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8755018108668416e-05,
+      "loss": 0.9359,
+      "step": 13436
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.874082675616887e-05,
+      "loss": 0.841,
+      "step": 13437
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.872663640500742e-05,
+      "loss": 0.9131,
+      "step": 13438
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.871244705552845e-05,
+      "loss": 0.8721,
+      "step": 13439
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.869825870807637e-05,
+      "loss": 0.7831,
+      "step": 13440
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.868407136299545e-05,
+      "loss": 0.8337,
+      "step": 13441
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.866988502063009e-05,
+      "loss": 0.8936,
+      "step": 13442
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8655699681324516e-05,
+      "loss": 0.8317,
+      "step": 13443
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8641515345423036e-05,
+      "loss": 0.8044,
+      "step": 13444
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.862733201326991e-05,
+      "loss": 0.7743,
+      "step": 13445
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.86131496852093e-05,
+      "loss": 0.8625,
+      "step": 13446
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.859896836158548e-05,
+      "loss": 0.9684,
+      "step": 13447
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8584788042742566e-05,
+      "loss": 1.0838,
+      "step": 13448
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.857060872902476e-05,
+      "loss": 0.9623,
+      "step": 13449
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.855643042077611e-05,
+      "loss": 0.8741,
+      "step": 13450
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8542253118340784e-05,
+      "loss": 0.9244,
+      "step": 13451
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.85280768220628e-05,
+      "loss": 0.8379,
+      "step": 13452
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.851390153228626e-05,
+      "loss": 0.8807,
+      "step": 13453
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8499727249355175e-05,
+      "loss": 0.8013,
+      "step": 13454
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.848555397361354e-05,
+      "loss": 0.8839,
+      "step": 13455
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.847138170540535e-05,
+      "loss": 0.8821,
+      "step": 13456
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.845721044507453e-05,
+      "loss": 0.8149,
+      "step": 13457
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.844304019296501e-05,
+      "loss": 0.9103,
+      "step": 13458
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.8428870949420735e-05,
+      "loss": 0.9531,
+      "step": 13459
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.841470271478555e-05,
+      "loss": 0.8478,
+      "step": 13460
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.840053548940332e-05,
+      "loss": 0.9698,
+      "step": 13461
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.838636927361787e-05,
+      "loss": 0.8367,
+      "step": 13462
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.837220406777306e-05,
+      "loss": 0.8283,
+      "step": 13463
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.835803987221255e-05,
+      "loss": 0.6706,
+      "step": 13464
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.834387668728023e-05,
+      "loss": 1.0246,
+      "step": 13465
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8329714513319743e-05,
+      "loss": 0.7995,
+      "step": 13466
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8315553350674804e-05,
+      "loss": 0.9457,
+      "step": 13467
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.830139319968917e-05,
+      "loss": 0.7991,
+      "step": 13468
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.828723406070643e-05,
+      "loss": 0.876,
+      "step": 13469
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.827307593407021e-05,
+      "loss": 0.7291,
+      "step": 13470
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8258918820124196e-05,
+      "loss": 0.849,
+      "step": 13471
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.824476271921193e-05,
+      "loss": 0.8489,
+      "step": 13472
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.823060763167696e-05,
+      "loss": 0.9058,
+      "step": 13473
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.82164535578628e-05,
+      "loss": 0.8796,
+      "step": 13474
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8202300498113036e-05,
+      "loss": 0.9843,
+      "step": 13475
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8188148452771105e-05,
+      "loss": 0.9201,
+      "step": 13476
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8173997422180446e-05,
+      "loss": 0.8648,
+      "step": 13477
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.815984740668461e-05,
+      "loss": 0.9343,
+      "step": 13478
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.814569840662685e-05,
+      "loss": 1.0,
+      "step": 13479
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.813155042235068e-05,
+      "loss": 0.945,
+      "step": 13480
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.811740345419942e-05,
+      "loss": 0.7936,
+      "step": 13481
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8103257502516396e-05,
+      "loss": 0.8834,
+      "step": 13482
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.808911256764491e-05,
+      "loss": 1.3969,
+      "step": 13483
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.807496864992831e-05,
+      "loss": 1.1739,
+      "step": 13484
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8060825749709824e-05,
+      "loss": 0.9646,
+      "step": 13485
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.804668386733266e-05,
+      "loss": 0.997,
+      "step": 13486
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8032543003140115e-05,
+      "loss": 0.8898,
+      "step": 13487
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.801840315747533e-05,
+      "loss": 0.9642,
+      "step": 13488
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.8004264330681444e-05,
+      "loss": 0.9783,
+      "step": 13489
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.799012652310167e-05,
+      "loss": 0.9595,
+      "step": 13490
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.79759897350791e-05,
+      "loss": 0.871,
+      "step": 13491
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.796185396695679e-05,
+      "loss": 0.8697,
+      "step": 13492
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.794771921907785e-05,
+      "loss": 0.8894,
+      "step": 13493
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.793358549178529e-05,
+      "loss": 0.8524,
+      "step": 13494
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7919452785422124e-05,
+      "loss": 0.8276,
+      "step": 13495
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.790532110033138e-05,
+      "loss": 0.9284,
+      "step": 13496
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.789119043685602e-05,
+      "loss": 0.9043,
+      "step": 13497
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.787706079533898e-05,
+      "loss": 0.9392,
+      "step": 13498
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.786293217612313e-05,
+      "loss": 0.9725,
+      "step": 13499
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.784880457955146e-05,
+      "loss": 1.0125,
+      "step": 13500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.783467800596678e-05,
+      "loss": 1.524,
+      "step": 13501
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.782055245571191e-05,
+      "loss": 1.1789,
+      "step": 13502
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7806427929129735e-05,
+      "loss": 1.0313,
+      "step": 13503
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7792304426563024e-05,
+      "loss": 0.8562,
+      "step": 13504
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.777818194835452e-05,
+      "loss": 0.9909,
+      "step": 13505
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7764060494847e-05,
+      "loss": 0.9285,
+      "step": 13506
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.774994006638319e-05,
+      "loss": 0.7867,
+      "step": 13507
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.773582066330576e-05,
+      "loss": 0.9102,
+      "step": 13508
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7721702285957393e-05,
+      "loss": 0.9532,
+      "step": 13509
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7707584934680724e-05,
+      "loss": 0.835,
+      "step": 13510
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7693468609818346e-05,
+      "loss": 1.0455,
+      "step": 13511
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.767935331171292e-05,
+      "loss": 0.8407,
+      "step": 13512
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7665239040707e-05,
+      "loss": 0.8919,
+      "step": 13513
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7651125797143116e-05,
+      "loss": 0.8834,
+      "step": 13514
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.763701358136375e-05,
+      "loss": 1.3885,
+      "step": 13515
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7622902393711463e-05,
+      "loss": 1.1475,
+      "step": 13516
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.760879223452872e-05,
+      "loss": 0.9174,
+      "step": 13517
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.75946831041579e-05,
+      "loss": 0.791,
+      "step": 13518
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7580575002941516e-05,
+      "loss": 0.9711,
+      "step": 13519
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.756646793122191e-05,
+      "loss": 0.8625,
+      "step": 13520
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.755236188934143e-05,
+      "loss": 0.7978,
+      "step": 13521
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.753825687764254e-05,
+      "loss": 1.0891,
+      "step": 13522
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.752415289646743e-05,
+      "loss": 0.9168,
+      "step": 13523
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7510049946158405e-05,
+      "loss": 0.8311,
+      "step": 13524
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.749594802705781e-05,
+      "loss": 0.8419,
+      "step": 13525
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.748184713950786e-05,
+      "loss": 0.8564,
+      "step": 13526
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.746774728385072e-05,
+      "loss": 1.0003,
+      "step": 13527
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7453648460428685e-05,
+      "loss": 0.81,
+      "step": 13528
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.743955066958388e-05,
+      "loss": 0.8784,
+      "step": 13529
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.742545391165846e-05,
+      "loss": 0.995,
+      "step": 13530
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.741135818699448e-05,
+      "loss": 0.9415,
+      "step": 13531
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.739726349593413e-05,
+      "loss": 0.8127,
+      "step": 13532
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7383169838819436e-05,
+      "loss": 0.8879,
+      "step": 13533
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7369077215992415e-05,
+      "loss": 0.8311,
+      "step": 13534
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.735498562779515e-05,
+      "loss": 0.818,
+      "step": 13535
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7340895074569635e-05,
+      "loss": 0.8965,
+      "step": 13536
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.732680555665779e-05,
+      "loss": 0.7618,
+      "step": 13537
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7312717074401576e-05,
+      "loss": 0.8131,
+      "step": 13538
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7298629628142944e-05,
+      "loss": 0.9722,
+      "step": 13539
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7284543218223695e-05,
+      "loss": 0.8148,
+      "step": 13540
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.727045784498583e-05,
+      "loss": 0.9049,
+      "step": 13541
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.725637350877111e-05,
+      "loss": 0.8833,
+      "step": 13542
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.724229020992136e-05,
+      "loss": 0.8707,
+      "step": 13543
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.722820794877841e-05,
+      "loss": 0.9325,
+      "step": 13544
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.721412672568403e-05,
+      "loss": 1.339,
+      "step": 13545
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.720004654097993e-05,
+      "loss": 1.0546,
+      "step": 13546
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.718596739500779e-05,
+      "loss": 0.8138,
+      "step": 13547
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.717188928810939e-05,
+      "loss": 0.7494,
+      "step": 13548
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7157812220626383e-05,
+      "loss": 0.8194,
+      "step": 13549
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.714373619290034e-05,
+      "loss": 0.9495,
+      "step": 13550
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7129661205272976e-05,
+      "loss": 0.7873,
+      "step": 13551
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.711558725808583e-05,
+      "loss": 0.9315,
+      "step": 13552
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7101514351680484e-05,
+      "loss": 0.7917,
+      "step": 13553
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.708744248639848e-05,
+      "loss": 1.0356,
+      "step": 13554
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7073371662581307e-05,
+      "loss": 0.8859,
+      "step": 13555
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.705930188057045e-05,
+      "loss": 0.8078,
+      "step": 13556
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.704523314070743e-05,
+      "loss": 0.8237,
+      "step": 13557
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7031165443333665e-05,
+      "loss": 0.8763,
+      "step": 13558
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.7017098788790536e-05,
+      "loss": 0.8852,
+      "step": 13559
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.700303317741949e-05,
+      "loss": 0.8602,
+      "step": 13560
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.6988968609561865e-05,
+      "loss": 0.8455,
+      "step": 13561
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.697490508555896e-05,
+      "loss": 0.8924,
+      "step": 13562
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.696084260575218e-05,
+      "loss": 0.881,
+      "step": 13563
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.6946781170482744e-05,
+      "loss": 0.847,
+      "step": 13564
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.693272078009194e-05,
+      "loss": 0.8053,
+      "step": 13565
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.691866143492101e-05,
+      "loss": 0.8793,
+      "step": 13566
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.690460313531114e-05,
+      "loss": 0.8739,
+      "step": 13567
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6890545881603516e-05,
+      "loss": 0.7961,
+      "step": 13568
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.687648967413935e-05,
+      "loss": 0.8633,
+      "step": 13569
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6862434513259736e-05,
+      "loss": 0.8777,
+      "step": 13570
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.68483803993058e-05,
+      "loss": 0.8911,
+      "step": 13571
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6834327332618596e-05,
+      "loss": 0.9487,
+      "step": 13572
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.682027531353924e-05,
+      "loss": 0.8881,
+      "step": 13573
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.680622434240874e-05,
+      "loss": 0.9483,
+      "step": 13574
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.679217441956808e-05,
+      "loss": 0.9014,
+      "step": 13575
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.677812554535829e-05,
+      "loss": 0.881,
+      "step": 13576
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6764077720120325e-05,
+      "loss": 0.9057,
+      "step": 13577
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6750030944195045e-05,
+      "loss": 0.8812,
+      "step": 13578
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.673598521792347e-05,
+      "loss": 0.7947,
+      "step": 13579
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.672194054164641e-05,
+      "loss": 1.5158,
+      "step": 13580
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.670789691570475e-05,
+      "loss": 1.2913,
+      "step": 13581
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.669385434043931e-05,
+      "loss": 0.9229,
+      "step": 13582
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.667981281619089e-05,
+      "loss": 0.9312,
+      "step": 13583
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6665772343300237e-05,
+      "loss": 0.9204,
+      "step": 13584
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6651732922108194e-05,
+      "loss": 0.8712,
+      "step": 13585
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6637694552955426e-05,
+      "loss": 0.8752,
+      "step": 13586
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6623657236182656e-05,
+      "loss": 0.9681,
+      "step": 13587
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6609620972130514e-05,
+      "loss": 0.9648,
+      "step": 13588
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.659558576113974e-05,
+      "loss": 1.3431,
+      "step": 13589
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.658155160355092e-05,
+      "loss": 1.1758,
+      "step": 13590
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6567518499704606e-05,
+      "loss": 0.9235,
+      "step": 13591
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6553486449941464e-05,
+      "loss": 0.8439,
+      "step": 13592
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6539455454602e-05,
+      "loss": 0.8439,
+      "step": 13593
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6525425514026695e-05,
+      "loss": 0.8456,
+      "step": 13594
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.651139662855614e-05,
+      "loss": 0.9032,
+      "step": 13595
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.64973687985308e-05,
+      "loss": 0.964,
+      "step": 13596
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6483342024291e-05,
+      "loss": 0.9292,
+      "step": 13597
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6469316306177286e-05,
+      "loss": 0.9696,
+      "step": 13598
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6455291644530026e-05,
+      "loss": 0.9095,
+      "step": 13599
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6441268039689535e-05,
+      "loss": 0.8089,
+      "step": 13600
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.642724549199624e-05,
+      "loss": 0.8939,
+      "step": 13601
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6413224001790434e-05,
+      "loss": 0.711,
+      "step": 13602
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6399203569412385e-05,
+      "loss": 0.8949,
+      "step": 13603
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.638518419520235e-05,
+      "loss": 0.8607,
+      "step": 13604
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.637116587950064e-05,
+      "loss": 0.8614,
+      "step": 13605
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.635714862264743e-05,
+      "loss": 0.7158,
+      "step": 13606
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.634313242498286e-05,
+      "loss": 0.9,
+      "step": 13607
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.63291172868472e-05,
+      "loss": 1.3242,
+      "step": 13608
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.631510320858054e-05,
+      "loss": 1.189,
+      "step": 13609
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6301090190523e-05,
+      "loss": 0.9168,
+      "step": 13610
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.628707823301464e-05,
+      "loss": 1.3953,
+      "step": 13611
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.627306733639557e-05,
+      "loss": 1.1462,
+      "step": 13612
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.625905750100575e-05,
+      "loss": 0.7942,
+      "step": 13613
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.624504872718528e-05,
+      "loss": 0.7527,
+      "step": 13614
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.62310410152741e-05,
+      "loss": 0.9065,
+      "step": 13615
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6217034365612156e-05,
+      "loss": 0.8584,
+      "step": 13616
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6203028778539424e-05,
+      "loss": 0.793,
+      "step": 13617
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.618902425439581e-05,
+      "loss": 0.9141,
+      "step": 13618
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6175020793521124e-05,
+      "loss": 0.7798,
+      "step": 13619
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.616101839625532e-05,
+      "loss": 0.8789,
+      "step": 13620
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.614701706293819e-05,
+      "loss": 0.8279,
+      "step": 13621
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6133016793909524e-05,
+      "loss": 1.3336,
+      "step": 13622
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.611901758950908e-05,
+      "loss": 1.1631,
+      "step": 13623
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6105019450076667e-05,
+      "loss": 0.8995,
+      "step": 13624
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6091022375952005e-05,
+      "loss": 0.7697,
+      "step": 13625
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.607702636747477e-05,
+      "loss": 0.9752,
+      "step": 13626
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6063031424984646e-05,
+      "loss": 0.9884,
+      "step": 13627
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.604903754882128e-05,
+      "loss": 0.9935,
+      "step": 13628
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.603504473932426e-05,
+      "loss": 0.7662,
+      "step": 13629
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6021052996833265e-05,
+      "loss": 0.8804,
+      "step": 13630
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.6007062321687817e-05,
+      "loss": 0.8775,
+      "step": 13631
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5993072714227445e-05,
+      "loss": 0.9789,
+      "step": 13632
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.597908417479172e-05,
+      "loss": 0.9056,
+      "step": 13633
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.596509670372012e-05,
+      "loss": 0.7678,
+      "step": 13634
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5951110301352064e-05,
+      "loss": 0.8634,
+      "step": 13635
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5937124968027066e-05,
+      "loss": 0.9931,
+      "step": 13636
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.592314070408451e-05,
+      "loss": 0.8651,
+      "step": 13637
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.59091575098638e-05,
+      "loss": 0.8432,
+      "step": 13638
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5895175385704236e-05,
+      "loss": 0.7413,
+      "step": 13639
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.58811943319453e-05,
+      "loss": 0.8413,
+      "step": 13640
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.586721434892612e-05,
+      "loss": 1.0063,
+      "step": 13641
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.585323543698612e-05,
+      "loss": 0.9362,
+      "step": 13642
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.583925759646451e-05,
+      "loss": 0.9754,
+      "step": 13643
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.582528082770055e-05,
+      "loss": 0.8947,
+      "step": 13644
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.581130513103339e-05,
+      "loss": 1.0583,
+      "step": 13645
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.579733050680227e-05,
+      "loss": 0.891,
+      "step": 13646
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.578335695534634e-05,
+      "loss": 0.735,
+      "step": 13647
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.576938447700469e-05,
+      "loss": 0.8866,
+      "step": 13648
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5755413072116494e-05,
+      "loss": 0.9357,
+      "step": 13649
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.574144274102078e-05,
+      "loss": 0.835,
+      "step": 13650
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.572747348405658e-05,
+      "loss": 0.9728,
+      "step": 13651
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.571350530156298e-05,
+      "loss": 0.9804,
+      "step": 13652
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.569953819387897e-05,
+      "loss": 0.8039,
+      "step": 13653
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.568557216134351e-05,
+      "loss": 0.8094,
+      "step": 13654
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.567160720429554e-05,
+      "loss": 0.8904,
+      "step": 13655
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.565764332307397e-05,
+      "loss": 0.852,
+      "step": 13656
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.564368051801769e-05,
+      "loss": 0.9172,
+      "step": 13657
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.562971878946564e-05,
+      "loss": 0.9266,
+      "step": 13658
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.561575813775659e-05,
+      "loss": 0.9077,
+      "step": 13659
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.56017985632294e-05,
+      "loss": 0.9165,
+      "step": 13660
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.558784006622282e-05,
+      "loss": 0.8093,
+      "step": 13661
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.557388264707566e-05,
+      "loss": 0.9625,
+      "step": 13662
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.555992630612665e-05,
+      "loss": 0.9222,
+      "step": 13663
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.554597104371444e-05,
+      "loss": 0.8317,
+      "step": 13664
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5532016860177814e-05,
+      "loss": 0.9722,
+      "step": 13665
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.55180637558554e-05,
+      "loss": 0.9641,
+      "step": 13666
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5504111731085785e-05,
+      "loss": 0.9369,
+      "step": 13667
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.549016078620764e-05,
+      "loss": 0.7828,
+      "step": 13668
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5476210921559544e-05,
+      "loss": 0.9118,
+      "step": 13669
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.546226213747998e-05,
+      "loss": 0.9788,
+      "step": 13670
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5448314434307545e-05,
+      "loss": 0.871,
+      "step": 13671
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.543436781238074e-05,
+      "loss": 0.9512,
+      "step": 13672
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.542042227203796e-05,
+      "loss": 0.8877,
+      "step": 13673
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5406477813617785e-05,
+      "loss": 0.7812,
+      "step": 13674
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.539253443745858e-05,
+      "loss": 0.9209,
+      "step": 13675
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.537859214389871e-05,
+      "loss": 0.9113,
+      "step": 13676
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.536465093327656e-05,
+      "loss": 0.8184,
+      "step": 13677
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.535071080593053e-05,
+      "loss": 0.9525,
+      "step": 13678
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5336771762198894e-05,
+      "loss": 1.3351,
+      "step": 13679
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5322833802419914e-05,
+      "loss": 1.1834,
+      "step": 13680
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.530889692693193e-05,
+      "loss": 0.8793,
+      "step": 13681
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.529496113607315e-05,
+      "loss": 1.0276,
+      "step": 13682
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.528102643018176e-05,
+      "loss": 0.8456,
+      "step": 13683
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5267092809596046e-05,
+      "loss": 0.9858,
+      "step": 13684
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.525316027465406e-05,
+      "loss": 1.0041,
+      "step": 13685
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5239228825693925e-05,
+      "loss": 0.9111,
+      "step": 13686
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.522529846305384e-05,
+      "loss": 0.8461,
+      "step": 13687
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.521136918707186e-05,
+      "loss": 0.7274,
+      "step": 13688
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.519744099808598e-05,
+      "loss": 0.8912,
+      "step": 13689
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.518351389643433e-05,
+      "loss": 0.9137,
+      "step": 13690
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5169587882454834e-05,
+      "loss": 0.9728,
+      "step": 13691
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.515566295648547e-05,
+      "loss": 0.8326,
+      "step": 13692
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5141739118864256e-05,
+      "loss": 0.8738,
+      "step": 13693
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.512781636992908e-05,
+      "loss": 1.0331,
+      "step": 13694
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5113894710017824e-05,
+      "loss": 0.5898,
+      "step": 13695
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5099974139468325e-05,
+      "loss": 0.8486,
+      "step": 13696
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5086054658618534e-05,
+      "loss": 0.8605,
+      "step": 13697
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5072136267806186e-05,
+      "loss": 0.8162,
+      "step": 13698
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.505821896736911e-05,
+      "loss": 0.8473,
+      "step": 13699
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.504430275764505e-05,
+      "loss": 0.7924,
+      "step": 13700
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.503038763897175e-05,
+      "loss": 1.3968,
+      "step": 13701
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.501647361168688e-05,
+      "loss": 1.0966,
+      "step": 13702
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.500256067612821e-05,
+      "loss": 0.924,
+      "step": 13703
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.498864883263337e-05,
+      "loss": 1.4091,
+      "step": 13704
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.497473808153992e-05,
+      "loss": 1.6691,
+      "step": 13705
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4960828423185596e-05,
+      "loss": 1.0615,
+      "step": 13706
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4946919857907894e-05,
+      "loss": 0.9767,
+      "step": 13707
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.493301238604435e-05,
+      "loss": 0.8154,
+      "step": 13708
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4919106007932556e-05,
+      "loss": 0.7268,
+      "step": 13709
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.490520072390999e-05,
+      "loss": 0.8822,
+      "step": 13710
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.489129653431412e-05,
+      "loss": 0.7286,
+      "step": 13711
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.487739343948234e-05,
+      "loss": 1.0302,
+      "step": 13712
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.48634914397522e-05,
+      "loss": 1.3267,
+      "step": 13713
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.484959053546095e-05,
+      "loss": 1.0982,
+      "step": 13714
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.483569072694605e-05,
+      "loss": 0.8489,
+      "step": 13715
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4821792014544825e-05,
+      "loss": 0.9807,
+      "step": 13716
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.480789439859457e-05,
+      "loss": 0.8757,
+      "step": 13717
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4793997879432536e-05,
+      "loss": 0.729,
+      "step": 13718
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.478010245739608e-05,
+      "loss": 0.9291,
+      "step": 13719
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4766208132822375e-05,
+      "loss": 0.8447,
+      "step": 13720
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.47523149060486e-05,
+      "loss": 0.9848,
+      "step": 13721
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4738422777412014e-05,
+      "loss": 0.7934,
+      "step": 13722
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.472453174724974e-05,
+      "loss": 0.8321,
+      "step": 13723
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4710641815898865e-05,
+      "loss": 0.8641,
+      "step": 13724
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.469675298369654e-05,
+      "loss": 0.9794,
+      "step": 13725
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4682865250979856e-05,
+      "loss": 0.8337,
+      "step": 13726
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.46689786180858e-05,
+      "loss": 0.8116,
+      "step": 13727
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4655093085351436e-05,
+      "loss": 0.8451,
+      "step": 13728
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.464120865311373e-05,
+      "loss": 0.7964,
+      "step": 13729
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4627325321709645e-05,
+      "loss": 0.82,
+      "step": 13730
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4613443091476155e-05,
+      "loss": 0.8663,
+      "step": 13731
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4599561962750175e-05,
+      "loss": 0.9442,
+      "step": 13732
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.458568193586858e-05,
+      "loss": 0.7101,
+      "step": 13733
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.457180301116819e-05,
+      "loss": 0.8498,
+      "step": 13734
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4557925188985915e-05,
+      "loss": 0.8529,
+      "step": 13735
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.454404846965853e-05,
+      "loss": 0.9973,
+      "step": 13736
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4530172853522756e-05,
+      "loss": 0.9894,
+      "step": 13737
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.451629834091545e-05,
+      "loss": 0.8302,
+      "step": 13738
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.450242493217328e-05,
+      "loss": 0.7582,
+      "step": 13739
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.448855262763293e-05,
+      "loss": 0.7827,
+      "step": 13740
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4474681427631126e-05,
+      "loss": 0.9617,
+      "step": 13741
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.44608113325045e-05,
+      "loss": 1.0829,
+      "step": 13742
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.444694234258967e-05,
+      "loss": 0.841,
+      "step": 13743
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4433074458223196e-05,
+      "loss": 0.8587,
+      "step": 13744
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.441920767974168e-05,
+      "loss": 0.9167,
+      "step": 13745
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4405342007481605e-05,
+      "loss": 0.9248,
+      "step": 13746
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.439147744177957e-05,
+      "loss": 0.8351,
+      "step": 13747
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4377613982972006e-05,
+      "loss": 0.8868,
+      "step": 13748
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4363751631395397e-05,
+      "loss": 1.4319,
+      "step": 13749
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4349890387386126e-05,
+      "loss": 1.6877,
+      "step": 13750
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.433603025128067e-05,
+      "loss": 1.1305,
+      "step": 13751
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.432217122341536e-05,
+      "loss": 0.9804,
+      "step": 13752
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.430831330412652e-05,
+      "loss": 0.9175,
+      "step": 13753
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4294456493750555e-05,
+      "loss": 0.9043,
+      "step": 13754
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.428060079262373e-05,
+      "loss": 0.9003,
+      "step": 13755
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.426674620108225e-05,
+      "loss": 0.9694,
+      "step": 13756
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.42528927194625e-05,
+      "loss": 0.9743,
+      "step": 13757
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4239040348100566e-05,
+      "loss": 0.7425,
+      "step": 13758
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.422518908733265e-05,
+      "loss": 0.8043,
+      "step": 13759
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.421133893749497e-05,
+      "loss": 0.8714,
+      "step": 13760
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.419748989892366e-05,
+      "loss": 1.4744,
+      "step": 13761
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4183641971954755e-05,
+      "loss": 1.1393,
+      "step": 13762
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.416979515692442e-05,
+      "loss": 0.7476,
+      "step": 13763
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.415594945416869e-05,
+      "loss": 1.3309,
+      "step": 13764
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4142104864023536e-05,
+      "loss": 1.6462,
+      "step": 13765
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4128261386825054e-05,
+      "loss": 1.1621,
+      "step": 13766
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4114419022909155e-05,
+      "loss": 0.8474,
+      "step": 13767
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4100577772611805e-05,
+      "loss": 0.734,
+      "step": 13768
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.408673763626888e-05,
+      "loss": 0.8113,
+      "step": 13769
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.4072898614216336e-05,
+      "loss": 0.8679,
+      "step": 13770
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.405906070679003e-05,
+      "loss": 0.9023,
+      "step": 13771
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.404522391432578e-05,
+      "loss": 0.9088,
+      "step": 13772
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.40313882371594e-05,
+      "loss": 0.9036,
+      "step": 13773
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.401755367562666e-05,
+      "loss": 0.859,
+      "step": 13774
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.400372023006332e-05,
+      "loss": 0.7603,
+      "step": 13775
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.398988790080515e-05,
+      "loss": 0.8833,
+      "step": 13776
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3976056688187815e-05,
+      "loss": 0.8569,
+      "step": 13777
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3962226592546975e-05,
+      "loss": 0.9028,
+      "step": 13778
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.394839761421835e-05,
+      "loss": 0.8377,
+      "step": 13779
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.393456975353751e-05,
+      "loss": 0.9613,
+      "step": 13780
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.392074301084003e-05,
+      "loss": 0.8631,
+      "step": 13781
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.390691738646153e-05,
+      "loss": 0.9085,
+      "step": 13782
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.389309288073754e-05,
+      "loss": 0.8316,
+      "step": 13783
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3879269494003546e-05,
+      "loss": 0.8279,
+      "step": 13784
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.386544722659501e-05,
+      "loss": 0.9256,
+      "step": 13785
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.385162607884748e-05,
+      "loss": 0.8517,
+      "step": 13786
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.383780605109634e-05,
+      "loss": 0.8167,
+      "step": 13787
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.382398714367698e-05,
+      "loss": 0.8068,
+      "step": 13788
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.381016935692479e-05,
+      "loss": 0.9186,
+      "step": 13789
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3796352691175125e-05,
+      "loss": 1.0188,
+      "step": 13790
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3782537146763254e-05,
+      "loss": 0.7589,
+      "step": 13791
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.376872272402458e-05,
+      "loss": 0.792,
+      "step": 13792
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.375490942329429e-05,
+      "loss": 0.8152,
+      "step": 13793
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.374109724490762e-05,
+      "loss": 1.2174,
+      "step": 13794
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.372728618919986e-05,
+      "loss": 1.0724,
+      "step": 13795
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3713476256506135e-05,
+      "loss": 1.0013,
+      "step": 13796
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3699667447161595e-05,
+      "loss": 0.8767,
+      "step": 13797
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3685859761501425e-05,
+      "loss": 0.6815,
+      "step": 13798
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3672053199860706e-05,
+      "loss": 1.2981,
+      "step": 13799
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.365824776257451e-05,
+      "loss": 1.0541,
+      "step": 13800
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.364444344997788e-05,
+      "loss": 0.7193,
+      "step": 13801
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3630640262405864e-05,
+      "loss": 0.6243,
+      "step": 13802
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.36168382001934e-05,
+      "loss": 0.7592,
+      "step": 13803
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.360303726367554e-05,
+      "loss": 0.9124,
+      "step": 13804
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.358923745318718e-05,
+      "loss": 0.9193,
+      "step": 13805
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.357543876906324e-05,
+      "loss": 0.8032,
+      "step": 13806
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.356164121163856e-05,
+      "loss": 0.8805,
+      "step": 13807
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.35478447812481e-05,
+      "loss": 0.9034,
+      "step": 13808
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.353404947822662e-05,
+      "loss": 1.0571,
+      "step": 13809
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3520255302908916e-05,
+      "loss": 0.6637,
+      "step": 13810
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.350646225562983e-05,
+      "loss": 0.8541,
+      "step": 13811
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.349267033672406e-05,
+      "loss": 0.8949,
+      "step": 13812
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.347887954652631e-05,
+      "loss": 0.8895,
+      "step": 13813
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3465089885371346e-05,
+      "loss": 0.9737,
+      "step": 13814
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.34513013535938e-05,
+      "loss": 0.8948,
+      "step": 13815
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.34375139515283e-05,
+      "loss": 0.9463,
+      "step": 13816
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.342372767950946e-05,
+      "loss": 0.8996,
+      "step": 13817
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3409942537871885e-05,
+      "loss": 0.7242,
+      "step": 13818
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.339615852695008e-05,
+      "loss": 0.8875,
+      "step": 13819
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3382375647078654e-05,
+      "loss": 0.9654,
+      "step": 13820
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.336859389859208e-05,
+      "loss": 0.8146,
+      "step": 13821
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.335481328182477e-05,
+      "loss": 0.9353,
+      "step": 13822
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3341033797111285e-05,
+      "loss": 0.8146,
+      "step": 13823
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3327255444785976e-05,
+      "loss": 0.9358,
+      "step": 13824
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3313478225183245e-05,
+      "loss": 0.8175,
+      "step": 13825
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.329970213863744e-05,
+      "loss": 0.8407,
+      "step": 13826
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3285927185482954e-05,
+      "loss": 0.7424,
+      "step": 13827
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.327215336605407e-05,
+      "loss": 1.0069,
+      "step": 13828
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3258380680685004e-05,
+      "loss": 0.9341,
+      "step": 13829
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3244609129710146e-05,
+      "loss": 0.8013,
+      "step": 13830
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.323083871346366e-05,
+      "loss": 0.7881,
+      "step": 13831
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3217069432279684e-05,
+      "loss": 0.9586,
+      "step": 13832
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.320330128649247e-05,
+      "loss": 0.8353,
+      "step": 13833
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.318953427643615e-05,
+      "loss": 0.9786,
+      "step": 13834
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.317576840244478e-05,
+      "loss": 0.8881,
+      "step": 13835
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.316200366485256e-05,
+      "loss": 0.8639,
+      "step": 13836
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3148240063993485e-05,
+      "loss": 0.9008,
+      "step": 13837
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.313447760020156e-05,
+      "loss": 0.9099,
+      "step": 13838
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3120716273810866e-05,
+      "loss": 0.8465,
+      "step": 13839
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3106956085155344e-05,
+      "loss": 0.8235,
+      "step": 13840
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.309319703456897e-05,
+      "loss": 0.9816,
+      "step": 13841
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.30794391223856e-05,
+      "loss": 0.8173,
+      "step": 13842
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.306568234893923e-05,
+      "loss": 0.8646,
+      "step": 13843
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.305192671456368e-05,
+      "loss": 0.9869,
+      "step": 13844
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.303817221959275e-05,
+      "loss": 0.8713,
+      "step": 13845
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.3024418864360384e-05,
+      "loss": 0.9196,
+      "step": 13846
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.301066664920023e-05,
+      "loss": 0.6819,
+      "step": 13847
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.299691557444607e-05,
+      "loss": 0.9436,
+      "step": 13848
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.29831656404317e-05,
+      "loss": 0.867,
+      "step": 13849
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.296941684749079e-05,
+      "loss": 0.9673,
+      "step": 13850
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2955669195956966e-05,
+      "loss": 0.9163,
+      "step": 13851
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.294192268616398e-05,
+      "loss": 0.8876,
+      "step": 13852
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.292817731844538e-05,
+      "loss": 1.3045,
+      "step": 13853
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2914433093134755e-05,
+      "loss": 1.2026,
+      "step": 13854
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.29006900105657e-05,
+      "loss": 0.958,
+      "step": 13855
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.288694807107177e-05,
+      "loss": 0.8003,
+      "step": 13856
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2873207274986434e-05,
+      "loss": 0.8067,
+      "step": 13857
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.285946762264315e-05,
+      "loss": 0.8077,
+      "step": 13858
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.284572911437544e-05,
+      "loss": 0.9479,
+      "step": 13859
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.283199175051672e-05,
+      "loss": 0.8213,
+      "step": 13860
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.281825553140035e-05,
+      "loss": 0.8987,
+      "step": 13861
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.280452045735973e-05,
+      "loss": 0.9721,
+      "step": 13862
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2790786528728175e-05,
+      "loss": 0.9361,
+      "step": 13863
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.277705374583899e-05,
+      "loss": 0.9859,
+      "step": 13864
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2763322109025524e-05,
+      "loss": 0.7315,
+      "step": 13865
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2749591618620995e-05,
+      "loss": 0.8138,
+      "step": 13866
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.273586227495861e-05,
+      "loss": 0.7329,
+      "step": 13867
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.272213407837164e-05,
+      "loss": 0.7975,
+      "step": 13868
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.270840702919323e-05,
+      "loss": 0.8182,
+      "step": 13869
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2694681127756485e-05,
+      "loss": 0.8101,
+      "step": 13870
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2680956374394586e-05,
+      "loss": 0.7712,
+      "step": 13871
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.266723276944061e-05,
+      "loss": 0.8972,
+      "step": 13872
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.265351031322763e-05,
+      "loss": 0.8488,
+      "step": 13873
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2639789006088615e-05,
+      "loss": 0.8574,
+      "step": 13874
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.262606884835669e-05,
+      "loss": 0.8664,
+      "step": 13875
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2612349840364716e-05,
+      "loss": 0.8707,
+      "step": 13876
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.2598631982445726e-05,
+      "loss": 0.9587,
+      "step": 13877
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2584915274932624e-05,
+      "loss": 0.8187,
+      "step": 13878
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.257119971815831e-05,
+      "loss": 0.8157,
+      "step": 13879
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.255748531245559e-05,
+      "loss": 0.9877,
+      "step": 13880
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.254377205815742e-05,
+      "loss": 0.9158,
+      "step": 13881
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2530059955596545e-05,
+      "loss": 1.2389,
+      "step": 13882
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2516349005105736e-05,
+      "loss": 1.1102,
+      "step": 13883
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.25026392070178e-05,
+      "loss": 0.8959,
+      "step": 13884
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.248893056166544e-05,
+      "loss": 0.8912,
+      "step": 13885
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.247522306938133e-05,
+      "loss": 0.8275,
+      "step": 13886
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.246151673049822e-05,
+      "loss": 0.9565,
+      "step": 13887
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.244781154534869e-05,
+      "loss": 0.9651,
+      "step": 13888
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.243410751426539e-05,
+      "loss": 0.8909,
+      "step": 13889
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.242040463758089e-05,
+      "loss": 0.6993,
+      "step": 13890
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.240670291562775e-05,
+      "loss": 0.8422,
+      "step": 13891
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.239300234873847e-05,
+      "loss": 0.8643,
+      "step": 13892
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.237930293724564e-05,
+      "loss": 0.9474,
+      "step": 13893
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.23656046814817e-05,
+      "loss": 0.8889,
+      "step": 13894
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.235190758177906e-05,
+      "loss": 0.8459,
+      "step": 13895
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.233821163847019e-05,
+      "loss": 0.721,
+      "step": 13896
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2324516851887483e-05,
+      "loss": 0.8924,
+      "step": 13897
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.231082322236329e-05,
+      "loss": 0.8903,
+      "step": 13898
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2297130750229895e-05,
+      "loss": 0.8572,
+      "step": 13899
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.22834394358197e-05,
+      "loss": 0.6953,
+      "step": 13900
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2269749279464954e-05,
+      "loss": 0.827,
+      "step": 13901
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2256060281497854e-05,
+      "loss": 0.9565,
+      "step": 13902
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.224237244225072e-05,
+      "loss": 0.7748,
+      "step": 13903
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.222868576205573e-05,
+      "loss": 0.8982,
+      "step": 13904
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.221500024124494e-05,
+      "loss": 0.7113,
+      "step": 13905
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2201315880150606e-05,
+      "loss": 0.9081,
+      "step": 13906
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2187632679104826e-05,
+      "loss": 0.7835,
+      "step": 13907
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.217395063843961e-05,
+      "loss": 0.8332,
+      "step": 13908
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2160269758487116e-05,
+      "loss": 0.8437,
+      "step": 13909
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2146590039579314e-05,
+      "loss": 0.9066,
+      "step": 13910
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.213291148204817e-05,
+      "loss": 0.8964,
+      "step": 13911
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2119234086225745e-05,
+      "loss": 0.9065,
+      "step": 13912
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.210555785244393e-05,
+      "loss": 0.799,
+      "step": 13913
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.209188278103465e-05,
+      "loss": 1.0128,
+      "step": 13914
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2078208872329756e-05,
+      "loss": 0.8867,
+      "step": 13915
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.206453612666117e-05,
+      "loss": 0.8853,
+      "step": 13916
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.205086454436069e-05,
+      "loss": 0.907,
+      "step": 13917
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2037194125760067e-05,
+      "loss": 0.8231,
+      "step": 13918
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.202352487119122e-05,
+      "loss": 0.8989,
+      "step": 13919
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.2009856780985754e-05,
+      "loss": 0.9164,
+      "step": 13920
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.19961898554754e-05,
+      "loss": 0.957,
+      "step": 13921
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.198252409499191e-05,
+      "loss": 0.9345,
+      "step": 13922
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1968859499866917e-05,
+      "loss": 0.937,
+      "step": 13923
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.195519607043201e-05,
+      "loss": 0.8743,
+      "step": 13924
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1941533807018875e-05,
+      "loss": 0.7702,
+      "step": 13925
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.192787270995905e-05,
+      "loss": 0.901,
+      "step": 13926
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.191421277958402e-05,
+      "loss": 0.8687,
+      "step": 13927
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.190055401622542e-05,
+      "loss": 0.8871,
+      "step": 13928
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.188689642021468e-05,
+      "loss": 0.7635,
+      "step": 13929
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.187323999188327e-05,
+      "loss": 0.825,
+      "step": 13930
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1859584731562573e-05,
+      "loss": 0.9204,
+      "step": 13931
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.184593063958408e-05,
+      "loss": 0.8643,
+      "step": 13932
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.183227771627914e-05,
+      "loss": 1.0174,
+      "step": 13933
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.181862596197909e-05,
+      "loss": 1.3849,
+      "step": 13934
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1804975377015254e-05,
+      "loss": 1.1277,
+      "step": 13935
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1791325961718926e-05,
+      "loss": 0.7144,
+      "step": 13936
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.177767771642132e-05,
+      "loss": 0.9769,
+      "step": 13937
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1764030641453785e-05,
+      "loss": 0.89,
+      "step": 13938
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.175038473714744e-05,
+      "loss": 0.9988,
+      "step": 13939
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.173674000383346e-05,
+      "loss": 0.8347,
+      "step": 13940
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1723096441843057e-05,
+      "loss": 1.0738,
+      "step": 13941
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.170945405150731e-05,
+      "loss": 0.9005,
+      "step": 13942
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.169581283315729e-05,
+      "loss": 0.8943,
+      "step": 13943
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.168217278712413e-05,
+      "loss": 0.9398,
+      "step": 13944
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.166853391373882e-05,
+      "loss": 0.8955,
+      "step": 13945
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.165489621333239e-05,
+      "loss": 0.8306,
+      "step": 13946
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1641259686235774e-05,
+      "loss": 0.8353,
+      "step": 13947
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1627624332780033e-05,
+      "loss": 0.8886,
+      "step": 13948
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.161399015329593e-05,
+      "loss": 0.8945,
+      "step": 13949
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1600357148114475e-05,
+      "loss": 0.9169,
+      "step": 13950
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.158672531756652e-05,
+      "loss": 0.8002,
+      "step": 13951
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.157309466198286e-05,
+      "loss": 1.369,
+      "step": 13952
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.15594651816943e-05,
+      "loss": 0.9899,
+      "step": 13953
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.154583687703169e-05,
+      "loss": 0.9002,
+      "step": 13954
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.153220974832573e-05,
+      "loss": 0.9268,
+      "step": 13955
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.151858379590713e-05,
+      "loss": 0.8842,
+      "step": 13956
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1504959020106635e-05,
+      "loss": 0.8501,
+      "step": 13957
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1491335421254884e-05,
+      "loss": 1.0749,
+      "step": 13958
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.147771299968247e-05,
+      "loss": 0.9391,
+      "step": 13959
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.14640917557201e-05,
+      "loss": 0.8693,
+      "step": 13960
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.145047168969829e-05,
+      "loss": 0.9397,
+      "step": 13961
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.143685280194756e-05,
+      "loss": 0.8819,
+      "step": 13962
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.142323509279856e-05,
+      "loss": 0.8911,
+      "step": 13963
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.140961856258165e-05,
+      "loss": 0.9317,
+      "step": 13964
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.139600321162732e-05,
+      "loss": 1.3626,
+      "step": 13965
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.138238904026606e-05,
+      "loss": 1.1864,
+      "step": 13966
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.136877604882825e-05,
+      "loss": 0.8798,
+      "step": 13967
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1355164237644235e-05,
+      "loss": 0.8135,
+      "step": 13968
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1341553607044445e-05,
+      "loss": 0.8733,
+      "step": 13969
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.132794415735915e-05,
+      "loss": 0.8727,
+      "step": 13970
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.131433588891867e-05,
+      "loss": 0.8628,
+      "step": 13971
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.130072880205321e-05,
+      "loss": 0.9223,
+      "step": 13972
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.128712289709309e-05,
+      "loss": 0.8984,
+      "step": 13973
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1273518174368473e-05,
+      "loss": 0.8846,
+      "step": 13974
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1259914634209514e-05,
+      "loss": 0.8225,
+      "step": 13975
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.124631227694643e-05,
+      "loss": 0.7804,
+      "step": 13976
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.1232711102909304e-05,
+      "loss": 0.8311,
+      "step": 13977
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.121911111242823e-05,
+      "loss": 0.8631,
+      "step": 13978
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.120551230583329e-05,
+      "loss": 0.8655,
+      "step": 13979
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.11919146834545e-05,
+      "loss": 1.0075,
+      "step": 13980
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.117831824562183e-05,
+      "loss": 0.8708,
+      "step": 13981
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1164722992665336e-05,
+      "loss": 1.0002,
+      "step": 13982
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.115112892491494e-05,
+      "loss": 1.3689,
+      "step": 13983
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1137536042700504e-05,
+      "loss": 1.0926,
+      "step": 13984
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.112394434635202e-05,
+      "loss": 0.854,
+      "step": 13985
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.111035383619929e-05,
+      "loss": 0.7772,
+      "step": 13986
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.109676451257217e-05,
+      "loss": 0.7888,
+      "step": 13987
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1083176375800424e-05,
+      "loss": 0.8346,
+      "step": 13988
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1069589426213894e-05,
+      "loss": 0.9583,
+      "step": 13989
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.105600366414228e-05,
+      "loss": 1.0281,
+      "step": 13990
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1042419089915295e-05,
+      "loss": 0.9568,
+      "step": 13991
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.102883570386268e-05,
+      "loss": 0.788,
+      "step": 13992
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.101525350631411e-05,
+      "loss": 0.7845,
+      "step": 13993
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1001672497599087e-05,
+      "loss": 0.9292,
+      "step": 13994
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.098809267804735e-05,
+      "loss": 0.765,
+      "step": 13995
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.097451404798843e-05,
+      "loss": 0.8762,
+      "step": 13996
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.096093660775183e-05,
+      "loss": 0.8083,
+      "step": 13997
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.094736035766714e-05,
+      "loss": 0.7931,
+      "step": 13998
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.093378529806381e-05,
+      "loss": 0.953,
+      "step": 13999
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.092021142927127e-05,
+      "loss": 0.7685,
+      "step": 14000
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.090663875161903e-05,
+      "loss": 0.9808,
+      "step": 14001
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0893067265436435e-05,
+      "loss": 0.8781,
+      "step": 14002
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.087949697105288e-05,
+      "loss": 0.8409,
+      "step": 14003
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0865927868797646e-05,
+      "loss": 0.9731,
+      "step": 14004
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.085235995900015e-05,
+      "loss": 0.9134,
+      "step": 14005
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0838793241989635e-05,
+      "loss": 0.9115,
+      "step": 14006
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.082522771809534e-05,
+      "loss": 0.8821,
+      "step": 14007
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0811663387646504e-05,
+      "loss": 1.3434,
+      "step": 14008
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.079810025097232e-05,
+      "loss": 1.1104,
+      "step": 14009
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.078453830840193e-05,
+      "loss": 0.8566,
+      "step": 14010
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.077097756026455e-05,
+      "loss": 0.9541,
+      "step": 14011
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0757418006889257e-05,
+      "loss": 1.3357,
+      "step": 14012
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0743859648605086e-05,
+      "loss": 1.236,
+      "step": 14013
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0730302485741164e-05,
+      "loss": 0.9083,
+      "step": 14014
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0716746518626504e-05,
+      "loss": 0.81,
+      "step": 14015
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.070319174759002e-05,
+      "loss": 0.9373,
+      "step": 14016
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0689638172960795e-05,
+      "loss": 1.343,
+      "step": 14017
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0676085795067706e-05,
+      "loss": 1.5494,
+      "step": 14018
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.066253461423969e-05,
+      "loss": 1.2101,
+      "step": 14019
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.064898463080555e-05,
+      "loss": 0.7449,
+      "step": 14020
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.063543584509425e-05,
+      "loss": 0.837,
+      "step": 14021
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0621888257434534e-05,
+      "loss": 0.8873,
+      "step": 14022
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.060834186815523e-05,
+      "loss": 0.8586,
+      "step": 14023
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0594796677585085e-05,
+      "loss": 0.9787,
+      "step": 14024
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.05812526860528e-05,
+      "loss": 0.9389,
+      "step": 14025
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.056770989388716e-05,
+      "loss": 0.7718,
+      "step": 14026
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.055416830141679e-05,
+      "loss": 0.9297,
+      "step": 14027
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0540627908970364e-05,
+      "loss": 0.8519,
+      "step": 14028
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.052708871687643e-05,
+      "loss": 0.9614,
+      "step": 14029
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.051355072546367e-05,
+      "loss": 0.8798,
+      "step": 14030
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.05000139350606e-05,
+      "loss": 0.8329,
+      "step": 14031
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.048647834599571e-05,
+      "loss": 1.0249,
+      "step": 14032
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0472943958597595e-05,
+      "loss": 1.408,
+      "step": 14033
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0459410773194674e-05,
+      "loss": 1.1799,
+      "step": 14034
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.044587879011535e-05,
+      "loss": 0.919,
+      "step": 14035
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0432348009688125e-05,
+      "loss": 0.7366,
+      "step": 14036
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.041881843224139e-05,
+      "loss": 0.8964,
+      "step": 14037
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.040529005810335e-05,
+      "loss": 0.9184,
+      "step": 14038
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.039176288760248e-05,
+      "loss": 0.9372,
+      "step": 14039
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0378236921067025e-05,
+      "loss": 0.9429,
+      "step": 14040
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.036471215882523e-05,
+      "loss": 0.901,
+      "step": 14041
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.035118860120538e-05,
+      "loss": 0.8713,
+      "step": 14042
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.033766624853569e-05,
+      "loss": 0.9093,
+      "step": 14043
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.032414510114429e-05,
+      "loss": 0.8331,
+      "step": 14044
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0310625159359336e-05,
+      "loss": 0.7883,
+      "step": 14045
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0297106423509e-05,
+      "loss": 0.8956,
+      "step": 14046
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.028358889392134e-05,
+      "loss": 0.8793,
+      "step": 14047
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.02700725709244e-05,
+      "loss": 0.8218,
+      "step": 14048
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0256557454846255e-05,
+      "loss": 0.9973,
+      "step": 14049
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.024304354601491e-05,
+      "loss": 0.9006,
+      "step": 14050
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0229530844758324e-05,
+      "loss": 1.3457,
+      "step": 14051
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.021601935140443e-05,
+      "loss": 1.0391,
+      "step": 14052
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0202509066281165e-05,
+      "loss": 0.8599,
+      "step": 14053
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.018899998971636e-05,
+      "loss": 0.7136,
+      "step": 14054
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.017549212203797e-05,
+      "loss": 0.8263,
+      "step": 14055
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.016198546357377e-05,
+      "loss": 0.7994,
+      "step": 14056
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.014848001465153e-05,
+      "loss": 0.8625,
+      "step": 14057
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.013497577559909e-05,
+      "loss": 0.7263,
+      "step": 14058
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0121472746744146e-05,
+      "loss": 0.8122,
+      "step": 14059
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.010797092841443e-05,
+      "loss": 0.7842,
+      "step": 14060
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.009447032093757e-05,
+      "loss": 0.8734,
+      "step": 14061
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.00809709246413e-05,
+      "loss": 0.8656,
+      "step": 14062
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.006747273985319e-05,
+      "loss": 0.7223,
+      "step": 14063
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.005397576690083e-05,
+      "loss": 0.8509,
+      "step": 14064
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.004048000611183e-05,
+      "loss": 1.0018,
+      "step": 14065
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.002698545781374e-05,
+      "loss": 0.765,
+      "step": 14066
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.001349212233393e-05,
+      "loss": 0.8873,
+      "step": 14067
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.000000000000002e-05,
+      "loss": 0.8515,
+      "step": 14068
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.99865090911394e-05,
+      "loss": 0.9782,
+      "step": 14069
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.997301939607946e-05,
+      "loss": 0.8855,
+      "step": 14070
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9959530915147654e-05,
+      "loss": 0.8988,
+      "step": 14071
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9946043648671306e-05,
+      "loss": 0.8956,
+      "step": 14072
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.99325575969777e-05,
+      "loss": 0.7703,
+      "step": 14073
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9919072760394234e-05,
+      "loss": 0.7168,
+      "step": 14074
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.990558913924811e-05,
+      "loss": 0.9352,
+      "step": 14075
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.989210673386659e-05,
+      "loss": 0.8217,
+      "step": 14076
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.987862554457683e-05,
+      "loss": 0.7725,
+      "step": 14077
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.986514557170611e-05,
+      "loss": 0.8441,
+      "step": 14078
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.985166681558151e-05,
+      "loss": 0.6961,
+      "step": 14079
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9838189276530145e-05,
+      "loss": 0.9089,
+      "step": 14080
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.982471295487922e-05,
+      "loss": 0.7535,
+      "step": 14081
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.981123785095566e-05,
+      "loss": 0.8096,
+      "step": 14082
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9797763965086516e-05,
+      "loss": 0.7856,
+      "step": 14083
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.978429129759886e-05,
+      "loss": 0.8959,
+      "step": 14084
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.977081984881965e-05,
+      "loss": 0.8817,
+      "step": 14085
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.975734961907577e-05,
+      "loss": 0.8448,
+      "step": 14086
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.97438806086942e-05,
+      "loss": 0.9731,
+      "step": 14087
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.973041281800183e-05,
+      "loss": 1.0113,
+      "step": 14088
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9716946247325446e-05,
+      "loss": 1.2259,
+      "step": 14089
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.970348089699196e-05,
+      "loss": 1.101,
+      "step": 14090
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.969001676732812e-05,
+      "loss": 0.9583,
+      "step": 14091
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9676553858660713e-05,
+      "loss": 0.849,
+      "step": 14092
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.966309217131642e-05,
+      "loss": 0.8776,
+      "step": 14093
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9649631705622035e-05,
+      "loss": 0.8526,
+      "step": 14094
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.96361724619042e-05,
+      "loss": 0.8371,
+      "step": 14095
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.962271444048956e-05,
+      "loss": 0.9158,
+      "step": 14096
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.960925764170473e-05,
+      "loss": 0.7511,
+      "step": 14097
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.959580206587626e-05,
+      "loss": 0.8972,
+      "step": 14098
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.95823477133308e-05,
+      "loss": 0.8375,
+      "step": 14099
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.956889458439483e-05,
+      "loss": 0.8968,
+      "step": 14100
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9555442679394846e-05,
+      "loss": 1.2679,
+      "step": 14101
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.95419919986573e-05,
+      "loss": 1.1883,
+      "step": 14102
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.952854254250869e-05,
+      "loss": 0.8536,
+      "step": 14103
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.951509431127539e-05,
+      "loss": 1.0112,
+      "step": 14104
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.950164730528376e-05,
+      "loss": 0.9843,
+      "step": 14105
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.948820152486021e-05,
+      "loss": 0.9372,
+      "step": 14106
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.947475697033103e-05,
+      "loss": 1.2094,
+      "step": 14107
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9461313642022476e-05,
+      "loss": 1.0625,
+      "step": 14108
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.944787154026088e-05,
+      "loss": 0.8922,
+      "step": 14109
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9434430665372486e-05,
+      "loss": 0.9369,
+      "step": 14110
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.942099101768338e-05,
+      "loss": 0.8485,
+      "step": 14111
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9407552597519836e-05,
+      "loss": 0.8341,
+      "step": 14112
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9394115405207975e-05,
+      "loss": 0.8989,
+      "step": 14113
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.938067944107386e-05,
+      "loss": 0.9494,
+      "step": 14114
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.936724470544367e-05,
+      "loss": 0.9463,
+      "step": 14115
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9353811198643394e-05,
+      "loss": 0.7434,
+      "step": 14116
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9340378920999075e-05,
+      "loss": 0.7887,
+      "step": 14117
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.932694787283667e-05,
+      "loss": 0.8186,
+      "step": 14118
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9313518054482214e-05,
+      "loss": 0.8979,
+      "step": 14119
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.930008946626159e-05,
+      "loss": 0.9185,
+      "step": 14120
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.928666210850068e-05,
+      "loss": 0.8419,
+      "step": 14121
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.927323598152543e-05,
+      "loss": 0.6915,
+      "step": 14122
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9259811085661664e-05,
+      "loss": 0.9436,
+      "step": 14123
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9246387421235144e-05,
+      "loss": 0.9362,
+      "step": 14124
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.923296498857175e-05,
+      "loss": 0.8655,
+      "step": 14125
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.921954378799716e-05,
+      "loss": 0.8552,
+      "step": 14126
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.920612381983708e-05,
+      "loss": 0.8116,
+      "step": 14127
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9192705084417277e-05,
+      "loss": 0.8408,
+      "step": 14128
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.917928758206339e-05,
+      "loss": 0.9767,
+      "step": 14129
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.916587131310101e-05,
+      "loss": 0.891,
+      "step": 14130
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9152456277855816e-05,
+      "loss": 0.7517,
+      "step": 14131
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9139042476653364e-05,
+      "loss": 0.9182,
+      "step": 14132
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.912562990981917e-05,
+      "loss": 0.9368,
+      "step": 14133
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9112218577678736e-05,
+      "loss": 0.9609,
+      "step": 14134
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.909880848055761e-05,
+      "loss": 0.9367,
+      "step": 14135
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9085399618781224e-05,
+      "loss": 0.8884,
+      "step": 14136
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.907199199267496e-05,
+      "loss": 0.7889,
+      "step": 14137
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.905858560256428e-05,
+      "loss": 0.9081,
+      "step": 14138
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9045180448774516e-05,
+      "loss": 0.9006,
+      "step": 14139
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9031776531631015e-05,
+      "loss": 0.9357,
+      "step": 14140
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.901837385145908e-05,
+      "loss": 0.8442,
+      "step": 14141
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.900497240858397e-05,
+      "loss": 0.8159,
+      "step": 14142
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8991572203330907e-05,
+      "loss": 0.9064,
+      "step": 14143
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.897817323602518e-05,
+      "loss": 0.8538,
+      "step": 14144
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8964775506991944e-05,
+      "loss": 0.9104,
+      "step": 14145
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.895137901655631e-05,
+      "loss": 0.9623,
+      "step": 14146
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.893798376504347e-05,
+      "loss": 0.7956,
+      "step": 14147
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.892458975277851e-05,
+      "loss": 0.9116,
+      "step": 14148
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.891119698008646e-05,
+      "loss": 1.3549,
+      "step": 14149
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8897805447292334e-05,
+      "loss": 1.0152,
+      "step": 14150
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.888441515472122e-05,
+      "loss": 0.9639,
+      "step": 14151
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.887102610269805e-05,
+      "loss": 0.9182,
+      "step": 14152
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.885763829154773e-05,
+      "loss": 0.8771,
+      "step": 14153
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.884425172159529e-05,
+      "loss": 0.9189,
+      "step": 14154
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.883086639316546e-05,
+      "loss": 0.8326,
+      "step": 14155
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.881748230658322e-05,
+      "loss": 0.8595,
+      "step": 14156
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8804099462173336e-05,
+      "loss": 0.8654,
+      "step": 14157
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8790717860260614e-05,
+      "loss": 0.7335,
+      "step": 14158
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8777337501169784e-05,
+      "loss": 0.8495,
+      "step": 14159
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.876395838522565e-05,
+      "loss": 0.8363,
+      "step": 14160
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8750580512752884e-05,
+      "loss": 0.8976,
+      "step": 14161
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8737203884076124e-05,
+      "loss": 0.8652,
+      "step": 14162
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8723828499520075e-05,
+      "loss": 0.8402,
+      "step": 14163
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.871045435940932e-05,
+      "loss": 0.7614,
+      "step": 14164
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.869708146406842e-05,
+      "loss": 0.8438,
+      "step": 14165
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.868370981382198e-05,
+      "loss": 0.9284,
+      "step": 14166
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8670339408994494e-05,
+      "loss": 0.8576,
+      "step": 14167
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.865697024991045e-05,
+      "loss": 1.1794,
+      "step": 14168
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8643602336894326e-05,
+      "loss": 1.2022,
+      "step": 14169
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.863023567027053e-05,
+      "loss": 0.9224,
+      "step": 14170
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8616870250363454e-05,
+      "loss": 0.8504,
+      "step": 14171
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.860350607749752e-05,
+      "loss": 0.8648,
+      "step": 14172
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8590143151997045e-05,
+      "loss": 0.7875,
+      "step": 14173
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.857678147418634e-05,
+      "loss": 0.9071,
+      "step": 14174
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.856342104438964e-05,
+      "loss": 0.9444,
+      "step": 14175
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8550061862931296e-05,
+      "loss": 0.8809,
+      "step": 14176
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.853670393013545e-05,
+      "loss": 0.8875,
+      "step": 14177
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.852334724632628e-05,
+      "loss": 0.8945,
+      "step": 14178
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.850999181182801e-05,
+      "loss": 1.002,
+      "step": 14179
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8496637626964755e-05,
+      "loss": 0.9234,
+      "step": 14180
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.848328469206055e-05,
+      "loss": 0.7939,
+      "step": 14181
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.846993300743955e-05,
+      "loss": 0.8935,
+      "step": 14182
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8456582573425744e-05,
+      "loss": 1.3355,
+      "step": 14183
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8443233390343154e-05,
+      "loss": 1.1589,
+      "step": 14184
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.842988545851577e-05,
+      "loss": 0.8071,
+      "step": 14185
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8416538778267496e-05,
+      "loss": 0.7074,
+      "step": 14186
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.840319334992225e-05,
+      "loss": 1.5051,
+      "step": 14187
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.838984917380398e-05,
+      "loss": 1.2702,
+      "step": 14188
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.837650625023651e-05,
+      "loss": 0.8368,
+      "step": 14189
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.836316457954365e-05,
+      "loss": 0.9031,
+      "step": 14190
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8349824162049175e-05,
+      "loss": 0.8868,
+      "step": 14191
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8336484998076916e-05,
+      "loss": 0.8527,
+      "step": 14192
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.832314708795058e-05,
+      "loss": 0.7778,
+      "step": 14193
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.830981043199382e-05,
+      "loss": 0.7528,
+      "step": 14194
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.829647503053039e-05,
+      "loss": 0.8842,
+      "step": 14195
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.828314088388389e-05,
+      "loss": 1.0092,
+      "step": 14196
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8269807992377915e-05,
+      "loss": 1.2614,
+      "step": 14197
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.825647635633613e-05,
+      "loss": 0.9437,
+      "step": 14198
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8243145976082003e-05,
+      "loss": 0.8135,
+      "step": 14199
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.822981685193904e-05,
+      "loss": 0.9754,
+      "step": 14200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.82164889842308e-05,
+      "loss": 0.8624,
+      "step": 14201
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.820316237328072e-05,
+      "loss": 0.9557,
+      "step": 14202
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8189837019412184e-05,
+      "loss": 0.8459,
+      "step": 14203
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.817651292294868e-05,
+      "loss": 0.861,
+      "step": 14204
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.816319008421352e-05,
+      "loss": 1.0178,
+      "step": 14205
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.814986850353005e-05,
+      "loss": 0.8988,
+      "step": 14206
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.813654818122154e-05,
+      "loss": 1.2451,
+      "step": 14207
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8123229117611344e-05,
+      "loss": 1.102,
+      "step": 14208
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8109911313022674e-05,
+      "loss": 0.9065,
+      "step": 14209
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.809659476777869e-05,
+      "loss": 0.7637,
+      "step": 14210
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.808327948220268e-05,
+      "loss": 0.939,
+      "step": 14211
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.806996545661775e-05,
+      "loss": 0.9687,
+      "step": 14212
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8056652691347016e-05,
+      "loss": 0.949,
+      "step": 14213
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.804334118671357e-05,
+      "loss": 0.8003,
+      "step": 14214
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8030030943040494e-05,
+      "loss": 0.8896,
+      "step": 14215
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.801672196065078e-05,
+      "loss": 0.8753,
+      "step": 14216
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.8003414239867486e-05,
+      "loss": 0.8288,
+      "step": 14217
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.799010778101355e-05,
+      "loss": 0.8988,
+      "step": 14218
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7976802584411896e-05,
+      "loss": 0.886,
+      "step": 14219
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7963498650385484e-05,
+      "loss": 0.9328,
+      "step": 14220
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7950195979257176e-05,
+      "loss": 0.9066,
+      "step": 14221
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.79368945713498e-05,
+      "loss": 0.7881,
+      "step": 14222
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7923594426986164e-05,
+      "loss": 1.0084,
+      "step": 14223
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7910295546489104e-05,
+      "loss": 0.9075,
+      "step": 14224
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.789699793018135e-05,
+      "loss": 0.8614,
+      "step": 14225
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7883701578385585e-05,
+      "loss": 0.8132,
+      "step": 14226
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.78704064914246e-05,
+      "loss": 0.7364,
+      "step": 14227
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7857112669620994e-05,
+      "loss": 0.8817,
+      "step": 14228
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7843820113297425e-05,
+      "loss": 0.8995,
+      "step": 14229
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.783052882277648e-05,
+      "loss": 0.7677,
+      "step": 14230
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.781723879838073e-05,
+      "loss": 0.8896,
+      "step": 14231
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7803950040432686e-05,
+      "loss": 0.87,
+      "step": 14232
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7790662549254936e-05,
+      "loss": 0.8309,
+      "step": 14233
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.777737632516993e-05,
+      "loss": 0.8707,
+      "step": 14234
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7764091368500065e-05,
+      "loss": 0.9025,
+      "step": 14235
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7750807679567835e-05,
+      "loss": 0.9895,
+      "step": 14236
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.77375252586956e-05,
+      "loss": 0.9227,
+      "step": 14237
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.772424410620567e-05,
+      "loss": 1.0356,
+      "step": 14238
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.771096422242046e-05,
+      "loss": 0.7805,
+      "step": 14239
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7697685607662224e-05,
+      "loss": 0.9026,
+      "step": 14240
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7684408262253214e-05,
+      "loss": 0.8306,
+      "step": 14241
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.767113218651568e-05,
+      "loss": 1.0625,
+      "step": 14242
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.765785738077182e-05,
+      "loss": 0.8901,
+      "step": 14243
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.764458384534376e-05,
+      "loss": 0.9245,
+      "step": 14244
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.763131158055373e-05,
+      "loss": 0.832,
+      "step": 14245
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.76180405867238e-05,
+      "loss": 0.9434,
+      "step": 14246
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.760477086417604e-05,
+      "loss": 0.8756,
+      "step": 14247
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.759150241323247e-05,
+      "loss": 0.8503,
+      "step": 14248
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.757823523421519e-05,
+      "loss": 1.3556,
+      "step": 14249
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.756496932744613e-05,
+      "loss": 1.6923,
+      "step": 14250
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7551704693247236e-05,
+      "loss": 1.0676,
+      "step": 14251
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.753844133194049e-05,
+      "loss": 0.8987,
+      "step": 14252
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.752517924384774e-05,
+      "loss": 0.8549,
+      "step": 14253
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.751191842929084e-05,
+      "loss": 0.7768,
+      "step": 14254
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7498658888591675e-05,
+      "loss": 1.3688,
+      "step": 14255
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.748540062207202e-05,
+      "loss": 1.1317,
+      "step": 14256
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.747214363005365e-05,
+      "loss": 0.8321,
+      "step": 14257
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.745888791285829e-05,
+      "loss": 0.666,
+      "step": 14258
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.744563347080765e-05,
+      "loss": 0.7317,
+      "step": 14259
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.743238030422339e-05,
+      "loss": 0.8976,
+      "step": 14260
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.741912841342722e-05,
+      "loss": 0.8767,
+      "step": 14261
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7405877798740716e-05,
+      "loss": 0.9831,
+      "step": 14262
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.739262846048548e-05,
+      "loss": 0.9715,
+      "step": 14263
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7379380398983e-05,
+      "loss": 0.9078,
+      "step": 14264
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.73661336145549e-05,
+      "loss": 0.7628,
+      "step": 14265
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.735288810752262e-05,
+      "loss": 0.8892,
+      "step": 14266
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7339643878207596e-05,
+      "loss": 0.6945,
+      "step": 14267
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.732640092693132e-05,
+      "loss": 0.8216,
+      "step": 14268
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.731315925401516e-05,
+      "loss": 0.8223,
+      "step": 14269
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.729991885978045e-05,
+      "loss": 0.8656,
+      "step": 14270
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.728667974454859e-05,
+      "loss": 0.92,
+      "step": 14271
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7273441908640905e-05,
+      "loss": 0.8898,
+      "step": 14272
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.726020535237855e-05,
+      "loss": 0.8565,
+      "step": 14273
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.724697007608287e-05,
+      "loss": 0.7116,
+      "step": 14274
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7233736080075066e-05,
+      "loss": 0.9442,
+      "step": 14275
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.722050336467626e-05,
+      "loss": 0.8333,
+      "step": 14276
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7207271930207685e-05,
+      "loss": 0.8433,
+      "step": 14277
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.719404177699042e-05,
+      "loss": 0.8924,
+      "step": 14278
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.718081290534556e-05,
+      "loss": 0.9532,
+      "step": 14279
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.716758531559412e-05,
+      "loss": 0.9484,
+      "step": 14280
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7154359008057206e-05,
+      "loss": 0.8191,
+      "step": 14281
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.714113398305577e-05,
+      "loss": 0.8437,
+      "step": 14282
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.712791024091075e-05,
+      "loss": 0.8255,
+      "step": 14283
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7114687781943143e-05,
+      "loss": 0.8662,
+      "step": 14284
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.710146660647382e-05,
+      "loss": 1.0225,
+      "step": 14285
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.70882467148236e-05,
+      "loss": 0.7623,
+      "step": 14286
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7075028107313466e-05,
+      "loss": 0.9431,
+      "step": 14287
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7061810784264084e-05,
+      "loss": 0.8269,
+      "step": 14288
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.704859474599624e-05,
+      "loss": 0.7649,
+      "step": 14289
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.7035379992830766e-05,
+      "loss": 1.0078,
+      "step": 14290
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.702216652508832e-05,
+      "loss": 0.8372,
+      "step": 14291
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.700895434308956e-05,
+      "loss": 0.8686,
+      "step": 14292
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.699574344715522e-05,
+      "loss": 0.8181,
+      "step": 14293
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.698253383760588e-05,
+      "loss": 0.7141,
+      "step": 14294
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6969325514762074e-05,
+      "loss": 0.9603,
+      "step": 14295
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6956118478944465e-05,
+      "loss": 0.8603,
+      "step": 14296
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.694291273047352e-05,
+      "loss": 0.6683,
+      "step": 14297
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.692970826966976e-05,
+      "loss": 0.776,
+      "step": 14298
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.691650509685358e-05,
+      "loss": 0.8099,
+      "step": 14299
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.690330321234552e-05,
+      "loss": 0.7849,
+      "step": 14300
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.689010261646592e-05,
+      "loss": 0.9276,
+      "step": 14301
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6876903309535167e-05,
+      "loss": 1.0211,
+      "step": 14302
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6863705291873605e-05,
+      "loss": 0.8088,
+      "step": 14303
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.685050856380153e-05,
+      "loss": 0.7249,
+      "step": 14304
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.683731312563919e-05,
+      "loss": 0.8366,
+      "step": 14305
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6824118977706896e-05,
+      "loss": 0.8304,
+      "step": 14306
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.681092612032484e-05,
+      "loss": 0.8903,
+      "step": 14307
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6797734553813154e-05,
+      "loss": 0.779,
+      "step": 14308
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.678454427849208e-05,
+      "loss": 0.8792,
+      "step": 14309
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.677135529468169e-05,
+      "loss": 0.7327,
+      "step": 14310
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6758167602702044e-05,
+      "loss": 0.7293,
+      "step": 14311
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.674498120287327e-05,
+      "loss": 0.9064,
+      "step": 14312
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.673179609551536e-05,
+      "loss": 0.851,
+      "step": 14313
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.67186122809483e-05,
+      "loss": 1.0572,
+      "step": 14314
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6705429759492025e-05,
+      "loss": 0.682,
+      "step": 14315
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6692248531466585e-05,
+      "loss": 0.8643,
+      "step": 14316
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6679068597191734e-05,
+      "loss": 0.9672,
+      "step": 14317
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.666588995698743e-05,
+      "loss": 1.3493,
+      "step": 14318
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.66527126111735e-05,
+      "loss": 1.1625,
+      "step": 14319
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.663953656006974e-05,
+      "loss": 0.8254,
+      "step": 14320
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6626361803995886e-05,
+      "loss": 0.8832,
+      "step": 14321
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6613188343271774e-05,
+      "loss": 0.9697,
+      "step": 14322
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.660001617821705e-05,
+      "loss": 0.7313,
+      "step": 14323
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6586845309151395e-05,
+      "loss": 0.8707,
+      "step": 14324
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6573675736394496e-05,
+      "loss": 0.7431,
+      "step": 14325
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6560507460265955e-05,
+      "loss": 0.7117,
+      "step": 14326
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.654734048108532e-05,
+      "loss": 0.9308,
+      "step": 14327
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6534174799172226e-05,
+      "loss": 0.8725,
+      "step": 14328
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.652101041484615e-05,
+      "loss": 0.8925,
+      "step": 14329
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.650784732842658e-05,
+      "loss": 0.7928,
+      "step": 14330
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6494685540233e-05,
+      "loss": 0.9367,
+      "step": 14331
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.648152505058481e-05,
+      "loss": 0.8114,
+      "step": 14332
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.646836585980139e-05,
+      "loss": 0.7078,
+      "step": 14333
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.645520796820218e-05,
+      "loss": 0.9322,
+      "step": 14334
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.644205137610647e-05,
+      "loss": 0.8292,
+      "step": 14335
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.642889608383356e-05,
+      "loss": 0.7737,
+      "step": 14336
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6415742091702706e-05,
+      "loss": 0.9494,
+      "step": 14337
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.64025894000332e-05,
+      "loss": 0.8981,
+      "step": 14338
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6389438009144226e-05,
+      "loss": 0.9068,
+      "step": 14339
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.637628791935492e-05,
+      "loss": 0.7719,
+      "step": 14340
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6363139130984514e-05,
+      "loss": 0.8694,
+      "step": 14341
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6349991644352055e-05,
+      "loss": 0.8697,
+      "step": 14342
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6336845459776615e-05,
+      "loss": 0.8132,
+      "step": 14343
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.632370057757731e-05,
+      "loss": 0.8558,
+      "step": 14344
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.631055699807316e-05,
+      "loss": 0.8781,
+      "step": 14345
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6297414721583044e-05,
+      "loss": 0.8687,
+      "step": 14346
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.628427374842603e-05,
+      "loss": 0.8719,
+      "step": 14347
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6271134078921e-05,
+      "loss": 0.8038,
+      "step": 14348
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.625799571338682e-05,
+      "loss": 0.9926,
+      "step": 14349
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6244858652142396e-05,
+      "loss": 0.7634,
+      "step": 14350
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6231722895506556e-05,
+      "loss": 0.925,
+      "step": 14351
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.621858844379808e-05,
+      "loss": 0.8614,
+      "step": 14352
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.62054552973357e-05,
+      "loss": 0.8905,
+      "step": 14353
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.619232345643822e-05,
+      "loss": 0.8958,
+      "step": 14354
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.617919292142433e-05,
+      "loss": 0.8556,
+      "step": 14355
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6166063692612624e-05,
+      "loss": 0.7779,
+      "step": 14356
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.615293577032185e-05,
+      "loss": 0.852,
+      "step": 14357
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.613980915487056e-05,
+      "loss": 1.3101,
+      "step": 14358
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.612668384657729e-05,
+      "loss": 1.1892,
+      "step": 14359
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.611355984576071e-05,
+      "loss": 0.8752,
+      "step": 14360
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6100437152739215e-05,
+      "loss": 0.9161,
+      "step": 14361
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.60873157678313e-05,
+      "loss": 0.8579,
+      "step": 14362
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.607419569135545e-05,
+      "loss": 0.8745,
+      "step": 14363
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6061076923630075e-05,
+      "loss": 0.9302,
+      "step": 14364
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6047959464973524e-05,
+      "loss": 0.8687,
+      "step": 14365
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6034843315704215e-05,
+      "loss": 0.8545,
+      "step": 14366
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6021728476140425e-05,
+      "loss": 0.932,
+      "step": 14367
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.600861494660043e-05,
+      "loss": 0.7972,
+      "step": 14368
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.599550272740254e-05,
+      "loss": 0.9496,
+      "step": 14369
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.5982391818864966e-05,
+      "loss": 0.7305,
+      "step": 14370
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.596928222130589e-05,
+      "loss": 0.8397,
+      "step": 14371
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.5956173935043436e-05,
+      "loss": 0.8891,
+      "step": 14372
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.594306696039581e-05,
+      "loss": 1.4116,
+      "step": 14373
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.592996129768108e-05,
+      "loss": 1.1528,
+      "step": 14374
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.5916856947217314e-05,
+      "loss": 1.0189,
+      "step": 14375
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.5903753909322534e-05,
+      "loss": 1.3058,
+      "step": 14376
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.589065218431476e-05,
+      "loss": 1.15,
+      "step": 14377
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.587755177251192e-05,
+      "loss": 0.7153,
+      "step": 14378
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.586445267423203e-05,
+      "loss": 0.7718,
+      "step": 14379
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.585135488979295e-05,
+      "loss": 0.8897,
+      "step": 14380
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.5838258419512545e-05,
+      "loss": 0.9193,
+      "step": 14381
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.58251632637087e-05,
+      "loss": 0.9427,
+      "step": 14382
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.581206942269922e-05,
+      "loss": 0.8278,
+      "step": 14383
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.579897689680184e-05,
+      "loss": 0.8631,
+      "step": 14384
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.578588568633436e-05,
+      "loss": 0.7966,
+      "step": 14385
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.57727957916145e-05,
+      "loss": 0.9455,
+      "step": 14386
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.575970721295991e-05,
+      "loss": 0.8035,
+      "step": 14387
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.5746619950688217e-05,
+      "loss": 0.6755,
+      "step": 14388
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.573353400511716e-05,
+      "loss": 0.9474,
+      "step": 14389
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.572044937656418e-05,
+      "loss": 0.8182,
+      "step": 14390
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.570736606534692e-05,
+      "loss": 0.9213,
+      "step": 14391
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.569428407178291e-05,
+      "loss": 0.8439,
+      "step": 14392
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.568120339618961e-05,
+      "loss": 0.8183,
+      "step": 14393
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.566812403888446e-05,
+      "loss": 0.8835,
+      "step": 14394
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.565504600018495e-05,
+      "loss": 0.9058,
+      "step": 14395
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.564196928040846e-05,
+      "loss": 0.7744,
+      "step": 14396
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.562889387987229e-05,
+      "loss": 0.7911,
+      "step": 14397
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.561581979889388e-05,
+      "loss": 0.9222,
+      "step": 14398
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.560274703779046e-05,
+      "loss": 0.7149,
+      "step": 14399
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.558967559687929e-05,
+      "loss": 0.8653,
+      "step": 14400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.557660547647766e-05,
+      "loss": 0.8382,
+      "step": 14401
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.556353667690275e-05,
+      "loss": 0.9827,
+      "step": 14402
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.555046919847174e-05,
+      "loss": 1.3322,
+      "step": 14403
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5537403041501745e-05,
+      "loss": 1.0919,
+      "step": 14404
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.55243382063099e-05,
+      "loss": 0.8252,
+      "step": 14405
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5511274693213236e-05,
+      "loss": 0.8462,
+      "step": 14406
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5498212502528856e-05,
+      "loss": 0.7813,
+      "step": 14407
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.548515163457376e-05,
+      "loss": 0.8988,
+      "step": 14408
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.547209208966492e-05,
+      "loss": 0.9863,
+      "step": 14409
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.545903386811924e-05,
+      "loss": 0.8036,
+      "step": 14410
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.544597697025371e-05,
+      "loss": 0.8758,
+      "step": 14411
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.543292139638519e-05,
+      "loss": 0.8963,
+      "step": 14412
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5419867146830476e-05,
+      "loss": 0.8264,
+      "step": 14413
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.540681422190648e-05,
+      "loss": 0.8979,
+      "step": 14414
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5393762621929945e-05,
+      "loss": 0.8547,
+      "step": 14415
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.538071234721758e-05,
+      "loss": 0.9562,
+      "step": 14416
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.53676633980862e-05,
+      "loss": 0.751,
+      "step": 14417
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.535461577485245e-05,
+      "loss": 1.4668,
+      "step": 14418
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.534156947783298e-05,
+      "loss": 0.9926,
+      "step": 14419
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.532852450734444e-05,
+      "loss": 1.0719,
+      "step": 14420
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.53154808637034e-05,
+      "loss": 0.7843,
+      "step": 14421
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.530243854722639e-05,
+      "loss": 0.9365,
+      "step": 14422
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.528939755823003e-05,
+      "loss": 0.8581,
+      "step": 14423
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.527635789703076e-05,
+      "loss": 0.8244,
+      "step": 14424
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.526331956394506e-05,
+      "loss": 1.3303,
+      "step": 14425
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.525028255928932e-05,
+      "loss": 1.2488,
+      "step": 14426
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5237246883380026e-05,
+      "loss": 0.9035,
+      "step": 14427
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.52242125365335e-05,
+      "loss": 1.2802,
+      "step": 14428
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5211179519066026e-05,
+      "loss": 1.0233,
+      "step": 14429
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.519814783129401e-05,
+      "loss": 0.6195,
+      "step": 14430
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.518511747353369e-05,
+      "loss": 0.9364,
+      "step": 14431
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.517208844610123e-05,
+      "loss": 0.789,
+      "step": 14432
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5159060749312955e-05,
+      "loss": 0.8396,
+      "step": 14433
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5146034383485016e-05,
+      "loss": 0.8437,
+      "step": 14434
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5133009348933455e-05,
+      "loss": 0.8188,
+      "step": 14435
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5119985645974485e-05,
+      "loss": 0.7902,
+      "step": 14436
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.510696327492415e-05,
+      "loss": 0.9007,
+      "step": 14437
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.509394223609846e-05,
+      "loss": 0.7964,
+      "step": 14438
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.508092252981351e-05,
+      "loss": 0.7544,
+      "step": 14439
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5067904156385233e-05,
+      "loss": 0.7932,
+      "step": 14440
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.505488711612954e-05,
+      "loss": 0.9214,
+      "step": 14441
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.504187140936243e-05,
+      "loss": 0.889,
+      "step": 14442
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.502885703639975e-05,
+      "loss": 0.8429,
+      "step": 14443
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.5015843997557364e-05,
+      "loss": 0.7788,
+      "step": 14444
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.500283229315102e-05,
+      "loss": 0.9276,
+      "step": 14445
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.49898219234966e-05,
+      "loss": 0.8491,
+      "step": 14446
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.497681288890983e-05,
+      "loss": 0.8705,
+      "step": 14447
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4963805189706424e-05,
+      "loss": 0.903,
+      "step": 14448
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4950798826202054e-05,
+      "loss": 0.9309,
+      "step": 14449
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.493779379871241e-05,
+      "loss": 0.8568,
+      "step": 14450
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.492479010755306e-05,
+      "loss": 0.9832,
+      "step": 14451
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4911787753039683e-05,
+      "loss": 0.8193,
+      "step": 14452
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.489878673548779e-05,
+      "loss": 0.8714,
+      "step": 14453
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.488578705521287e-05,
+      "loss": 0.7802,
+      "step": 14454
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.48727887125305e-05,
+      "loss": 0.8738,
+      "step": 14455
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.48597917077561e-05,
+      "loss": 0.8642,
+      "step": 14456
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.484679604120508e-05,
+      "loss": 0.853,
+      "step": 14457
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.483380171319289e-05,
+      "loss": 0.7919,
+      "step": 14458
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.482080872403487e-05,
+      "loss": 0.8575,
+      "step": 14459
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4807817074046354e-05,
+      "loss": 0.9567,
+      "step": 14460
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4794826763542595e-05,
+      "loss": 0.914,
+      "step": 14461
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.478183779283895e-05,
+      "loss": 0.91,
+      "step": 14462
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4768850162250605e-05,
+      "loss": 0.8601,
+      "step": 14463
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.475586387209277e-05,
+      "loss": 1.0512,
+      "step": 14464
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.474287892268062e-05,
+      "loss": 0.8936,
+      "step": 14465
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.472989531432927e-05,
+      "loss": 0.8691,
+      "step": 14466
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.471691304735381e-05,
+      "loss": 0.7806,
+      "step": 14467
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.470393212206938e-05,
+      "loss": 0.7885,
+      "step": 14468
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.469095253879099e-05,
+      "loss": 0.9189,
+      "step": 14469
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.46779742978336e-05,
+      "loss": 0.9049,
+      "step": 14470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.466499739951226e-05,
+      "loss": 0.8391,
+      "step": 14471
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.465202184414188e-05,
+      "loss": 0.913,
+      "step": 14472
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4639047632037336e-05,
+      "loss": 0.7278,
+      "step": 14473
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.462607476351357e-05,
+      "loss": 0.7993,
+      "step": 14474
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.461310323888539e-05,
+      "loss": 0.7832,
+      "step": 14475
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.460013305846762e-05,
+      "loss": 0.9268,
+      "step": 14476
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4587164222574984e-05,
+      "loss": 0.8917,
+      "step": 14477
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4574196731522366e-05,
+      "loss": 0.8868,
+      "step": 14478
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.45612305856243e-05,
+      "loss": 0.7329,
+      "step": 14479
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.454826578519561e-05,
+      "loss": 0.9304,
+      "step": 14480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.453530233055088e-05,
+      "loss": 0.9366,
+      "step": 14481
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.452234022200474e-05,
+      "loss": 0.957,
+      "step": 14482
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.450937945987174e-05,
+      "loss": 0.8962,
+      "step": 14483
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.449642004446649e-05,
+      "loss": 0.8585,
+      "step": 14484
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.448346197610349e-05,
+      "loss": 0.9093,
+      "step": 14485
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.447050525509719e-05,
+      "loss": 0.9456,
+      "step": 14486
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4457549881762084e-05,
+      "loss": 0.9243,
+      "step": 14487
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.444459585641259e-05,
+      "loss": 0.7585,
+      "step": 14488
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4431643179363036e-05,
+      "loss": 0.9367,
+      "step": 14489
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.441869185092786e-05,
+      "loss": 0.9337,
+      "step": 14490
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4405741871421356e-05,
+      "loss": 0.8424,
+      "step": 14491
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4392793241157795e-05,
+      "loss": 1.2589,
+      "step": 14492
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4379845960451446e-05,
+      "loss": 1.1179,
+      "step": 14493
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.436690002961653e-05,
+      "loss": 0.8989,
+      "step": 14494
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.43539554489672e-05,
+      "loss": 0.7044,
+      "step": 14495
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.434101221881768e-05,
+      "loss": 0.8944,
+      "step": 14496
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4328070339482075e-05,
+      "loss": 0.8283,
+      "step": 14497
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4315129811274426e-05,
+      "loss": 0.8058,
+      "step": 14498
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.430219063450887e-05,
+      "loss": 0.8656,
+      "step": 14499
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.4289252809499405e-05,
+      "loss": 0.8333,
+      "step": 14500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.427631633656002e-05,
+      "loss": 0.8335,
+      "step": 14501
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.426338121600463e-05,
+      "loss": 0.7635,
+      "step": 14502
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.425044744814725e-05,
+      "loss": 0.855,
+      "step": 14503
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.4237515033301736e-05,
+      "loss": 0.8105,
+      "step": 14504
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.422458397178191e-05,
+      "loss": 0.8264,
+      "step": 14505
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.421165426390167e-05,
+      "loss": 0.9648,
+      "step": 14506
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.4198725909974837e-05,
+      "loss": 0.7254,
+      "step": 14507
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.418579891031504e-05,
+      "loss": 0.8484,
+      "step": 14508
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.417287326523613e-05,
+      "loss": 0.8137,
+      "step": 14509
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.415994897505177e-05,
+      "loss": 0.7605,
+      "step": 14510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.414702604007559e-05,
+      "loss": 0.9924,
+      "step": 14511
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.4134104460621286e-05,
+      "loss": 0.8272,
+      "step": 14512
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.412118423700243e-05,
+      "loss": 0.9245,
+      "step": 14513
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.4108265369532553e-05,
+      "loss": 1.2438,
+      "step": 14514
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.409534785852526e-05,
+      "loss": 1.0541,
+      "step": 14515
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.408243170429401e-05,
+      "loss": 1.2773,
+      "step": 14516
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.406951690715229e-05,
+      "loss": 1.0273,
+      "step": 14517
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.405660346741347e-05,
+      "loss": 0.8105,
+      "step": 14518
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.404369138539105e-05,
+      "loss": 0.7389,
+      "step": 14519
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.403078066139835e-05,
+      "loss": 1.0352,
+      "step": 14520
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.401787129574868e-05,
+      "loss": 0.7799,
+      "step": 14521
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.400496328875544e-05,
+      "loss": 0.9508,
+      "step": 14522
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3992056640731806e-05,
+      "loss": 0.945,
+      "step": 14523
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3979151351991e-05,
+      "loss": 0.7913,
+      "step": 14524
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.396624742284631e-05,
+      "loss": 0.9228,
+      "step": 14525
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3953344853610855e-05,
+      "loss": 0.8696,
+      "step": 14526
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.394044364459776e-05,
+      "loss": 0.8984,
+      "step": 14527
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.392754379612018e-05,
+      "loss": 0.893,
+      "step": 14528
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3914645308491186e-05,
+      "loss": 0.8993,
+      "step": 14529
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.390174818202374e-05,
+      "loss": 1.0227,
+      "step": 14530
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.388885241703095e-05,
+      "loss": 0.9552,
+      "step": 14531
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3875958013825736e-05,
+      "loss": 0.9125,
+      "step": 14532
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.386306497272105e-05,
+      "loss": 0.8757,
+      "step": 14533
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.385017329402975e-05,
+      "loss": 1.2343,
+      "step": 14534
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3837282978064796e-05,
+      "loss": 1.1879,
+      "step": 14535
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.382439402513899e-05,
+      "loss": 0.7976,
+      "step": 14536
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.381150643556513e-05,
+      "loss": 0.8116,
+      "step": 14537
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3798620209656e-05,
+      "loss": 0.8345,
+      "step": 14538
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3785735347724344e-05,
+      "loss": 0.9682,
+      "step": 14539
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.377285185008283e-05,
+      "loss": 0.8617,
+      "step": 14540
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.375996971704421e-05,
+      "loss": 0.8594,
+      "step": 14541
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.374708894892108e-05,
+      "loss": 0.8907,
+      "step": 14542
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.373420954602603e-05,
+      "loss": 0.9017,
+      "step": 14543
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.372133150867168e-05,
+      "loss": 0.7885,
+      "step": 14544
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3708454837170574e-05,
+      "loss": 0.866,
+      "step": 14545
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.369557953183516e-05,
+      "loss": 0.8066,
+      "step": 14546
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.368270559297799e-05,
+      "loss": 0.8168,
+      "step": 14547
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3669833020911485e-05,
+      "loss": 0.9084,
+      "step": 14548
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3656961815948036e-05,
+      "loss": 0.7853,
+      "step": 14549
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3644091978400005e-05,
+      "loss": 0.7126,
+      "step": 14550
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.363122350857983e-05,
+      "loss": 0.8054,
+      "step": 14551
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3618356406799684e-05,
+      "loss": 0.7298,
+      "step": 14552
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.360549067337195e-05,
+      "loss": 0.9184,
+      "step": 14553
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.359262630860883e-05,
+      "loss": 0.9031,
+      "step": 14554
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3579763312822553e-05,
+      "loss": 0.6902,
+      "step": 14555
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.356690168632524e-05,
+      "loss": 1.2117,
+      "step": 14556
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.355404142942912e-05,
+      "loss": 1.047,
+      "step": 14557
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.354118254244627e-05,
+      "loss": 0.9626,
+      "step": 14558
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3528325025688723e-05,
+      "loss": 0.7837,
+      "step": 14559
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.35154688794686e-05,
+      "loss": 0.7816,
+      "step": 14560
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3502614104097894e-05,
+      "loss": 0.9612,
+      "step": 14561
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3489760699888514e-05,
+      "loss": 0.843,
+      "step": 14562
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.34769086671525e-05,
+      "loss": 0.8509,
+      "step": 14563
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.346405800620173e-05,
+      "loss": 0.9429,
+      "step": 14564
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.345120871734808e-05,
+      "loss": 0.8745,
+      "step": 14565
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.343836080090339e-05,
+      "loss": 0.8145,
+      "step": 14566
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3425514257179455e-05,
+      "loss": 0.768,
+      "step": 14567
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3412669086488055e-05,
+      "loss": 0.7664,
+      "step": 14568
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3399825289140984e-05,
+      "loss": 0.9096,
+      "step": 14569
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3386982865449924e-05,
+      "loss": 0.8743,
+      "step": 14570
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.337414181572651e-05,
+      "loss": 0.9012,
+      "step": 14571
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.336130214028248e-05,
+      "loss": 0.8399,
+      "step": 14572
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3348463839429375e-05,
+      "loss": 0.9286,
+      "step": 14573
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.333562691347881e-05,
+      "loss": 0.9698,
+      "step": 14574
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.332279136274228e-05,
+      "loss": 0.9096,
+      "step": 14575
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.330995718753136e-05,
+      "loss": 0.8489,
+      "step": 14576
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3297124388157505e-05,
+      "loss": 0.9274,
+      "step": 14577
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3284292964932113e-05,
+      "loss": 1.0339,
+      "step": 14578
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.327146291816668e-05,
+      "loss": 0.9098,
+      "step": 14579
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.325863424817259e-05,
+      "loss": 0.7973,
+      "step": 14580
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.324580695526106e-05,
+      "loss": 0.7843,
+      "step": 14581
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.323298103974352e-05,
+      "loss": 0.751,
+      "step": 14582
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3220156501931195e-05,
+      "loss": 0.8544,
+      "step": 14583
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.320733334213532e-05,
+      "loss": 0.932,
+      "step": 14584
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.319451156066717e-05,
+      "loss": 0.769,
+      "step": 14585
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.318169115783789e-05,
+      "loss": 0.9189,
+      "step": 14586
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.316887213395858e-05,
+      "loss": 0.8891,
+      "step": 14587
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3156054489340426e-05,
+      "loss": 0.862,
+      "step": 14588
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3143238224294454e-05,
+      "loss": 0.9103,
+      "step": 14589
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3130423339131744e-05,
+      "loss": 0.8363,
+      "step": 14590
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.311760983416323e-05,
+      "loss": 0.9422,
+      "step": 14591
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.31047977097e-05,
+      "loss": 0.7512,
+      "step": 14592
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.309198696605292e-05,
+      "loss": 0.8592,
+      "step": 14593
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3079177603532906e-05,
+      "loss": 0.893,
+      "step": 14594
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.306636962245091e-05,
+      "loss": 0.8301,
+      "step": 14595
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.305356302311768e-05,
+      "loss": 0.7868,
+      "step": 14596
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3040757805844024e-05,
+      "loss": 0.7838,
+      "step": 14597
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3027953970940796e-05,
+      "loss": 0.7427,
+      "step": 14598
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.301515151871869e-05,
+      "loss": 0.9047,
+      "step": 14599
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.3002350449488384e-05,
+      "loss": 0.9693,
+      "step": 14600
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.2989550763560624e-05,
+      "loss": 1.2623,
+      "step": 14601
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.2976752461246025e-05,
+      "loss": 1.1549,
+      "step": 14602
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.296395554285515e-05,
+      "loss": 0.8666,
+      "step": 14603
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.295116000869864e-05,
+      "loss": 0.9078,
+      "step": 14604
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.293836585908702e-05,
+      "loss": 0.9303,
+      "step": 14605
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2925573094330773e-05,
+      "loss": 0.9026,
+      "step": 14606
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.291278171474035e-05,
+      "loss": 0.7898,
+      "step": 14607
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.289999172062625e-05,
+      "loss": 0.7886,
+      "step": 14608
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.288720311229887e-05,
+      "loss": 0.8182,
+      "step": 14609
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.287441589006856e-05,
+      "loss": 0.84,
+      "step": 14610
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.286163005424566e-05,
+      "loss": 0.7793,
+      "step": 14611
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.284884560514049e-05,
+      "loss": 0.8898,
+      "step": 14612
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.283606254306327e-05,
+      "loss": 0.7762,
+      "step": 14613
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2823280868324326e-05,
+      "loss": 0.7673,
+      "step": 14614
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2810500581233805e-05,
+      "loss": 1.3565,
+      "step": 14615
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.279772168210185e-05,
+      "loss": 1.1346,
+      "step": 14616
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2784944171238686e-05,
+      "loss": 0.7459,
+      "step": 14617
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.277216804895437e-05,
+      "loss": 0.7939,
+      "step": 14618
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.275939331555893e-05,
+      "loss": 0.92,
+      "step": 14619
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.274661997136247e-05,
+      "loss": 0.8285,
+      "step": 14620
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2733848016674974e-05,
+      "loss": 0.864,
+      "step": 14621
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.272107745180639e-05,
+      "loss": 0.9267,
+      "step": 14622
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.270830827706663e-05,
+      "loss": 1.2358,
+      "step": 14623
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2695540492765673e-05,
+      "loss": 0.9902,
+      "step": 14624
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.268277409921333e-05,
+      "loss": 0.7072,
+      "step": 14625
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2670009096719454e-05,
+      "loss": 0.7737,
+      "step": 14626
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.265724548559384e-05,
+      "loss": 0.77,
+      "step": 14627
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.264448326614625e-05,
+      "loss": 0.9567,
+      "step": 14628
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2631722438686374e-05,
+      "loss": 0.753,
+      "step": 14629
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2618963003523994e-05,
+      "loss": 0.708,
+      "step": 14630
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2606204960968734e-05,
+      "loss": 0.8171,
+      "step": 14631
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.259344831133019e-05,
+      "loss": 0.8659,
+      "step": 14632
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.258069305491803e-05,
+      "loss": 0.7332,
+      "step": 14633
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2567939192041786e-05,
+      "loss": 0.8121,
+      "step": 14634
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.255518672301093e-05,
+      "loss": 1.3777,
+      "step": 14635
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.254243564813507e-05,
+      "loss": 1.0733,
+      "step": 14636
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2529685967723595e-05,
+      "loss": 0.8907,
+      "step": 14637
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.251693768208591e-05,
+      "loss": 0.9446,
+      "step": 14638
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.250419079153152e-05,
+      "loss": 1.0375,
+      "step": 14639
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.249144529636968e-05,
+      "loss": 0.7695,
+      "step": 14640
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.247870119690971e-05,
+      "loss": 0.9106,
+      "step": 14641
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2465958493460966e-05,
+      "loss": 0.7851,
+      "step": 14642
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.245321718633268e-05,
+      "loss": 0.8902,
+      "step": 14643
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2440477275834055e-05,
+      "loss": 0.8228,
+      "step": 14644
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.242773876227433e-05,
+      "loss": 0.9115,
+      "step": 14645
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2415001645962636e-05,
+      "loss": 0.8856,
+      "step": 14646
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2402265927208095e-05,
+      "loss": 0.9828,
+      "step": 14647
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2389531606319766e-05,
+      "loss": 0.7923,
+      "step": 14648
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.237679868360677e-05,
+      "loss": 0.8129,
+      "step": 14649
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2364067159378085e-05,
+      "loss": 0.7639,
+      "step": 14650
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.235133703394269e-05,
+      "loss": 0.8588,
+      "step": 14651
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.233860830760956e-05,
+      "loss": 0.8282,
+      "step": 14652
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.232588098068764e-05,
+      "loss": 0.864,
+      "step": 14653
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2313155053485764e-05,
+      "loss": 0.7788,
+      "step": 14654
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.230043052631281e-05,
+      "loss": 0.9462,
+      "step": 14655
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2287707399477595e-05,
+      "loss": 0.8007,
+      "step": 14656
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.227498567328885e-05,
+      "loss": 0.7625,
+      "step": 14657
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.226226534805542e-05,
+      "loss": 0.776,
+      "step": 14658
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.224954642408596e-05,
+      "loss": 0.9823,
+      "step": 14659
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.223682890168914e-05,
+      "loss": 1.3391,
+      "step": 14660
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.222411278117365e-05,
+      "loss": 1.0799,
+      "step": 14661
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.22113980628481e-05,
+      "loss": 0.8888,
+      "step": 14662
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2198684747021046e-05,
+      "loss": 0.9549,
+      "step": 14663
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.218597283400101e-05,
+      "loss": 0.8396,
+      "step": 14664
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.217326232409657e-05,
+      "loss": 0.9512,
+      "step": 14665
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.216055321761616e-05,
+      "loss": 0.9122,
+      "step": 14666
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.214784551486821e-05,
+      "loss": 0.8022,
+      "step": 14667
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.213513921616118e-05,
+      "loss": 0.8296,
+      "step": 14668
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.212243432180345e-05,
+      "loss": 1.1106,
+      "step": 14669
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.210973083210324e-05,
+      "loss": 0.8986,
+      "step": 14670
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.209702874736901e-05,
+      "loss": 0.8503,
+      "step": 14671
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.208432806790894e-05,
+      "loss": 0.7658,
+      "step": 14672
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.207162879403127e-05,
+      "loss": 0.9214,
+      "step": 14673
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.205893092604426e-05,
+      "loss": 0.8754,
+      "step": 14674
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.204623446425605e-05,
+      "loss": 0.7617,
+      "step": 14675
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2033539408974744e-05,
+      "loss": 0.8726,
+      "step": 14676
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.202084576050851e-05,
+      "loss": 0.8409,
+      "step": 14677
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.2008153519165386e-05,
+      "loss": 0.9834,
+      "step": 14678
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.199546268525339e-05,
+      "loss": 0.9111,
+      "step": 14679
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1982773259080524e-05,
+      "loss": 0.8753,
+      "step": 14680
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1970085240954784e-05,
+      "loss": 0.7331,
+      "step": 14681
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.195739863118408e-05,
+      "loss": 0.8413,
+      "step": 14682
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.194471343007631e-05,
+      "loss": 1.0034,
+      "step": 14683
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.193202963793934e-05,
+      "loss": 0.8699,
+      "step": 14684
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.191934725508099e-05,
+      "loss": 0.9103,
+      "step": 14685
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1906666281809024e-05,
+      "loss": 0.9002,
+      "step": 14686
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1893986718431286e-05,
+      "loss": 0.886,
+      "step": 14687
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1881308565255454e-05,
+      "loss": 0.8755,
+      "step": 14688
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.186863182258919e-05,
+      "loss": 0.9047,
+      "step": 14689
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1855956490740214e-05,
+      "loss": 0.9511,
+      "step": 14690
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1843282570016104e-05,
+      "loss": 0.9076,
+      "step": 14691
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.183061006072445e-05,
+      "loss": 0.8268,
+      "step": 14692
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.181793896317284e-05,
+      "loss": 0.9426,
+      "step": 14693
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.180526927766879e-05,
+      "loss": 0.8199,
+      "step": 14694
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1792601004519774e-05,
+      "loss": 0.863,
+      "step": 14695
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.177993414403319e-05,
+      "loss": 0.8285,
+      "step": 14696
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1767268696516556e-05,
+      "loss": 0.8429,
+      "step": 14697
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1754604662277205e-05,
+      "loss": 0.8268,
+      "step": 14698
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.17419420416225e-05,
+      "loss": 0.9063,
+      "step": 14699
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.172928083485974e-05,
+      "loss": 1.299,
+      "step": 14700
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.171662104229617e-05,
+      "loss": 1.2167,
+      "step": 14701
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.170396266423911e-05,
+      "loss": 0.7642,
+      "step": 14702
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.169130570099575e-05,
+      "loss": 0.8634,
+      "step": 14703
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.167865015287325e-05,
+      "loss": 0.7551,
+      "step": 14704
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1665996020178744e-05,
+      "loss": 0.6971,
+      "step": 14705
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.165334330321937e-05,
+      "loss": 0.7253,
+      "step": 14706
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.164069200230221e-05,
+      "loss": 0.9074,
+      "step": 14707
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1628042117734236e-05,
+      "loss": 0.7978,
+      "step": 14708
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.1615393649822544e-05,
+      "loss": 0.9459,
+      "step": 14709
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1602746598874066e-05,
+      "loss": 1.2804,
+      "step": 14710
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.15901009651957e-05,
+      "loss": 1.1569,
+      "step": 14711
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.157745674909442e-05,
+      "loss": 0.8531,
+      "step": 14712
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.15648139508771e-05,
+      "loss": 0.864,
+      "step": 14713
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.155217257085047e-05,
+      "loss": 0.9363,
+      "step": 14714
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1539532609321416e-05,
+      "loss": 0.9489,
+      "step": 14715
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1526894066596685e-05,
+      "loss": 0.8583,
+      "step": 14716
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.151425694298298e-05,
+      "loss": 0.7211,
+      "step": 14717
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1501621238787045e-05,
+      "loss": 0.8432,
+      "step": 14718
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.148898695431552e-05,
+      "loss": 0.8708,
+      "step": 14719
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1476354089875036e-05,
+      "loss": 0.7593,
+      "step": 14720
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.146372264577214e-05,
+      "loss": 0.9035,
+      "step": 14721
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.145109262231347e-05,
+      "loss": 0.9434,
+      "step": 14722
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.14384640198055e-05,
+      "loss": 0.9785,
+      "step": 14723
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.142583683855471e-05,
+      "loss": 1.3502,
+      "step": 14724
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.14132110788676e-05,
+      "loss": 1.2197,
+      "step": 14725
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.140058674105057e-05,
+      "loss": 0.9477,
+      "step": 14726
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.138796382541e-05,
+      "loss": 0.7758,
+      "step": 14727
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1375342332252256e-05,
+      "loss": 0.8785,
+      "step": 14728
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1362722261883624e-05,
+      "loss": 0.7835,
+      "step": 14729
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.135010361461037e-05,
+      "loss": 0.7406,
+      "step": 14730
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.133748639073882e-05,
+      "loss": 1.0342,
+      "step": 14731
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1324870590575134e-05,
+      "loss": 0.9158,
+      "step": 14732
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.131225621442548e-05,
+      "loss": 0.8682,
+      "step": 14733
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.129964326259604e-05,
+      "loss": 0.8261,
+      "step": 14734
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1287031735392925e-05,
+      "loss": 0.8755,
+      "step": 14735
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.127442163312217e-05,
+      "loss": 0.8927,
+      "step": 14736
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.126181295608982e-05,
+      "loss": 0.8879,
+      "step": 14737
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.124920570460192e-05,
+      "loss": 0.8714,
+      "step": 14738
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1236599878964424e-05,
+      "loss": 0.7528,
+      "step": 14739
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1223995479483224e-05,
+      "loss": 0.8617,
+      "step": 14740
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.12113925064643e-05,
+      "loss": 1.3302,
+      "step": 14741
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.119879096021352e-05,
+      "loss": 1.0899,
+      "step": 14742
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.11861908410366e-05,
+      "loss": 0.8733,
+      "step": 14743
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1173592149239446e-05,
+      "loss": 0.7547,
+      "step": 14744
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1160994885127805e-05,
+      "loss": 0.8904,
+      "step": 14745
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1148399049007345e-05,
+      "loss": 0.803,
+      "step": 14746
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.113580464118385e-05,
+      "loss": 0.9268,
+      "step": 14747
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.112321166196293e-05,
+      "loss": 0.8125,
+      "step": 14748
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.111062011165019e-05,
+      "loss": 0.847,
+      "step": 14749
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.109802999055127e-05,
+      "loss": 0.7519,
+      "step": 14750
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1085441298971714e-05,
+      "loss": 0.9829,
+      "step": 14751
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.107285403721702e-05,
+      "loss": 0.8153,
+      "step": 14752
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.106026820559265e-05,
+      "loss": 0.9946,
+      "step": 14753
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.104768380440415e-05,
+      "loss": 0.8727,
+      "step": 14754
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.1035100833956855e-05,
+      "loss": 0.927,
+      "step": 14755
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.102251929455614e-05,
+      "loss": 0.9536,
+      "step": 14756
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.100993918650744e-05,
+      "loss": 0.7269,
+      "step": 14757
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.099736051011599e-05,
+      "loss": 0.8916,
+      "step": 14758
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0984783265687044e-05,
+      "loss": 0.7711,
+      "step": 14759
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.097220745352592e-05,
+      "loss": 0.8445,
+      "step": 14760
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.095963307393781e-05,
+      "loss": 0.8312,
+      "step": 14761
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.094706012722782e-05,
+      "loss": 0.9037,
+      "step": 14762
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.093448861370118e-05,
+      "loss": 0.8402,
+      "step": 14763
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.092191853366296e-05,
+      "loss": 0.763,
+      "step": 14764
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.090934988741818e-05,
+      "loss": 0.7388,
+      "step": 14765
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.089678267527196e-05,
+      "loss": 0.9676,
+      "step": 14766
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.088421689752925e-05,
+      "loss": 0.7303,
+      "step": 14767
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.087165255449502e-05,
+      "loss": 0.9038,
+      "step": 14768
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.085908964647418e-05,
+      "loss": 0.9492,
+      "step": 14769
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.084652817377167e-05,
+      "loss": 0.8206,
+      "step": 14770
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0833968136692335e-05,
+      "loss": 0.7805,
+      "step": 14771
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.082140953554099e-05,
+      "loss": 0.8563,
+      "step": 14772
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.080885237062242e-05,
+      "loss": 0.8647,
+      "step": 14773
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0796296642241363e-05,
+      "loss": 0.8391,
+      "step": 14774
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.078374235070259e-05,
+      "loss": 0.914,
+      "step": 14775
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.077118949631078e-05,
+      "loss": 0.8771,
+      "step": 14776
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.075863807937056e-05,
+      "loss": 0.7653,
+      "step": 14777
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.074608810018651e-05,
+      "loss": 0.9088,
+      "step": 14778
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.073353955906329e-05,
+      "loss": 0.8339,
+      "step": 14779
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.072099245630542e-05,
+      "loss": 0.7251,
+      "step": 14780
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0708446792217356e-05,
+      "loss": 0.8311,
+      "step": 14781
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.069590256710366e-05,
+      "loss": 0.9771,
+      "step": 14782
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.068335978126874e-05,
+      "loss": 1.4173,
+      "step": 14783
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.067081843501696e-05,
+      "loss": 1.1129,
+      "step": 14784
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.065827852865275e-05,
+      "loss": 0.8795,
+      "step": 14785
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0645740062480476e-05,
+      "loss": 0.8471,
+      "step": 14786
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.063320303680431e-05,
+      "loss": 0.7321,
+      "step": 14787
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.062066745192864e-05,
+      "loss": 1.1712,
+      "step": 14788
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0608133308157646e-05,
+      "loss": 1.0654,
+      "step": 14789
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0595600605795506e-05,
+      "loss": 1.3462,
+      "step": 14790
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0583069345146454e-05,
+      "loss": 1.1089,
+      "step": 14791
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.057053952651456e-05,
+      "loss": 0.7622,
+      "step": 14792
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.055801115020394e-05,
+      "loss": 0.9336,
+      "step": 14793
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0545484216518606e-05,
+      "loss": 0.7943,
+      "step": 14794
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.053295872576265e-05,
+      "loss": 0.9455,
+      "step": 14795
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0520434678240015e-05,
+      "loss": 0.9559,
+      "step": 14796
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.050791207425464e-05,
+      "loss": 0.7779,
+      "step": 14797
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.049539091411049e-05,
+      "loss": 0.8187,
+      "step": 14798
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.048287119811143e-05,
+      "loss": 0.8445,
+      "step": 14799
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.047035292656125e-05,
+      "loss": 0.9583,
+      "step": 14800
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.045783609976389e-05,
+      "loss": 0.8833,
+      "step": 14801
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0445320718023004e-05,
+      "loss": 0.9413,
+      "step": 14802
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.043280678164234e-05,
+      "loss": 0.7523,
+      "step": 14803
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.042029429092568e-05,
+      "loss": 0.7527,
+      "step": 14804
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.040778324617666e-05,
+      "loss": 0.7872,
+      "step": 14805
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.039527364769886e-05,
+      "loss": 0.8677,
+      "step": 14806
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0382765495795984e-05,
+      "loss": 0.8344,
+      "step": 14807
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.037025879077154e-05,
+      "loss": 0.8623,
+      "step": 14808
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0357753532929075e-05,
+      "loss": 0.8087,
+      "step": 14809
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.034524972257204e-05,
+      "loss": 0.9455,
+      "step": 14810
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.033274736000396e-05,
+      "loss": 0.9092,
+      "step": 14811
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0320246445528244e-05,
+      "loss": 1.0201,
+      "step": 14812
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.0307746979448225e-05,
+      "loss": 1.47,
+      "step": 14813
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0295248962067345e-05,
+      "loss": 1.1091,
+      "step": 14814
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.028275239368888e-05,
+      "loss": 0.8476,
+      "step": 14815
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.027025727461613e-05,
+      "loss": 0.7802,
+      "step": 14816
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.025776360515232e-05,
+      "loss": 0.9147,
+      "step": 14817
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.024527138560069e-05,
+      "loss": 0.8038,
+      "step": 14818
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0232780616264375e-05,
+      "loss": 0.7931,
+      "step": 14819
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.022029129744658e-05,
+      "loss": 0.8327,
+      "step": 14820
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0207803429450386e-05,
+      "loss": 0.8511,
+      "step": 14821
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.019531701257884e-05,
+      "loss": 0.9442,
+      "step": 14822
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.018283204713503e-05,
+      "loss": 0.7845,
+      "step": 14823
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.017034853342194e-05,
+      "loss": 0.7196,
+      "step": 14824
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.015786647174254e-05,
+      "loss": 0.8142,
+      "step": 14825
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0145385862399733e-05,
+      "loss": 1.0026,
+      "step": 14826
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0132906705696474e-05,
+      "loss": 0.7689,
+      "step": 14827
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.012042900193559e-05,
+      "loss": 0.8517,
+      "step": 14828
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.010795275141987e-05,
+      "loss": 0.8535,
+      "step": 14829
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.009547795445222e-05,
+      "loss": 0.8876,
+      "step": 14830
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.008300461133526e-05,
+      "loss": 0.8471,
+      "step": 14831
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.007053272237181e-05,
+      "loss": 0.9042,
+      "step": 14832
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.005806228786453e-05,
+      "loss": 0.9155,
+      "step": 14833
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.004559330811606e-05,
+      "loss": 0.751,
+      "step": 14834
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.003312578342897e-05,
+      "loss": 0.8356,
+      "step": 14835
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0020659714105935e-05,
+      "loss": 0.8963,
+      "step": 14836
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.0008195100449455e-05,
+      "loss": 0.8452,
+      "step": 14837
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9995731942761994e-05,
+      "loss": 0.695,
+      "step": 14838
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.998327024134611e-05,
+      "loss": 0.8627,
+      "step": 14839
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.997080999650421e-05,
+      "loss": 0.8053,
+      "step": 14840
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9958351208538644e-05,
+      "loss": 0.8239,
+      "step": 14841
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.994589387775186e-05,
+      "loss": 0.912,
+      "step": 14842
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.993343800444617e-05,
+      "loss": 0.8652,
+      "step": 14843
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.992098358892384e-05,
+      "loss": 0.867,
+      "step": 14844
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.990853063148716e-05,
+      "loss": 0.9651,
+      "step": 14845
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9896079132438346e-05,
+      "loss": 0.9468,
+      "step": 14846
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9883629092079554e-05,
+      "loss": 0.8578,
+      "step": 14847
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9871180510713004e-05,
+      "loss": 0.7009,
+      "step": 14848
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.98587333886408e-05,
+      "loss": 0.9572,
+      "step": 14849
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.984628772616501e-05,
+      "loss": 0.8739,
+      "step": 14850
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9833843523587645e-05,
+      "loss": 0.9548,
+      "step": 14851
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.982140078121081e-05,
+      "loss": 0.84,
+      "step": 14852
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.980895949933643e-05,
+      "loss": 1.0497,
+      "step": 14853
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.979651967826642e-05,
+      "loss": 0.757,
+      "step": 14854
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9784081318302756e-05,
+      "loss": 0.813,
+      "step": 14855
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9771644419747276e-05,
+      "loss": 0.7366,
+      "step": 14856
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.975920898290179e-05,
+      "loss": 0.8099,
+      "step": 14857
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.974677500806816e-05,
+      "loss": 0.8848,
+      "step": 14858
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.973434249554811e-05,
+      "loss": 0.9348,
+      "step": 14859
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9721911445643387e-05,
+      "loss": 0.9113,
+      "step": 14860
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.970948185865568e-05,
+      "loss": 0.8065,
+      "step": 14861
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.969705373488665e-05,
+      "loss": 0.8943,
+      "step": 14862
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.968462707463787e-05,
+      "loss": 0.7953,
+      "step": 14863
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.967220187821102e-05,
+      "loss": 0.7783,
+      "step": 14864
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9659778145907614e-05,
+      "loss": 1.057,
+      "step": 14865
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.964735587802915e-05,
+      "loss": 0.7263,
+      "step": 14866
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.96349350748771e-05,
+      "loss": 1.1463,
+      "step": 14867
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.962251573675296e-05,
+      "loss": 1.1364,
+      "step": 14868
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.961009786395813e-05,
+      "loss": 0.9352,
+      "step": 14869
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.959768145679392e-05,
+      "loss": 0.8719,
+      "step": 14870
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.958526651556177e-05,
+      "loss": 0.9421,
+      "step": 14871
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.957285304056292e-05,
+      "loss": 0.8351,
+      "step": 14872
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9560441032098626e-05,
+      "loss": 0.8365,
+      "step": 14873
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.954803049047021e-05,
+      "loss": 0.8737,
+      "step": 14874
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9535621415978774e-05,
+      "loss": 0.8502,
+      "step": 14875
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.952321380892549e-05,
+      "loss": 0.8691,
+      "step": 14876
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.951080766961153e-05,
+      "loss": 0.8698,
+      "step": 14877
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9498402998337955e-05,
+      "loss": 1.2127,
+      "step": 14878
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.948599979540579e-05,
+      "loss": 1.2205,
+      "step": 14879
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.947359806111613e-05,
+      "loss": 0.732,
+      "step": 14880
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.946119779576992e-05,
+      "loss": 0.8783,
+      "step": 14881
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9448798999668103e-05,
+      "loss": 1.264,
+      "step": 14882
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9436401673111554e-05,
+      "loss": 1.0378,
+      "step": 14883
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9424005816401235e-05,
+      "loss": 0.777,
+      "step": 14884
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9411611429837925e-05,
+      "loss": 0.9158,
+      "step": 14885
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.939921851372242e-05,
+      "loss": 0.7791,
+      "step": 14886
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9386827068355546e-05,
+      "loss": 0.8267,
+      "step": 14887
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9374437094038006e-05,
+      "loss": 0.8016,
+      "step": 14888
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.93620485910705e-05,
+      "loss": 0.7701,
+      "step": 14889
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9349661559753694e-05,
+      "loss": 0.9725,
+      "step": 14890
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.93372760003882e-05,
+      "loss": 0.8353,
+      "step": 14891
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.932489191327458e-05,
+      "loss": 0.7875,
+      "step": 14892
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9312509298713465e-05,
+      "loss": 0.9652,
+      "step": 14893
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.930012815700533e-05,
+      "loss": 0.8562,
+      "step": 14894
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.928774848845064e-05,
+      "loss": 0.8461,
+      "step": 14895
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.92753702933499e-05,
+      "loss": 0.8005,
+      "step": 14896
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9262993572003485e-05,
+      "loss": 0.9376,
+      "step": 14897
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.925061832471178e-05,
+      "loss": 0.8572,
+      "step": 14898
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9238244551775084e-05,
+      "loss": 0.8699,
+      "step": 14899
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9225872253493776e-05,
+      "loss": 0.9455,
+      "step": 14900
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9213501430168084e-05,
+      "loss": 0.9277,
+      "step": 14901
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9201132082098215e-05,
+      "loss": 0.9072,
+      "step": 14902
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.918876420958444e-05,
+      "loss": 1.2957,
+      "step": 14903
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9176397812926866e-05,
+      "loss": 0.996,
+      "step": 14904
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.916403289242564e-05,
+      "loss": 0.8125,
+      "step": 14905
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9151669448380836e-05,
+      "loss": 0.9475,
+      "step": 14906
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9139307481092514e-05,
+      "loss": 1.2521,
+      "step": 14907
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.912694699086066e-05,
+      "loss": 1.5863,
+      "step": 14908
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.911458797798533e-05,
+      "loss": 1.0439,
+      "step": 14909
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.910223044276643e-05,
+      "loss": 0.8048,
+      "step": 14910
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.908987438550383e-05,
+      "loss": 0.8275,
+      "step": 14911
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9077519806497486e-05,
+      "loss": 0.8357,
+      "step": 14912
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.906516670604719e-05,
+      "loss": 0.6956,
+      "step": 14913
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.905281508445273e-05,
+      "loss": 0.856,
+      "step": 14914
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.9040464942013924e-05,
+      "loss": 0.6924,
+      "step": 14915
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.902811627903048e-05,
+      "loss": 0.8528,
+      "step": 14916
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.901576909580208e-05,
+      "loss": 0.9012,
+      "step": 14917
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.900342339262837e-05,
+      "loss": 0.8087,
+      "step": 14918
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8991079169809055e-05,
+      "loss": 0.8701,
+      "step": 14919
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.897873642764359e-05,
+      "loss": 0.8103,
+      "step": 14920
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.896639516643165e-05,
+      "loss": 0.7976,
+      "step": 14921
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8954055386472684e-05,
+      "loss": 0.754,
+      "step": 14922
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.89417170880662e-05,
+      "loss": 1.3616,
+      "step": 14923
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.892938027151159e-05,
+      "loss": 0.9605,
+      "step": 14924
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8917044937108336e-05,
+      "loss": 0.8377,
+      "step": 14925
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.890471108515579e-05,
+      "loss": 0.8888,
+      "step": 14926
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8892378715953223e-05,
+      "loss": 0.9285,
+      "step": 14927
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.888004782980004e-05,
+      "loss": 0.833,
+      "step": 14928
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.886771842699544e-05,
+      "loss": 0.7965,
+      "step": 14929
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.885539050783862e-05,
+      "loss": 1.1526,
+      "step": 14930
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.884306407262886e-05,
+      "loss": 1.1375,
+      "step": 14931
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.883073912166527e-05,
+      "loss": 0.9931,
+      "step": 14932
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.881841565524696e-05,
+      "loss": 0.8349,
+      "step": 14933
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.880609367367304e-05,
+      "loss": 0.9055,
+      "step": 14934
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8793773177242546e-05,
+      "loss": 0.6855,
+      "step": 14935
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8781454166254436e-05,
+      "loss": 0.8707,
+      "step": 14936
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8769136641007775e-05,
+      "loss": 0.8149,
+      "step": 14937
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.875682060180147e-05,
+      "loss": 0.9306,
+      "step": 14938
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.874450604893441e-05,
+      "loss": 0.7467,
+      "step": 14939
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.873219298270545e-05,
+      "loss": 0.836,
+      "step": 14940
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.871988140341347e-05,
+      "loss": 0.8618,
+      "step": 14941
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.870757131135724e-05,
+      "loss": 0.8416,
+      "step": 14942
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.869526270683548e-05,
+      "loss": 0.7579,
+      "step": 14943
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.868295559014698e-05,
+      "loss": 0.7729,
+      "step": 14944
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.867064996159041e-05,
+      "loss": 1.2773,
+      "step": 14945
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.865834582146437e-05,
+      "loss": 1.0975,
+      "step": 14946
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8646043170067546e-05,
+      "loss": 0.9213,
+      "step": 14947
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8633742007698524e-05,
+      "loss": 0.8979,
+      "step": 14948
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8621442334655735e-05,
+      "loss": 0.8217,
+      "step": 14949
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.86091441512378e-05,
+      "loss": 0.864,
+      "step": 14950
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8596847457743135e-05,
+      "loss": 0.8381,
+      "step": 14951
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.858455225447016e-05,
+      "loss": 0.9077,
+      "step": 14952
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8572258541717335e-05,
+      "loss": 0.7522,
+      "step": 14953
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.855996631978298e-05,
+      "loss": 0.8374,
+      "step": 14954
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.854767558896543e-05,
+      "loss": 0.7942,
+      "step": 14955
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.853538634956294e-05,
+      "loss": 0.8236,
+      "step": 14956
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.852309860187382e-05,
+      "loss": 0.7603,
+      "step": 14957
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.851081234619627e-05,
+      "loss": 0.8611,
+      "step": 14958
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.849852758282843e-05,
+      "loss": 0.7912,
+      "step": 14959
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.848624431206851e-05,
+      "loss": 0.8757,
+      "step": 14960
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.847396253421458e-05,
+      "loss": 0.8672,
+      "step": 14961
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.846168224956469e-05,
+      "loss": 0.837,
+      "step": 14962
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.844940345841697e-05,
+      "loss": 0.6754,
+      "step": 14963
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8437126161069315e-05,
+      "loss": 0.8049,
+      "step": 14964
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.84248503578197e-05,
+      "loss": 0.8351,
+      "step": 14965
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.841257604896611e-05,
+      "loss": 0.8536,
+      "step": 14966
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.840030323480641e-05,
+      "loss": 0.7592,
+      "step": 14967
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.838803191563841e-05,
+      "loss": 0.9443,
+      "step": 14968
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.837576209176e-05,
+      "loss": 0.784,
+      "step": 14969
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.836349376346894e-05,
+      "loss": 0.7801,
+      "step": 14970
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.835122693106296e-05,
+      "loss": 0.8744,
+      "step": 14971
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.833896159483974e-05,
+      "loss": 0.9532,
+      "step": 14972
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.832669775509702e-05,
+      "loss": 0.7468,
+      "step": 14973
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.831443541213241e-05,
+      "loss": 0.8004,
+      "step": 14974
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.830217456624347e-05,
+      "loss": 1.2243,
+      "step": 14975
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.828991521772784e-05,
+      "loss": 1.1407,
+      "step": 14976
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8277657366883e-05,
+      "loss": 0.8802,
+      "step": 14977
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8265401014006466e-05,
+      "loss": 0.9734,
+      "step": 14978
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.825314615939566e-05,
+      "loss": 0.6487,
+      "step": 14979
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8240892803348016e-05,
+      "loss": 0.8348,
+      "step": 14980
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8228640946160886e-05,
+      "loss": 0.7684,
+      "step": 14981
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.821639058813169e-05,
+      "loss": 0.8885,
+      "step": 14982
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.82041417295577e-05,
+      "loss": 0.8901,
+      "step": 14983
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8191894370736136e-05,
+      "loss": 0.8387,
+      "step": 14984
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.817964851196432e-05,
+      "loss": 0.8033,
+      "step": 14985
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.816740415353942e-05,
+      "loss": 0.9083,
+      "step": 14986
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.815516129575856e-05,
+      "loss": 0.7169,
+      "step": 14987
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.814291993891894e-05,
+      "loss": 0.8252,
+      "step": 14988
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.813068008331762e-05,
+      "loss": 0.8749,
+      "step": 14989
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.811844172925164e-05,
+      "loss": 0.8034,
+      "step": 14990
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.810620487701799e-05,
+      "loss": 0.9933,
+      "step": 14991
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.809396952691377e-05,
+      "loss": 0.8901,
+      "step": 14992
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8081735679235775e-05,
+      "loss": 0.8917,
+      "step": 14993
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.806950333428102e-05,
+      "loss": 0.8394,
+      "step": 14994
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.805727249234634e-05,
+      "loss": 0.9262,
+      "step": 14995
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.804504315372857e-05,
+      "loss": 0.8386,
+      "step": 14996
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.803281531872448e-05,
+      "loss": 0.8473,
+      "step": 14997
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.802058898763089e-05,
+      "loss": 0.8617,
+      "step": 14998
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8008364160744514e-05,
+      "loss": 0.7246,
+      "step": 14999
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.7996140838361996e-05,
+      "loss": 0.9283,
+      "step": 15000
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.798391902078005e-05,
+      "loss": 0.8696,
+      "step": 15001
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.797169870829527e-05,
+      "loss": 1.3746,
+      "step": 15002
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.79594799012042e-05,
+      "loss": 1.1568,
+      "step": 15003
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.794726259980345e-05,
+      "loss": 0.8681,
+      "step": 15004
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.793504680438949e-05,
+      "loss": 0.7635,
+      "step": 15005
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.79228325152588e-05,
+      "loss": 0.8075,
+      "step": 15006
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.791061973270781e-05,
+      "loss": 0.7257,
+      "step": 15007
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.789840845703291e-05,
+      "loss": 0.854,
+      "step": 15008
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.788619868853044e-05,
+      "loss": 0.9185,
+      "step": 15009
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.787399042749677e-05,
+      "loss": 0.9072,
+      "step": 15010
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.786178367422819e-05,
+      "loss": 0.9487,
+      "step": 15011
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.784957842902094e-05,
+      "loss": 0.9346,
+      "step": 15012
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.783737469217117e-05,
+      "loss": 0.7205,
+      "step": 15013
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.782517246397515e-05,
+      "loss": 0.8887,
+      "step": 15014
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.781297174472901e-05,
+      "loss": 0.8425,
+      "step": 15015
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.780077253472878e-05,
+      "loss": 0.8362,
+      "step": 15016
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.7788574834270615e-05,
+      "loss": 0.8783,
+      "step": 15017
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.777637864365051e-05,
+      "loss": 0.7565,
+      "step": 15018
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.776418396316444e-05,
+      "loss": 1.2615,
+      "step": 15019
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.77519907931084e-05,
+      "loss": 0.9246,
+      "step": 15020
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.773979913377834e-05,
+      "loss": 0.8816,
+      "step": 15021
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.772760898547003e-05,
+      "loss": 0.9391,
+      "step": 15022
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.771542034847943e-05,
+      "loss": 0.7934,
+      "step": 15023
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.770323322310231e-05,
+      "loss": 0.8475,
+      "step": 15024
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.769104760963441e-05,
+      "loss": 0.9363,
+      "step": 15025
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.767886350837153e-05,
+      "loss": 0.8514,
+      "step": 15026
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.766668091960937e-05,
+      "loss": 0.9329,
+      "step": 15027
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7654499843643556e-05,
+      "loss": 0.7158,
+      "step": 15028
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7642320280769696e-05,
+      "loss": 0.7766,
+      "step": 15029
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.763014223128345e-05,
+      "loss": 0.9565,
+      "step": 15030
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.761796569548035e-05,
+      "loss": 0.845,
+      "step": 15031
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.760579067365586e-05,
+      "loss": 0.7265,
+      "step": 15032
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7593617166105545e-05,
+      "loss": 0.931,
+      "step": 15033
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7581445173124816e-05,
+      "loss": 0.8561,
+      "step": 15034
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.756927469500903e-05,
+      "loss": 0.8292,
+      "step": 15035
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.755710573205368e-05,
+      "loss": 0.8906,
+      "step": 15036
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7544938284553974e-05,
+      "loss": 0.8918,
+      "step": 15037
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.753277235280523e-05,
+      "loss": 0.8079,
+      "step": 15038
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.752060793710278e-05,
+      "loss": 0.9385,
+      "step": 15039
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.75084450377418e-05,
+      "loss": 0.858,
+      "step": 15040
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.749628365501745e-05,
+      "loss": 0.8712,
+      "step": 15041
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7484123789224946e-05,
+      "loss": 0.8482,
+      "step": 15042
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7471965440659364e-05,
+      "loss": 0.8297,
+      "step": 15043
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.745980860961576e-05,
+      "loss": 0.8196,
+      "step": 15044
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.744765329638922e-05,
+      "loss": 0.8119,
+      "step": 15045
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.743549950127473e-05,
+      "loss": 0.9203,
+      "step": 15046
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.742334722456724e-05,
+      "loss": 0.78,
+      "step": 15047
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.741119646656167e-05,
+      "loss": 0.9769,
+      "step": 15048
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.739904722755296e-05,
+      "loss": 0.8533,
+      "step": 15049
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.738689950783595e-05,
+      "loss": 0.8492,
+      "step": 15050
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.737475330770543e-05,
+      "loss": 0.7967,
+      "step": 15051
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7362608627456206e-05,
+      "loss": 0.8957,
+      "step": 15052
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.735046546738302e-05,
+      "loss": 0.8224,
+      "step": 15053
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.733832382778054e-05,
+      "loss": 0.7798,
+      "step": 15054
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7326183708943506e-05,
+      "loss": 1.3487,
+      "step": 15055
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.731404511116652e-05,
+      "loss": 1.0009,
+      "step": 15056
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.730190803474414e-05,
+      "loss": 0.9115,
+      "step": 15057
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7289772479970996e-05,
+      "loss": 0.7079,
+      "step": 15058
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.727763844714159e-05,
+      "loss": 0.8336,
+      "step": 15059
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.726550593655036e-05,
+      "loss": 1.0003,
+      "step": 15060
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.725337494849183e-05,
+      "loss": 0.853,
+      "step": 15061
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7241245483260365e-05,
+      "loss": 0.8818,
+      "step": 15062
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.722911754115037e-05,
+      "loss": 0.6988,
+      "step": 15063
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7216991122456124e-05,
+      "loss": 0.8562,
+      "step": 15064
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7204866227472e-05,
+      "loss": 0.6823,
+      "step": 15065
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.719274285649225e-05,
+      "loss": 0.7577,
+      "step": 15066
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7180621009811066e-05,
+      "loss": 0.9195,
+      "step": 15067
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.716850068772266e-05,
+      "loss": 0.8765,
+      "step": 15068
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7156381890521194e-05,
+      "loss": 0.877,
+      "step": 15069
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.714426461850073e-05,
+      "loss": 0.7885,
+      "step": 15070
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.713214887195544e-05,
+      "loss": 0.8665,
+      "step": 15071
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7120034651179316e-05,
+      "loss": 1.0016,
+      "step": 15072
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.710792195646632e-05,
+      "loss": 0.8575,
+      "step": 15073
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.709581078811051e-05,
+      "loss": 0.8989,
+      "step": 15074
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.708370114640577e-05,
+      "loss": 0.9321,
+      "step": 15075
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.707159303164597e-05,
+      "loss": 0.811,
+      "step": 15076
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.705948644412504e-05,
+      "loss": 0.8271,
+      "step": 15077
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.7047381384136735e-05,
+      "loss": 1.0347,
+      "step": 15078
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.703527785197488e-05,
+      "loss": 0.9487,
+      "step": 15079
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.702317584793319e-05,
+      "loss": 1.0108,
+      "step": 15080
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.701107537230539e-05,
+      "loss": 0.852,
+      "step": 15081
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.699897642538511e-05,
+      "loss": 0.8193,
+      "step": 15082
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6986879007466066e-05,
+      "loss": 1.0234,
+      "step": 15083
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.69747831188418e-05,
+      "loss": 0.8297,
+      "step": 15084
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.69626887598059e-05,
+      "loss": 0.8736,
+      "step": 15085
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6950595930651834e-05,
+      "loss": 0.8631,
+      "step": 15086
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.693850463167316e-05,
+      "loss": 0.8558,
+      "step": 15087
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6926414863163305e-05,
+      "loss": 0.7807,
+      "step": 15088
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.691432662541564e-05,
+      "loss": 0.9242,
+      "step": 15089
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.69022399187236e-05,
+      "loss": 0.8135,
+      "step": 15090
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.68901547433805e-05,
+      "loss": 0.8419,
+      "step": 15091
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.68780710996796e-05,
+      "loss": 0.8621,
+      "step": 15092
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.686598898791425e-05,
+      "loss": 1.0942,
+      "step": 15093
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6853908408377615e-05,
+      "loss": 0.8158,
+      "step": 15094
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6841829361362886e-05,
+      "loss": 0.859,
+      "step": 15095
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.682975184716324e-05,
+      "loss": 1.1891,
+      "step": 15096
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.681767586607178e-05,
+      "loss": 1.0884,
+      "step": 15097
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.680560141838153e-05,
+      "loss": 0.9105,
+      "step": 15098
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.679352850438562e-05,
+      "loss": 0.95,
+      "step": 15099
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6781457124377026e-05,
+      "loss": 0.8453,
+      "step": 15100
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.676938727864869e-05,
+      "loss": 0.9545,
+      "step": 15101
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6757318967493525e-05,
+      "loss": 0.9562,
+      "step": 15102
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6745252191204485e-05,
+      "loss": 1.3716,
+      "step": 15103
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6733186950074394e-05,
+      "loss": 1.6227,
+      "step": 15104
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6721123244396015e-05,
+      "loss": 1.4607,
+      "step": 15105
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.670906107446222e-05,
+      "loss": 1.0656,
+      "step": 15106
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.669700044056572e-05,
+      "loss": 0.7813,
+      "step": 15107
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.668494134299916e-05,
+      "loss": 0.8335,
+      "step": 15108
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.66728837820553e-05,
+      "loss": 1.3203,
+      "step": 15109
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.666082775802675e-05,
+      "loss": 1.1712,
+      "step": 15110
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6648773271206024e-05,
+      "loss": 0.8929,
+      "step": 15111
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.663672032188576e-05,
+      "loss": 0.85,
+      "step": 15112
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.662466891035846e-05,
+      "loss": 0.6862,
+      "step": 15113
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.661261903691656e-05,
+      "loss": 0.859,
+      "step": 15114
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6600570701852565e-05,
+      "loss": 0.907,
+      "step": 15115
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6588523905458874e-05,
+      "loss": 0.8352,
+      "step": 15116
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.657647864802779e-05,
+      "loss": 0.7023,
+      "step": 15117
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6564434929851734e-05,
+      "loss": 1.2098,
+      "step": 15118
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.655239275122296e-05,
+      "loss": 1.0819,
+      "step": 15119
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6540352112433726e-05,
+      "loss": 0.8722,
+      "step": 15120
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.652831301377623e-05,
+      "loss": 0.8826,
+      "step": 15121
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6516275455542694e-05,
+      "loss": 0.866,
+      "step": 15122
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6504239438025256e-05,
+      "loss": 0.7884,
+      "step": 15123
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.649220496151601e-05,
+      "loss": 0.8304,
+      "step": 15124
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.648017202630703e-05,
+      "loss": 1.3374,
+      "step": 15125
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.646814063269034e-05,
+      "loss": 1.5681,
+      "step": 15126
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.645611078095792e-05,
+      "loss": 1.0798,
+      "step": 15127
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6444082471401784e-05,
+      "loss": 0.9266,
+      "step": 15128
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.643205570431383e-05,
+      "loss": 0.7166,
+      "step": 15129
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.642003047998588e-05,
+      "loss": 0.7799,
+      "step": 15130
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6408006798709884e-05,
+      "loss": 0.777,
+      "step": 15131
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.639598466077758e-05,
+      "loss": 0.8226,
+      "step": 15132
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6383964066480734e-05,
+      "loss": 0.9312,
+      "step": 15133
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.637194501611114e-05,
+      "loss": 0.8041,
+      "step": 15134
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6359927509960446e-05,
+      "loss": 0.839,
+      "step": 15135
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6347911548320336e-05,
+      "loss": 0.859,
+      "step": 15136
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.633589713148237e-05,
+      "loss": 0.8026,
+      "step": 15137
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6323884259738204e-05,
+      "loss": 0.912,
+      "step": 15138
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6311872933379375e-05,
+      "loss": 1.3374,
+      "step": 15139
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.629986315269737e-05,
+      "loss": 1.1014,
+      "step": 15140
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6287854917983655e-05,
+      "loss": 0.885,
+      "step": 15141
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.627584822952967e-05,
+      "loss": 0.965,
+      "step": 15142
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.626384308762679e-05,
+      "loss": 0.9844,
+      "step": 15143
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.625183949256643e-05,
+      "loss": 0.8659,
+      "step": 15144
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.623983744463987e-05,
+      "loss": 0.9192,
+      "step": 15145
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6227836944138374e-05,
+      "loss": 0.8582,
+      "step": 15146
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.621583799135324e-05,
+      "loss": 0.7909,
+      "step": 15147
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.620384058657565e-05,
+      "loss": 0.7856,
+      "step": 15148
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.619184473009676e-05,
+      "loss": 0.9072,
+      "step": 15149
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6179850422207725e-05,
+      "loss": 0.6962,
+      "step": 15150
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.616785766319966e-05,
+      "loss": 0.9287,
+      "step": 15151
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6155866453363574e-05,
+      "loss": 0.9572,
+      "step": 15152
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6143876792990496e-05,
+      "loss": 0.7794,
+      "step": 15153
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6131888682371476e-05,
+      "loss": 0.7076,
+      "step": 15154
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6119902121797346e-05,
+      "loss": 0.9149,
+      "step": 15155
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.61079171115591e-05,
+      "loss": 0.8517,
+      "step": 15156
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6095933651947576e-05,
+      "loss": 0.8599,
+      "step": 15157
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.608395174325362e-05,
+      "loss": 0.7522,
+      "step": 15158
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.607197138576798e-05,
+      "loss": 0.7923,
+      "step": 15159
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6059992579781484e-05,
+      "loss": 1.4154,
+      "step": 15160
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.6048015325584813e-05,
+      "loss": 1.1324,
+      "step": 15161
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.603603962346863e-05,
+      "loss": 0.9409,
+      "step": 15162
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.602406547372364e-05,
+      "loss": 0.6408,
+      "step": 15163
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.601209287664039e-05,
+      "loss": 0.8777,
+      "step": 15164
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.600012183250945e-05,
+      "loss": 0.8239,
+      "step": 15165
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.598815234162141e-05,
+      "loss": 0.8059,
+      "step": 15166
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5976184404266714e-05,
+      "loss": 0.7442,
+      "step": 15167
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.596421802073582e-05,
+      "loss": 1.3479,
+      "step": 15168
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5952253191319165e-05,
+      "loss": 1.2024,
+      "step": 15169
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.594028991630711e-05,
+      "loss": 0.7774,
+      "step": 15170
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.592832819598998e-05,
+      "loss": 0.7111,
+      "step": 15171
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.591636803065812e-05,
+      "loss": 0.8261,
+      "step": 15172
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.590440942060179e-05,
+      "loss": 1.0111,
+      "step": 15173
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5892452366111185e-05,
+      "loss": 0.9102,
+      "step": 15174
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.588049686747654e-05,
+      "loss": 0.9148,
+      "step": 15175
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5868542924988004e-05,
+      "loss": 0.6752,
+      "step": 15176
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5856590538935674e-05,
+      "loss": 0.8953,
+      "step": 15177
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5844639709609596e-05,
+      "loss": 0.8171,
+      "step": 15178
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5832690437299874e-05,
+      "loss": 0.7779,
+      "step": 15179
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5820742722296486e-05,
+      "loss": 0.8675,
+      "step": 15180
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.580879656488937e-05,
+      "loss": 0.8872,
+      "step": 15181
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.579685196536851e-05,
+      "loss": 0.8045,
+      "step": 15182
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5784908924023775e-05,
+      "loss": 0.8829,
+      "step": 15183
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.577296744114496e-05,
+      "loss": 0.7752,
+      "step": 15184
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.576102751702194e-05,
+      "loss": 0.7881,
+      "step": 15185
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.574908915194447e-05,
+      "loss": 0.9226,
+      "step": 15186
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.573715234620226e-05,
+      "loss": 1.5059,
+      "step": 15187
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5725217100085076e-05,
+      "loss": 1.1644,
+      "step": 15188
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.571328341388255e-05,
+      "loss": 0.9733,
+      "step": 15189
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5701351287884255e-05,
+      "loss": 0.8887,
+      "step": 15190
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.568942072237985e-05,
+      "loss": 0.8019,
+      "step": 15191
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.567749171765887e-05,
+      "loss": 0.9556,
+      "step": 15192
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.566556427401081e-05,
+      "loss": 0.846,
+      "step": 15193
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5653638391725095e-05,
+      "loss": 0.8694,
+      "step": 15194
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.564171407109126e-05,
+      "loss": 0.7213,
+      "step": 15195
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5629791312398644e-05,
+      "loss": 0.9912,
+      "step": 15196
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.561787011593658e-05,
+      "loss": 0.7759,
+      "step": 15197
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.560595048199449e-05,
+      "loss": 0.8219,
+      "step": 15198
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.559403241086154e-05,
+      "loss": 0.836,
+      "step": 15199
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.558211590282701e-05,
+      "loss": 0.8315,
+      "step": 15200
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5570200958180143e-05,
+      "loss": 0.8726,
+      "step": 15201
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.555828757721009e-05,
+      "loss": 1.4035,
+      "step": 15202
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.554637576020594e-05,
+      "loss": 1.0691,
+      "step": 15203
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.553446550745687e-05,
+      "loss": 0.911,
+      "step": 15204
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5522556819251883e-05,
+      "loss": 0.738,
+      "step": 15205
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.551064969587997e-05,
+      "loss": 0.8003,
+      "step": 15206
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.549874413763019e-05,
+      "loss": 0.8477,
+      "step": 15207
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5486840144791436e-05,
+      "loss": 0.8642,
+      "step": 15208
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5474937717652604e-05,
+      "loss": 0.7441,
+      "step": 15209
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.546303685650254e-05,
+      "loss": 0.8388,
+      "step": 15210
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.545113756163014e-05,
+      "loss": 0.9511,
+      "step": 15211
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.543923983332416e-05,
+      "loss": 1.0113,
+      "step": 15212
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5427343671873333e-05,
+      "loss": 0.8657,
+      "step": 15213
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.54154490775664e-05,
+      "loss": 0.8435,
+      "step": 15214
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.540355605069201e-05,
+      "loss": 0.9579,
+      "step": 15215
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5391664591538784e-05,
+      "loss": 0.7209,
+      "step": 15216
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5379774700395395e-05,
+      "loss": 1.3878,
+      "step": 15217
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.536788637755035e-05,
+      "loss": 1.073,
+      "step": 15218
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5355999623292144e-05,
+      "loss": 0.8678,
+      "step": 15219
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.534411443790935e-05,
+      "loss": 1.294,
+      "step": 15220
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.533223082169036e-05,
+      "loss": 1.1504,
+      "step": 15221
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.532034877492355e-05,
+      "loss": 0.7804,
+      "step": 15222
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5308468297897356e-05,
+      "loss": 0.7622,
+      "step": 15223
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.52965893909001e-05,
+      "loss": 0.8063,
+      "step": 15224
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.528471205422006e-05,
+      "loss": 0.8765,
+      "step": 15225
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.527283628814545e-05,
+      "loss": 0.6995,
+      "step": 15226
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.52609620929646e-05,
+      "loss": 1.319,
+      "step": 15227
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.5249089468965555e-05,
+      "loss": 1.0188,
+      "step": 15228
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.523721841643657e-05,
+      "loss": 0.7458,
+      "step": 15229
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.522534893566569e-05,
+      "loss": 0.8018,
+      "step": 15230
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.521348102694101e-05,
+      "loss": 0.7943,
+      "step": 15231
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.52016146905505e-05,
+      "loss": 0.7901,
+      "step": 15232
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.5189749926782224e-05,
+      "loss": 1.3151,
+      "step": 15233
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.51778867359241e-05,
+      "loss": 1.096,
+      "step": 15234
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.5166025118264024e-05,
+      "loss": 1.21,
+      "step": 15235
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.5154165074089905e-05,
+      "loss": 1.1487,
+      "step": 15236
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.514230660368958e-05,
+      "loss": 0.8552,
+      "step": 15237
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.513044970735079e-05,
+      "loss": 0.8015,
+      "step": 15238
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.511859438536138e-05,
+      "loss": 0.9469,
+      "step": 15239
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.510674063800903e-05,
+      "loss": 0.8179,
+      "step": 15240
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.509488846558142e-05,
+      "loss": 0.7862,
+      "step": 15241
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.5083037868366194e-05,
+      "loss": 0.8208,
+      "step": 15242
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.507118884665098e-05,
+      "loss": 0.8291,
+      "step": 15243
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.505934140072329e-05,
+      "loss": 0.8246,
+      "step": 15244
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.504749553087074e-05,
+      "loss": 0.8288,
+      "step": 15245
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.503565123738079e-05,
+      "loss": 0.8791,
+      "step": 15246
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.5023808520540835e-05,
+      "loss": 0.8365,
+      "step": 15247
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.501196738063839e-05,
+      "loss": 0.8338,
+      "step": 15248
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.500012781796077e-05,
+      "loss": 0.8844,
+      "step": 15249
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.498828983279535e-05,
+      "loss": 0.8709,
+      "step": 15250
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.497645342542937e-05,
+      "loss": 0.8367,
+      "step": 15251
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.496461859615017e-05,
+      "loss": 0.8464,
+      "step": 15252
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4952785345244945e-05,
+      "loss": 0.8072,
+      "step": 15253
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4940953673000834e-05,
+      "loss": 0.8596,
+      "step": 15254
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.492912357970507e-05,
+      "loss": 0.7879,
+      "step": 15255
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.491729506564472e-05,
+      "loss": 0.884,
+      "step": 15256
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4905468131106855e-05,
+      "loss": 0.8679,
+      "step": 15257
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.489364277637852e-05,
+      "loss": 0.7809,
+      "step": 15258
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.48818190017467e-05,
+      "loss": 0.8311,
+      "step": 15259
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.486999680749832e-05,
+      "loss": 0.8586,
+      "step": 15260
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.485817619392036e-05,
+      "loss": 0.8005,
+      "step": 15261
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.484635716129967e-05,
+      "loss": 0.8571,
+      "step": 15262
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.483453970992305e-05,
+      "loss": 1.2429,
+      "step": 15263
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.48227238400774e-05,
+      "loss": 1.0691,
+      "step": 15264
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.481090955204942e-05,
+      "loss": 0.8154,
+      "step": 15265
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.479909684612584e-05,
+      "loss": 1.0115,
+      "step": 15266
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.478728572259334e-05,
+      "loss": 0.818,
+      "step": 15267
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.477547618173861e-05,
+      "loss": 0.8632,
+      "step": 15268
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4763668223848236e-05,
+      "loss": 0.912,
+      "step": 15269
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.475186184920875e-05,
+      "loss": 0.6674,
+      "step": 15270
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.474005705810681e-05,
+      "loss": 0.8819,
+      "step": 15271
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.472825385082878e-05,
+      "loss": 1.075,
+      "step": 15272
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4716452227661146e-05,
+      "loss": 1.0003,
+      "step": 15273
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.470465218889038e-05,
+      "loss": 0.7711,
+      "step": 15274
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.469285373480283e-05,
+      "loss": 0.7105,
+      "step": 15275
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4681056865684805e-05,
+      "loss": 0.7776,
+      "step": 15276
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.466926158182268e-05,
+      "loss": 0.8216,
+      "step": 15277
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.465746788350267e-05,
+      "loss": 0.9563,
+      "step": 15278
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4645675771010986e-05,
+      "loss": 0.7759,
+      "step": 15279
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.463388524463389e-05,
+      "loss": 0.858,
+      "step": 15280
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.462209630465748e-05,
+      "loss": 0.8666,
+      "step": 15281
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.461030895136786e-05,
+      "loss": 0.8988,
+      "step": 15282
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.45985231850511e-05,
+      "loss": 0.8915,
+      "step": 15283
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4586739005993285e-05,
+      "loss": 0.8223,
+      "step": 15284
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.457495641448037e-05,
+      "loss": 0.9255,
+      "step": 15285
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.456317541079832e-05,
+      "loss": 0.8055,
+      "step": 15286
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.455139599523306e-05,
+      "loss": 0.9653,
+      "step": 15287
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.453961816807045e-05,
+      "loss": 0.7744,
+      "step": 15288
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.452784192959633e-05,
+      "loss": 0.823,
+      "step": 15289
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.451606728009653e-05,
+      "loss": 0.7678,
+      "step": 15290
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.450429421985681e-05,
+      "loss": 0.8572,
+      "step": 15291
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.449252274916286e-05,
+      "loss": 0.685,
+      "step": 15292
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4480752868300426e-05,
+      "loss": 0.8861,
+      "step": 15293
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.446898457755512e-05,
+      "loss": 0.8017,
+      "step": 15294
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.445721787721252e-05,
+      "loss": 0.9565,
+      "step": 15295
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.444545276755828e-05,
+      "loss": 0.7934,
+      "step": 15296
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.443368924887789e-05,
+      "loss": 1.0283,
+      "step": 15297
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4421927321456834e-05,
+      "loss": 0.7544,
+      "step": 15298
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.441016698558055e-05,
+      "loss": 0.7711,
+      "step": 15299
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.439840824153451e-05,
+      "loss": 0.751,
+      "step": 15300
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.438665108960406e-05,
+      "loss": 0.78,
+      "step": 15301
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.437489553007455e-05,
+      "loss": 0.7314,
+      "step": 15302
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4363141563231274e-05,
+      "loss": 0.7615,
+      "step": 15303
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4351389189359495e-05,
+      "loss": 0.982,
+      "step": 15304
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.43396384087444e-05,
+      "loss": 0.7069,
+      "step": 15305
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.432788922167125e-05,
+      "loss": 0.9405,
+      "step": 15306
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4316141628425155e-05,
+      "loss": 0.8216,
+      "step": 15307
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4304395629291184e-05,
+      "loss": 0.8773,
+      "step": 15308
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4292651224554476e-05,
+      "loss": 1.3823,
+      "step": 15309
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4280908414500026e-05,
+      "loss": 1.134,
+      "step": 15310
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4269167199412796e-05,
+      "loss": 0.8322,
+      "step": 15311
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4257427579577804e-05,
+      "loss": 0.8291,
+      "step": 15312
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4245689555279936e-05,
+      "loss": 0.9488,
+      "step": 15313
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.423395312680402e-05,
+      "loss": 0.9458,
+      "step": 15314
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4222218294435004e-05,
+      "loss": 0.8642,
+      "step": 15315
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4210485058457585e-05,
+      "loss": 0.9328,
+      "step": 15316
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4198753419156524e-05,
+      "loss": 0.8453,
+      "step": 15317
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.418702337681662e-05,
+      "loss": 0.8614,
+      "step": 15318
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.417529493172249e-05,
+      "loss": 0.9457,
+      "step": 15319
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.416356808415877e-05,
+      "loss": 0.9111,
+      "step": 15320
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4151842834410105e-05,
+      "loss": 0.7141,
+      "step": 15321
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.414011918276107e-05,
+      "loss": 0.8802,
+      "step": 15322
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.412839712949615e-05,
+      "loss": 0.9218,
+      "step": 15323
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.411667667489982e-05,
+      "loss": 0.9117,
+      "step": 15324
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.410495781925659e-05,
+      "loss": 0.8873,
+      "step": 15325
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.409324056285085e-05,
+      "loss": 0.7583,
+      "step": 15326
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.408152490596691e-05,
+      "loss": 0.8118,
+      "step": 15327
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4069810848889194e-05,
+      "loss": 0.8477,
+      "step": 15328
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.405809839190196e-05,
+      "loss": 0.8986,
+      "step": 15329
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.404638753528945e-05,
+      "loss": 0.9437,
+      "step": 15330
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.4034678279335895e-05,
+      "loss": 0.7968,
+      "step": 15331
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.402297062432547e-05,
+      "loss": 0.766,
+      "step": 15332
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.401126457054227e-05,
+      "loss": 0.8238,
+      "step": 15333
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3999560118270454e-05,
+      "loss": 0.9281,
+      "step": 15334
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.398785726779408e-05,
+      "loss": 0.8049,
+      "step": 15335
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.397615601939711e-05,
+      "loss": 0.8978,
+      "step": 15336
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3964456373363615e-05,
+      "loss": 0.7351,
+      "step": 15337
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.395275832997748e-05,
+      "loss": 0.8324,
+      "step": 15338
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.394106188952264e-05,
+      "loss": 0.8385,
+      "step": 15339
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.39293670522829e-05,
+      "loss": 0.8449,
+      "step": 15340
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3917673818542176e-05,
+      "loss": 0.741,
+      "step": 15341
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3905982188584196e-05,
+      "loss": 1.3135,
+      "step": 15342
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.389429216269271e-05,
+      "loss": 1.0628,
+      "step": 15343
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.388260374115147e-05,
+      "loss": 0.8343,
+      "step": 15344
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3870916924244166e-05,
+      "loss": 0.9505,
+      "step": 15345
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.385923171225431e-05,
+      "loss": 0.9055,
+      "step": 15346
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.384754810546562e-05,
+      "loss": 0.9112,
+      "step": 15347
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.383586610416159e-05,
+      "loss": 0.7641,
+      "step": 15348
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.382418570862573e-05,
+      "loss": 0.8823,
+      "step": 15349
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.381250691914157e-05,
+      "loss": 0.7552,
+      "step": 15350
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3800829735992514e-05,
+      "loss": 0.7884,
+      "step": 15351
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3789154159461934e-05,
+      "loss": 0.8023,
+      "step": 15352
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.377748018983324e-05,
+      "loss": 0.8802,
+      "step": 15353
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.376580782738975e-05,
+      "loss": 0.8594,
+      "step": 15354
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3754137072414714e-05,
+      "loss": 0.8977,
+      "step": 15355
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.374246792519137e-05,
+      "loss": 0.6946,
+      "step": 15356
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3730800386002955e-05,
+      "loss": 0.8524,
+      "step": 15357
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.371913445513263e-05,
+      "loss": 0.8987,
+      "step": 15358
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3707470132863515e-05,
+      "loss": 0.8872,
+      "step": 15359
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3695807419478684e-05,
+      "loss": 0.9275,
+      "step": 15360
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.36841463152612e-05,
+      "loss": 0.8951,
+      "step": 15361
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.367248682049403e-05,
+      "loss": 0.9319,
+      "step": 15362
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.366082893546021e-05,
+      "loss": 0.8582,
+      "step": 15363
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.364917266044264e-05,
+      "loss": 0.8447,
+      "step": 15364
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.363751799572419e-05,
+      "loss": 0.7871,
+      "step": 15365
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.362586494158775e-05,
+      "loss": 0.9991,
+      "step": 15366
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.361421349831613e-05,
+      "loss": 0.8263,
+      "step": 15367
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3602563666192044e-05,
+      "loss": 0.9566,
+      "step": 15368
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.359091544549832e-05,
+      "loss": 0.8247,
+      "step": 15369
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.357926883651761e-05,
+      "loss": 0.947,
+      "step": 15370
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3567623839532555e-05,
+      "loss": 0.8845,
+      "step": 15371
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3555980454825766e-05,
+      "loss": 0.789,
+      "step": 15372
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.354433868267988e-05,
+      "loss": 0.9112,
+      "step": 15373
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3532698523377395e-05,
+      "loss": 0.864,
+      "step": 15374
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.352105997720081e-05,
+      "loss": 0.7782,
+      "step": 15375
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3509423044432595e-05,
+      "loss": 0.8815,
+      "step": 15376
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.349778772535515e-05,
+      "loss": 0.7337,
+      "step": 15377
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.348615402025089e-05,
+      "loss": 0.8747,
+      "step": 15378
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.347452192940217e-05,
+      "loss": 0.9033,
+      "step": 15379
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.346289145309125e-05,
+      "loss": 0.7189,
+      "step": 15380
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3451262591600396e-05,
+      "loss": 0.8023,
+      "step": 15381
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.343963534521188e-05,
+      "loss": 0.8156,
+      "step": 15382
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.342800971420788e-05,
+      "loss": 0.8285,
+      "step": 15383
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.341638569887048e-05,
+      "loss": 0.8425,
+      "step": 15384
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.340476329948187e-05,
+      "loss": 0.8449,
+      "step": 15385
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3393142516324083e-05,
+      "loss": 0.8462,
+      "step": 15386
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.338152334967912e-05,
+      "loss": 0.8804,
+      "step": 15387
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.336990579982905e-05,
+      "loss": 0.9308,
+      "step": 15388
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.33582898670558e-05,
+      "loss": 0.7897,
+      "step": 15389
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.334667555164121e-05,
+      "loss": 0.8184,
+      "step": 15390
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.333506285386722e-05,
+      "loss": 1.3893,
+      "step": 15391
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.332345177401566e-05,
+      "loss": 1.0037,
+      "step": 15392
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.331184231236827e-05,
+      "loss": 0.8736,
+      "step": 15393
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3300234469206884e-05,
+      "loss": 0.8773,
+      "step": 15394
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.328862824481318e-05,
+      "loss": 0.7597,
+      "step": 15395
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3277023639468855e-05,
+      "loss": 0.8964,
+      "step": 15396
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.326542065345548e-05,
+      "loss": 0.8408,
+      "step": 15397
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.325381928705473e-05,
+      "loss": 0.7304,
+      "step": 15398
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.324221954054816e-05,
+      "loss": 0.8277,
+      "step": 15399
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.323062141421721e-05,
+      "loss": 0.8704,
+      "step": 15400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.321902490834346e-05,
+      "loss": 0.9184,
+      "step": 15401
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.320743002320831e-05,
+      "loss": 0.7,
+      "step": 15402
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.319583675909312e-05,
+      "loss": 0.888,
+      "step": 15403
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.318424511627936e-05,
+      "loss": 0.862,
+      "step": 15404
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3172655095048256e-05,
+      "loss": 0.8114,
+      "step": 15405
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.316106669568108e-05,
+      "loss": 0.7595,
+      "step": 15406
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.314947991845915e-05,
+      "loss": 0.8978,
+      "step": 15407
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.313789476366364e-05,
+      "loss": 0.9513,
+      "step": 15408
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.312631123157568e-05,
+      "loss": 0.9193,
+      "step": 15409
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.311472932247646e-05,
+      "loss": 0.8221,
+      "step": 15410
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.310314903664703e-05,
+      "loss": 0.9225,
+      "step": 15411
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3091570374368456e-05,
+      "loss": 0.7352,
+      "step": 15412
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.307999333592169e-05,
+      "loss": 0.914,
+      "step": 15413
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.306841792158778e-05,
+      "loss": 1.1803,
+      "step": 15414
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3056844131647614e-05,
+      "loss": 1.1139,
+      "step": 15415
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.3045271966382066e-05,
+      "loss": 0.8694,
+      "step": 15416
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.303370142607203e-05,
+      "loss": 0.8733,
+      "step": 15417
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.302213251099832e-05,
+      "loss": 0.887,
+      "step": 15418
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.301056522144162e-05,
+      "loss": 0.8145,
+      "step": 15419
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.299899955768276e-05,
+      "loss": 0.899,
+      "step": 15420
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2987435520002394e-05,
+      "loss": 0.8971,
+      "step": 15421
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.297587310868114e-05,
+      "loss": 0.9259,
+      "step": 15422
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.296431232399968e-05,
+      "loss": 0.9266,
+      "step": 15423
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2952753166238556e-05,
+      "loss": 0.8843,
+      "step": 15424
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.294119563567828e-05,
+      "loss": 0.8003,
+      "step": 15425
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2929639732599396e-05,
+      "loss": 1.4216,
+      "step": 15426
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2918085457282336e-05,
+      "loss": 1.0947,
+      "step": 15427
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2906532810007506e-05,
+      "loss": 0.8044,
+      "step": 15428
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2894981791055254e-05,
+      "loss": 0.8295,
+      "step": 15429
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2883432400706e-05,
+      "loss": 0.8336,
+      "step": 15430
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.287188463923998e-05,
+      "loss": 0.9551,
+      "step": 15431
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.286033850693744e-05,
+      "loss": 0.7486,
+      "step": 15432
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.284879400407869e-05,
+      "loss": 0.8773,
+      "step": 15433
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.283725113094382e-05,
+      "loss": 0.9099,
+      "step": 15434
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.282570988781294e-05,
+      "loss": 0.8281,
+      "step": 15435
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.281417027496625e-05,
+      "loss": 0.7223,
+      "step": 15436
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.2802632292683755e-05,
+      "loss": 0.8629,
+      "step": 15437
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.279109594124545e-05,
+      "loss": 0.8133,
+      "step": 15438
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.27795612209314e-05,
+      "loss": 0.7371,
+      "step": 15439
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.276802813202148e-05,
+      "loss": 0.9397,
+      "step": 15440
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.275649667479558e-05,
+      "loss": 0.9395,
+      "step": 15441
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.274496684953363e-05,
+      "loss": 0.9826,
+      "step": 15442
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.27334386565154e-05,
+      "loss": 0.7104,
+      "step": 15443
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.27219120960207e-05,
+      "loss": 0.8708,
+      "step": 15444
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.271038716832921e-05,
+      "loss": 0.9156,
+      "step": 15445
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.269886387372073e-05,
+      "loss": 0.9858,
+      "step": 15446
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2687342212474884e-05,
+      "loss": 0.8109,
+      "step": 15447
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.267582218487129e-05,
+      "loss": 0.8407,
+      "step": 15448
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.266430379118951e-05,
+      "loss": 0.8322,
+      "step": 15449
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2652787031709096e-05,
+      "loss": 0.8406,
+      "step": 15450
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2641271906709606e-05,
+      "loss": 0.7313,
+      "step": 15451
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.262975841647047e-05,
+      "loss": 0.7346,
+      "step": 15452
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.261824656127111e-05,
+      "loss": 0.8129,
+      "step": 15453
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.260673634139089e-05,
+      "loss": 0.8182,
+      "step": 15454
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.25952277571092e-05,
+      "loss": 0.9329,
+      "step": 15455
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2583720808705356e-05,
+      "loss": 0.708,
+      "step": 15456
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.257221549645855e-05,
+      "loss": 0.786,
+      "step": 15457
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.256071182064809e-05,
+      "loss": 0.954,
+      "step": 15458
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.254920978155315e-05,
+      "loss": 0.9244,
+      "step": 15459
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.253770937945282e-05,
+      "loss": 0.8581,
+      "step": 15460
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2526210614626284e-05,
+      "loss": 0.8126,
+      "step": 15461
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.251471348735262e-05,
+      "loss": 0.9327,
+      "step": 15462
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.250321799791074e-05,
+      "loss": 0.8627,
+      "step": 15463
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.249172414657975e-05,
+      "loss": 0.9341,
+      "step": 15464
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.248023193363855e-05,
+      "loss": 0.8274,
+      "step": 15465
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.246874135936603e-05,
+      "loss": 1.286,
+      "step": 15466
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2457252424041126e-05,
+      "loss": 1.1287,
+      "step": 15467
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.244576512794263e-05,
+      "loss": 0.8568,
+      "step": 15468
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.243427947134933e-05,
+      "loss": 0.8848,
+      "step": 15469
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.242279545453995e-05,
+      "loss": 0.9631,
+      "step": 15470
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.241131307779325e-05,
+      "loss": 0.6814,
+      "step": 15471
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.239983234138789e-05,
+      "loss": 0.841,
+      "step": 15472
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.238835324560246e-05,
+      "loss": 0.9316,
+      "step": 15473
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.237687579071562e-05,
+      "loss": 0.7335,
+      "step": 15474
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2365399977005895e-05,
+      "loss": 0.9184,
+      "step": 15475
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.235392580475175e-05,
+      "loss": 1.2357,
+      "step": 15476
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.234245327423176e-05,
+      "loss": 1.5286,
+      "step": 15477
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.233098238572425e-05,
+      "loss": 1.1508,
+      "step": 15478
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.231951313950764e-05,
+      "loss": 0.8558,
+      "step": 15479
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.230804553586032e-05,
+      "loss": 0.8088,
+      "step": 15480
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.229657957506058e-05,
+      "loss": 0.7755,
+      "step": 15481
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.228511525738667e-05,
+      "loss": 1.3083,
+      "step": 15482
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2273652583116865e-05,
+      "loss": 1.1692,
+      "step": 15483
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2262191552529356e-05,
+      "loss": 0.819,
+      "step": 15484
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2250732165902265e-05,
+      "loss": 0.8412,
+      "step": 15485
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2239274423513686e-05,
+      "loss": 0.7398,
+      "step": 15486
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.222781832564177e-05,
+      "loss": 0.7667,
+      "step": 15487
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.221636387256449e-05,
+      "loss": 0.8348,
+      "step": 15488
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2204911064559825e-05,
+      "loss": 0.7701,
+      "step": 15489
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.21934599019058e-05,
+      "loss": 0.8985,
+      "step": 15490
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.218201038488029e-05,
+      "loss": 0.9211,
+      "step": 15491
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.217056251376116e-05,
+      "loss": 0.7944,
+      "step": 15492
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2159116288826253e-05,
+      "loss": 0.799,
+      "step": 15493
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.214767171035336e-05,
+      "loss": 0.7185,
+      "step": 15494
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.213622877862019e-05,
+      "loss": 0.8705,
+      "step": 15495
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.212478749390454e-05,
+      "loss": 0.9671,
+      "step": 15496
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2113347856484044e-05,
+      "loss": 0.8874,
+      "step": 15497
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2101909866636315e-05,
+      "loss": 0.6952,
+      "step": 15498
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2090473524639e-05,
+      "loss": 0.7716,
+      "step": 15499
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.20790388307696e-05,
+      "loss": 0.7701,
+      "step": 15500
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.206760578530568e-05,
+      "loss": 0.9018,
+      "step": 15501
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.205617438852463e-05,
+      "loss": 0.7992,
+      "step": 15502
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2044744640703984e-05,
+      "loss": 0.9531,
+      "step": 15503
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2033316542121084e-05,
+      "loss": 0.8634,
+      "step": 15504
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.202189009305326e-05,
+      "loss": 0.7727,
+      "step": 15505
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.201046529377793e-05,
+      "loss": 0.8241,
+      "step": 15506
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1999042144572245e-05,
+      "loss": 0.7754,
+      "step": 15507
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1987620645713456e-05,
+      "loss": 0.8714,
+      "step": 15508
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.197620079747883e-05,
+      "loss": 0.8108,
+      "step": 15509
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.196478260014548e-05,
+      "loss": 0.828,
+      "step": 15510
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1953366053990475e-05,
+      "loss": 0.765,
+      "step": 15511
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1941951159290976e-05,
+      "loss": 1.4513,
+      "step": 15512
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1930537916323966e-05,
+      "loss": 0.9853,
+      "step": 15513
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.191912632536641e-05,
+      "loss": 1.3289,
+      "step": 15514
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.190771638669533e-05,
+      "loss": 1.0945,
+      "step": 15515
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.189630810058761e-05,
+      "loss": 0.8159,
+      "step": 15516
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1884901467320084e-05,
+      "loss": 0.8729,
+      "step": 15517
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.187349648716964e-05,
+      "loss": 0.8393,
+      "step": 15518
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.186209316041307e-05,
+      "loss": 0.722,
+      "step": 15519
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.185069148732709e-05,
+      "loss": 0.8868,
+      "step": 15520
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1839291468188436e-05,
+      "loss": 0.6684,
+      "step": 15521
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.182789310327376e-05,
+      "loss": 1.3514,
+      "step": 15522
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1816496392859693e-05,
+      "loss": 1.0417,
+      "step": 15523
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1805101337222865e-05,
+      "loss": 0.857,
+      "step": 15524
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.17937079366398e-05,
+      "loss": 0.8338,
+      "step": 15525
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1782316191387026e-05,
+      "loss": 0.8908,
+      "step": 15526
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.177092610174096e-05,
+      "loss": 1.3565,
+      "step": 15527
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1759537667978125e-05,
+      "loss": 1.1429,
+      "step": 15528
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.174815089037484e-05,
+      "loss": 0.8688,
+      "step": 15529
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.173676576920747e-05,
+      "loss": 0.8911,
+      "step": 15530
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.172538230475235e-05,
+      "loss": 1.005,
+      "step": 15531
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.171400049728573e-05,
+      "loss": 0.8621,
+      "step": 15532
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.170262034708383e-05,
+      "loss": 0.9251,
+      "step": 15533
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.169124185442288e-05,
+      "loss": 0.6632,
+      "step": 15534
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1679865019579016e-05,
+      "loss": 0.8646,
+      "step": 15535
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.166848984282833e-05,
+      "loss": 0.7058,
+      "step": 15536
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.16571163244469e-05,
+      "loss": 0.789,
+      "step": 15537
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.164574446471076e-05,
+      "loss": 0.8124,
+      "step": 15538
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.163437426389585e-05,
+      "loss": 0.8974,
+      "step": 15539
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.16230057222782e-05,
+      "loss": 0.916,
+      "step": 15540
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1611638840133686e-05,
+      "loss": 1.2992,
+      "step": 15541
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.160027361773816e-05,
+      "loss": 1.0317,
+      "step": 15542
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.158891005536744e-05,
+      "loss": 0.8604,
+      "step": 15543
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.157754815329736e-05,
+      "loss": 1.2738,
+      "step": 15544
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1566187911803646e-05,
+      "loss": 1.2164,
+      "step": 15545
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.155482933116196e-05,
+      "loss": 0.7753,
+      "step": 15546
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.154347241164805e-05,
+      "loss": 0.8042,
+      "step": 15547
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.15321171535375e-05,
+      "loss": 0.8168,
+      "step": 15548
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.152076355710586e-05,
+      "loss": 0.8684,
+      "step": 15549
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.150941162262875e-05,
+      "loss": 0.8472,
+      "step": 15550
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1498061350381666e-05,
+      "loss": 0.7685,
+      "step": 15551
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.148671274063999e-05,
+      "loss": 0.7905,
+      "step": 15552
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1475365793679234e-05,
+      "loss": 1.534,
+      "step": 15553
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.146402050977474e-05,
+      "loss": 1.1612,
+      "step": 15554
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.145267688920184e-05,
+      "loss": 0.7851,
+      "step": 15555
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.144133493223589e-05,
+      "loss": 0.8979,
+      "step": 15556
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.142999463915213e-05,
+      "loss": 0.8704,
+      "step": 15557
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1418656010225765e-05,
+      "loss": 0.7892,
+      "step": 15558
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1407319045731965e-05,
+      "loss": 0.7857,
+      "step": 15559
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1395983745945924e-05,
+      "loss": 0.6892,
+      "step": 15560
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1384650111142725e-05,
+      "loss": 0.8111,
+      "step": 15561
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1373318141597376e-05,
+      "loss": 0.8063,
+      "step": 15562
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1361987837584986e-05,
+      "loss": 0.8628,
+      "step": 15563
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.135065919938047e-05,
+      "loss": 0.8055,
+      "step": 15564
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.133933222725881e-05,
+      "loss": 0.7598,
+      "step": 15565
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.132800692149488e-05,
+      "loss": 0.877,
+      "step": 15566
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.131668328236355e-05,
+      "loss": 0.8444,
+      "step": 15567
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.130536131013958e-05,
+      "loss": 0.748,
+      "step": 15568
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.129404100509785e-05,
+      "loss": 0.7719,
+      "step": 15569
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.128272236751304e-05,
+      "loss": 0.7637,
+      "step": 15570
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.127140539765984e-05,
+      "loss": 0.7796,
+      "step": 15571
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.126009009581293e-05,
+      "loss": 0.8307,
+      "step": 15572
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1248776462246946e-05,
+      "loss": 0.812,
+      "step": 15573
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1237464497236416e-05,
+      "loss": 0.7902,
+      "step": 15574
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.122615420105587e-05,
+      "loss": 0.9008,
+      "step": 15575
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.121484557397986e-05,
+      "loss": 0.9734,
+      "step": 15576
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.120353861628281e-05,
+      "loss": 0.824,
+      "step": 15577
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1192233328239104e-05,
+      "loss": 0.8535,
+      "step": 15578
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1180929710123174e-05,
+      "loss": 0.8384,
+      "step": 15579
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.116962776220933e-05,
+      "loss": 0.8163,
+      "step": 15580
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.115832748477185e-05,
+      "loss": 0.9036,
+      "step": 15581
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1147028878084994e-05,
+      "loss": 0.8995,
+      "step": 15582
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.113573194242298e-05,
+      "loss": 0.9322,
+      "step": 15583
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.112443667805993e-05,
+      "loss": 0.8763,
+      "step": 15584
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.111314308527006e-05,
+      "loss": 0.919,
+      "step": 15585
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.11018511643274e-05,
+      "loss": 0.7579,
+      "step": 15586
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.109056091550599e-05,
+      "loss": 0.7746,
+      "step": 15587
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.107927233907988e-05,
+      "loss": 1.2138,
+      "step": 15588
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1067985435323036e-05,
+      "loss": 1.0111,
+      "step": 15589
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1056700204509326e-05,
+      "loss": 0.8566,
+      "step": 15590
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1045416646912716e-05,
+      "loss": 0.9017,
+      "step": 15591
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.103413476280702e-05,
+      "loss": 0.9307,
+      "step": 15592
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.102285455246603e-05,
+      "loss": 0.7221,
+      "step": 15593
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.101157601616349e-05,
+      "loss": 0.8103,
+      "step": 15594
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1000299154173216e-05,
+      "loss": 0.8494,
+      "step": 15595
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.098902396676876e-05,
+      "loss": 0.8076,
+      "step": 15596
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0977750454223874e-05,
+      "loss": 1.2715,
+      "step": 15597
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0966478616812125e-05,
+      "loss": 1.093,
+      "step": 15598
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.095520845480705e-05,
+      "loss": 0.7781,
+      "step": 15599
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.094393996848217e-05,
+      "loss": 0.8865,
+      "step": 15600
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0932673158111014e-05,
+      "loss": 1.0008,
+      "step": 15601
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.092140802396698e-05,
+      "loss": 0.802,
+      "step": 15602
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0910144566323465e-05,
+      "loss": 0.9825,
+      "step": 15603
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.089888278545386e-05,
+      "loss": 0.8182,
+      "step": 15604
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.088762268163147e-05,
+      "loss": 0.8593,
+      "step": 15605
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.087636425512952e-05,
+      "loss": 0.86,
+      "step": 15606
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0865107506221336e-05,
+      "loss": 0.7955,
+      "step": 15607
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.085385243518005e-05,
+      "loss": 0.8164,
+      "step": 15608
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0842599042278856e-05,
+      "loss": 0.6864,
+      "step": 15609
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0831347327790827e-05,
+      "loss": 0.7301,
+      "step": 15610
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.082009729198907e-05,
+      "loss": 1.3309,
+      "step": 15611
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.080884893514656e-05,
+      "loss": 1.0779,
+      "step": 15612
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.079760225753637e-05,
+      "loss": 0.9395,
+      "step": 15613
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.078635725943141e-05,
+      "loss": 1.3654,
+      "step": 15614
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0775113941104594e-05,
+      "loss": 1.0709,
+      "step": 15615
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.076387230282874e-05,
+      "loss": 0.8733,
+      "step": 15616
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0752632344876785e-05,
+      "loss": 0.8759,
+      "step": 15617
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.074139406752144e-05,
+      "loss": 0.9245,
+      "step": 15618
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.073015747103545e-05,
+      "loss": 0.7059,
+      "step": 15619
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0718922555691556e-05,
+      "loss": 0.8149,
+      "step": 15620
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.070768932176242e-05,
+      "loss": 0.8228,
+      "step": 15621
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0696457769520626e-05,
+      "loss": 0.7177,
+      "step": 15622
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.068522789923881e-05,
+      "loss": 0.8059,
+      "step": 15623
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0673999711189525e-05,
+      "loss": 0.8123,
+      "step": 15624
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.066277320564518e-05,
+      "loss": 0.816,
+      "step": 15625
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.065154838287833e-05,
+      "loss": 0.8615,
+      "step": 15626
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.064032524316135e-05,
+      "loss": 0.9583,
+      "step": 15627
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0629103786766597e-05,
+      "loss": 0.7618,
+      "step": 15628
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.061788401396649e-05,
+      "loss": 0.8481,
+      "step": 15629
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0606665925033274e-05,
+      "loss": 1.2122,
+      "step": 15630
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0595449520239216e-05,
+      "loss": 1.2714,
+      "step": 15631
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0584234799856496e-05,
+      "loss": 0.8083,
+      "step": 15632
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.057302176415735e-05,
+      "loss": 0.9316,
+      "step": 15633
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0561810413413895e-05,
+      "loss": 0.7791,
+      "step": 15634
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0550600747898175e-05,
+      "loss": 1.2718,
+      "step": 15635
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0539392767882305e-05,
+      "loss": 1.0314,
+      "step": 15636
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0528186473638285e-05,
+      "loss": 0.8946,
+      "step": 15637
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0516981865438032e-05,
+      "loss": 0.9155,
+      "step": 15638
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0505778943553587e-05,
+      "loss": 0.7216,
+      "step": 15639
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0494577708256744e-05,
+      "loss": 0.8994,
+      "step": 15640
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0483378159819342e-05,
+      "loss": 1.2105,
+      "step": 15641
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0472180298513253e-05,
+      "loss": 1.163,
+      "step": 15642
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0460984124610213e-05,
+      "loss": 1.3315,
+      "step": 15643
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.044978963838191e-05,
+      "loss": 1.0919,
+      "step": 15644
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0438596840100107e-05,
+      "loss": 0.9045,
+      "step": 15645
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0427405730036396e-05,
+      "loss": 0.7249,
+      "step": 15646
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.04162163084624e-05,
+      "loss": 0.8873,
+      "step": 15647
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.040502857564963e-05,
+      "loss": 0.7695,
+      "step": 15648
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0393842531869675e-05,
+      "loss": 0.8745,
+      "step": 15649
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0382658177393986e-05,
+      "loss": 0.8634,
+      "step": 15650
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0371475512493975e-05,
+      "loss": 0.9162,
+      "step": 15651
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0360294537441092e-05,
+      "loss": 0.8174,
+      "step": 15652
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0349115252506666e-05,
+      "loss": 0.8707,
+      "step": 15653
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.033793765796201e-05,
+      "loss": 0.8378,
+      "step": 15654
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0326761754078402e-05,
+      "loss": 0.8985,
+      "step": 15655
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.031558754112708e-05,
+      "loss": 0.92,
+      "step": 15656
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0304415019379184e-05,
+      "loss": 0.9775,
+      "step": 15657
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.029324418910595e-05,
+      "loss": 0.8178,
+      "step": 15658
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0282075050578452e-05,
+      "loss": 0.7264,
+      "step": 15659
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.027090760406771e-05,
+      "loss": 0.7508,
+      "step": 15660
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0259741849844846e-05,
+      "loss": 0.8827,
+      "step": 15661
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0248577788180787e-05,
+      "loss": 0.8982,
+      "step": 15662
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.023741541934647e-05,
+      "loss": 0.9156,
+      "step": 15663
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0226254743612847e-05,
+      "loss": 0.741,
+      "step": 15664
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0215095761250757e-05,
+      "loss": 0.8104,
+      "step": 15665
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0203938472531012e-05,
+      "loss": 0.7243,
+      "step": 15666
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.019278287772438e-05,
+      "loss": 0.6664,
+      "step": 15667
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0181628977101682e-05,
+      "loss": 0.8771,
+      "step": 15668
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0170476770933498e-05,
+      "loss": 0.8105,
+      "step": 15669
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0159326259490572e-05,
+      "loss": 0.8847,
+      "step": 15670
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.01481774430435e-05,
+      "loss": 0.8595,
+      "step": 15671
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.013703032186286e-05,
+      "loss": 0.8858,
+      "step": 15672
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0125884896219136e-05,
+      "loss": 0.8084,
+      "step": 15673
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.01147411663829e-05,
+      "loss": 0.8587,
+      "step": 15674
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.010359913262457e-05,
+      "loss": 0.8286,
+      "step": 15675
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.009245879521453e-05,
+      "loss": 0.8297,
+      "step": 15676
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0081320154423208e-05,
+      "loss": 0.8604,
+      "step": 15677
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0070183210520897e-05,
+      "loss": 0.7663,
+      "step": 15678
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0059047963777865e-05,
+      "loss": 0.7993,
+      "step": 15679
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0047914414464408e-05,
+      "loss": 0.9976,
+      "step": 15680
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0036782562850708e-05,
+      "loss": 0.6932,
+      "step": 15681
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0025652409206928e-05,
+      "loss": 0.8701,
+      "step": 15682
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0014523953803187e-05,
+      "loss": 0.9449,
+      "step": 15683
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.0003397196909576e-05,
+      "loss": 0.9491,
+      "step": 15684
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9992272138796097e-05,
+      "loss": 0.776,
+      "step": 15685
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9981148779732816e-05,
+      "loss": 0.9083,
+      "step": 15686
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.997002711998965e-05,
+      "loss": 1.1793,
+      "step": 15687
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9958907159836524e-05,
+      "loss": 1.1076,
+      "step": 15688
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9947788899543283e-05,
+      "loss": 0.8502,
+      "step": 15689
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9936672339379825e-05,
+      "loss": 0.8328,
+      "step": 15690
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9925557479615905e-05,
+      "loss": 0.7808,
+      "step": 15691
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.991444432052124e-05,
+      "loss": 0.7998,
+      "step": 15692
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.990333286236562e-05,
+      "loss": 0.8304,
+      "step": 15693
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9892223105418672e-05,
+      "loss": 0.8425,
+      "step": 15694
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9881115049949982e-05,
+      "loss": 0.932,
+      "step": 15695
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9870008696229213e-05,
+      "loss": 0.8068,
+      "step": 15696
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9858904044525883e-05,
+      "loss": 0.8193,
+      "step": 15697
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9847801095109486e-05,
+      "loss": 0.8214,
+      "step": 15698
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9836699848249496e-05,
+      "loss": 0.9115,
+      "step": 15699
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9825600304215316e-05,
+      "loss": 0.812,
+      "step": 15700
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9814502463276305e-05,
+      "loss": 0.8292,
+      "step": 15701
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9803406325701868e-05,
+      "loss": 0.9836,
+      "step": 15702
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.979231189176126e-05,
+      "loss": 1.0094,
+      "step": 15703
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.978121916172375e-05,
+      "loss": 0.8407,
+      "step": 15704
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9770128135858523e-05,
+      "loss": 0.9545,
+      "step": 15705
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.975903881443479e-05,
+      "loss": 0.8473,
+      "step": 15706
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.974795119772168e-05,
+      "loss": 0.9108,
+      "step": 15707
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9736865285988234e-05,
+      "loss": 0.8369,
+      "step": 15708
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9725781079503577e-05,
+      "loss": 0.8398,
+      "step": 15709
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9714698578536683e-05,
+      "loss": 0.9181,
+      "step": 15710
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9703617783356484e-05,
+      "loss": 0.8777,
+      "step": 15711
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9692538694231986e-05,
+      "loss": 0.8768,
+      "step": 15712
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9681461311432002e-05,
+      "loss": 0.8355,
+      "step": 15713
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.967038563522536e-05,
+      "loss": 0.8239,
+      "step": 15714
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.965931166588093e-05,
+      "loss": 0.8626,
+      "step": 15715
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9648239403667434e-05,
+      "loss": 0.8893,
+      "step": 15716
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9637168848853557e-05,
+      "loss": 1.0373,
+      "step": 15717
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9626100001708046e-05,
+      "loss": 1.022,
+      "step": 15718
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9615032862499502e-05,
+      "loss": 0.7968,
+      "step": 15719
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9603967431496482e-05,
+      "loss": 0.7406,
+      "step": 15720
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.959290370896761e-05,
+      "loss": 0.9376,
+      "step": 15721
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9581841695181357e-05,
+      "loss": 0.7359,
+      "step": 15722
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.95707813904062e-05,
+      "loss": 0.7485,
+      "step": 15723
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.955972279491053e-05,
+      "loss": 0.7651,
+      "step": 15724
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.954866590896279e-05,
+      "loss": 0.8295,
+      "step": 15725
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.95376107328313e-05,
+      "loss": 0.9294,
+      "step": 15726
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9526557266784372e-05,
+      "loss": 1.2087,
+      "step": 15727
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9515505511090247e-05,
+      "loss": 1.0683,
+      "step": 15728
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9504455466017167e-05,
+      "loss": 0.9132,
+      "step": 15729
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9493407131833262e-05,
+      "loss": 0.8728,
+      "step": 15730
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.948236050880675e-05,
+      "loss": 0.8756,
+      "step": 15731
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9471315597205673e-05,
+      "loss": 0.8824,
+      "step": 15732
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9460272397298073e-05,
+      "loss": 0.8132,
+      "step": 15733
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9449230909352022e-05,
+      "loss": 0.9133,
+      "step": 15734
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.943819113363544e-05,
+      "loss": 0.6747,
+      "step": 15735
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9427153070416258e-05,
+      "loss": 0.7916,
+      "step": 15736
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9416116719962394e-05,
+      "loss": 0.8134,
+      "step": 15737
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9405082082541692e-05,
+      "loss": 0.8773,
+      "step": 15738
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.939404915842193e-05,
+      "loss": 0.9102,
+      "step": 15739
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.938301794787086e-05,
+      "loss": 0.7499,
+      "step": 15740
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9371988451156253e-05,
+      "loss": 0.8816,
+      "step": 15741
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9360960668545768e-05,
+      "loss": 0.9613,
+      "step": 15742
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9349934600307037e-05,
+      "loss": 0.763,
+      "step": 15743
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9338910246707662e-05,
+      "loss": 0.8082,
+      "step": 15744
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9327887608015193e-05,
+      "loss": 0.7832,
+      "step": 15745
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9316866684497114e-05,
+      "loss": 0.748,
+      "step": 15746
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9305847476420957e-05,
+      "loss": 0.849,
+      "step": 15747
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.929482998405413e-05,
+      "loss": 0.8849,
+      "step": 15748
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9283814207663985e-05,
+      "loss": 0.9668,
+      "step": 15749
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9272800147517922e-05,
+      "loss": 0.8432,
+      "step": 15750
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.926178780388322e-05,
+      "loss": 0.8577,
+      "step": 15751
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9250777177027133e-05,
+      "loss": 0.7913,
+      "step": 15752
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9239768267216916e-05,
+      "loss": 0.7405,
+      "step": 15753
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9228761074719735e-05,
+      "loss": 0.8863,
+      "step": 15754
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.921775559980271e-05,
+      "loss": 0.9735,
+      "step": 15755
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.920675184273297e-05,
+      "loss": 0.8565,
+      "step": 15756
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.919574980377754e-05,
+      "loss": 0.8765,
+      "step": 15757
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9184749483203422e-05,
+      "loss": 0.7831,
+      "step": 15758
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9173750881277627e-05,
+      "loss": 0.7487,
+      "step": 15759
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9162753998267088e-05,
+      "loss": 0.8663,
+      "step": 15760
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9151758834438668e-05,
+      "loss": 0.8396,
+      "step": 15761
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.914076539005919e-05,
+      "loss": 0.7876,
+      "step": 15762
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.912977366539552e-05,
+      "loss": 0.8524,
+      "step": 15763
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.911878366071439e-05,
+      "loss": 0.8457,
+      "step": 15764
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9107795376282487e-05,
+      "loss": 1.3236,
+      "step": 15765
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.909680881236656e-05,
+      "loss": 1.1547,
+      "step": 15766
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9085823969233216e-05,
+      "loss": 0.8806,
+      "step": 15767
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9074840847149022e-05,
+      "loss": 0.9364,
+      "step": 15768
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9063859446380582e-05,
+      "loss": 0.8934,
+      "step": 15769
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9052879767194397e-05,
+      "loss": 0.8421,
+      "step": 15770
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.904190180985692e-05,
+      "loss": 0.9526,
+      "step": 15771
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9030925574634584e-05,
+      "loss": 0.9301,
+      "step": 15772
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9019951061793784e-05,
+      "loss": 0.9401,
+      "step": 15773
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.900897827160083e-05,
+      "loss": 0.8919,
+      "step": 15774
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8998007204322085e-05,
+      "loss": 0.7735,
+      "step": 15775
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8987037860223774e-05,
+      "loss": 0.9186,
+      "step": 15776
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.897607023957214e-05,
+      "loss": 0.7418,
+      "step": 15777
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.896510434263331e-05,
+      "loss": 0.6404,
+      "step": 15778
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8954140169673493e-05,
+      "loss": 0.7641,
+      "step": 15779
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.894317772095875e-05,
+      "loss": 0.9147,
+      "step": 15780
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8932216996755092e-05,
+      "loss": 0.9306,
+      "step": 15781
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.89212579973286e-05,
+      "loss": 0.6521,
+      "step": 15782
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8910300722945226e-05,
+      "loss": 0.8647,
+      "step": 15783
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.889934517387084e-05,
+      "loss": 0.7651,
+      "step": 15784
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8888391350371413e-05,
+      "loss": 0.9389,
+      "step": 15785
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8877439252712767e-05,
+      "loss": 0.7592,
+      "step": 15786
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8866488881160626e-05,
+      "loss": 0.7013,
+      "step": 15787
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.885554023598084e-05,
+      "loss": 0.7327,
+      "step": 15788
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8844593317439094e-05,
+      "loss": 0.7997,
+      "step": 15789
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8833648125801028e-05,
+      "loss": 0.8658,
+      "step": 15790
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8822704661332346e-05,
+      "loss": 1.0307,
+      "step": 15791
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8811762924298612e-05,
+      "loss": 0.8303,
+      "step": 15792
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8800822914965332e-05,
+      "loss": 0.8101,
+      "step": 15793
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8789884633598086e-05,
+      "loss": 0.6111,
+      "step": 15794
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8778948080462297e-05,
+      "loss": 0.7511,
+      "step": 15795
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.876801325582341e-05,
+      "loss": 0.9437,
+      "step": 15796
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8757080159946758e-05,
+      "loss": 0.922,
+      "step": 15797
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.874614879309775e-05,
+      "loss": 0.9599,
+      "step": 15798
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.873521915554166e-05,
+      "loss": 0.6822,
+      "step": 15799
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8724291247543733e-05,
+      "loss": 0.7612,
+      "step": 15800
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8713365069369192e-05,
+      "loss": 0.9994,
+      "step": 15801
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.870244062128321e-05,
+      "loss": 0.7824,
+      "step": 15802
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.869151790355088e-05,
+      "loss": 0.7476,
+      "step": 15803
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8680596916437353e-05,
+      "loss": 0.9039,
+      "step": 15804
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8669677660207638e-05,
+      "loss": 0.7805,
+      "step": 15805
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8658760135126727e-05,
+      "loss": 0.9449,
+      "step": 15806
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.864784434145964e-05,
+      "loss": 0.846,
+      "step": 15807
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8636930279471252e-05,
+      "loss": 0.951,
+      "step": 15808
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.862601794942642e-05,
+      "loss": 0.8253,
+      "step": 15809
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8615107351590053e-05,
+      "loss": 0.9957,
+      "step": 15810
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8604198486226898e-05,
+      "loss": 0.9026,
+      "step": 15811
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8593291353601705e-05,
+      "loss": 0.8169,
+      "step": 15812
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8582385953979173e-05,
+      "loss": 0.8624,
+      "step": 15813
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8571482287624006e-05,
+      "loss": 0.9055,
+      "step": 15814
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8560580354800825e-05,
+      "loss": 0.8678,
+      "step": 15815
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8549680155774205e-05,
+      "loss": 0.8099,
+      "step": 15816
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.853878169080868e-05,
+      "loss": 0.9564,
+      "step": 15817
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.852788496016875e-05,
+      "loss": 0.7973,
+      "step": 15818
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8516989964118855e-05,
+      "loss": 0.8396,
+      "step": 15819
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8506096702923468e-05,
+      "loss": 0.8161,
+      "step": 15820
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8495205176846918e-05,
+      "loss": 0.7797,
+      "step": 15821
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8484315386153527e-05,
+      "loss": 0.9278,
+      "step": 15822
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8473427331107627e-05,
+      "loss": 0.8282,
+      "step": 15823
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8462541011973454e-05,
+      "loss": 0.7028,
+      "step": 15824
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.845165642901516e-05,
+      "loss": 0.8834,
+      "step": 15825
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8440773582496992e-05,
+      "loss": 0.8612,
+      "step": 15826
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8429892472683018e-05,
+      "loss": 0.9,
+      "step": 15827
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8419013099837333e-05,
+      "loss": 0.9406,
+      "step": 15828
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.840813546422394e-05,
+      "loss": 0.8865,
+      "step": 15829
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8397259566106916e-05,
+      "loss": 0.8347,
+      "step": 15830
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8386385405750105e-05,
+      "loss": 0.9175,
+      "step": 15831
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8375512983417496e-05,
+      "loss": 0.925,
+      "step": 15832
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8364642299372934e-05,
+      "loss": 0.9095,
+      "step": 15833
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.835377335388024e-05,
+      "loss": 0.7559,
+      "step": 15834
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8342906147203185e-05,
+      "loss": 0.8124,
+      "step": 15835
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8332040679605544e-05,
+      "loss": 0.8366,
+      "step": 15836
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.832117695135099e-05,
+      "loss": 0.89,
+      "step": 15837
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8310314962703167e-05,
+      "loss": 0.8582,
+      "step": 15838
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.829945471392573e-05,
+      "loss": 0.771,
+      "step": 15839
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8288596205282248e-05,
+      "loss": 0.7646,
+      "step": 15840
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.82777394370362e-05,
+      "loss": 1.2668,
+      "step": 15841
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8266884409451134e-05,
+      "loss": 1.0588,
+      "step": 15842
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.825603112279046e-05,
+      "loss": 0.8524,
+      "step": 15843
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8245179577317604e-05,
+      "loss": 0.7858,
+      "step": 15844
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.823432977329592e-05,
+      "loss": 0.9473,
+      "step": 15845
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.822348171098872e-05,
+      "loss": 0.8594,
+      "step": 15846
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8212635390659246e-05,
+      "loss": 0.7507,
+      "step": 15847
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8201790812570793e-05,
+      "loss": 0.8191,
+      "step": 15848
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8190947976986548e-05,
+      "loss": 0.8127,
+      "step": 15849
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.81801068841696e-05,
+      "loss": 0.8543,
+      "step": 15850
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8169267534383127e-05,
+      "loss": 0.668,
+      "step": 15851
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.8158429927890174e-05,
+      "loss": 1.3008,
+      "step": 15852
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8147594064953754e-05,
+      "loss": 1.0946,
+      "step": 15853
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.813675994583681e-05,
+      "loss": 0.8322,
+      "step": 15854
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.812592757080236e-05,
+      "loss": 0.937,
+      "step": 15855
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.811509694011325e-05,
+      "loss": 0.8193,
+      "step": 15856
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8104268054032324e-05,
+      "loss": 0.827,
+      "step": 15857
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8093440912822434e-05,
+      "loss": 0.8503,
+      "step": 15858
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.808261551674637e-05,
+      "loss": 0.9362,
+      "step": 15859
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8071791866066743e-05,
+      "loss": 0.7191,
+      "step": 15860
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8060969961046346e-05,
+      "loss": 0.9656,
+      "step": 15861
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8050149801947777e-05,
+      "loss": 0.7862,
+      "step": 15862
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.803933138903362e-05,
+      "loss": 0.7948,
+      "step": 15863
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.802851472256649e-05,
+      "loss": 0.7574,
+      "step": 15864
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.801769980280886e-05,
+      "loss": 0.9178,
+      "step": 15865
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.8006886630023177e-05,
+      "loss": 0.7482,
+      "step": 15866
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.799607520447195e-05,
+      "loss": 0.8149,
+      "step": 15867
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7985265526417504e-05,
+      "loss": 0.8761,
+      "step": 15868
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7974457596122205e-05,
+      "loss": 0.9975,
+      "step": 15869
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7963651413848325e-05,
+      "loss": 0.8616,
+      "step": 15870
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7952846979858173e-05,
+      "loss": 0.8262,
+      "step": 15871
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.794204429441395e-05,
+      "loss": 0.8214,
+      "step": 15872
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.793124335777779e-05,
+      "loss": 0.7981,
+      "step": 15873
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7920444170211923e-05,
+      "loss": 0.7483,
+      "step": 15874
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7909646731978355e-05,
+      "loss": 0.7388,
+      "step": 15875
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7898851043339126e-05,
+      "loss": 0.802,
+      "step": 15876
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7888057104556297e-05,
+      "loss": 0.9179,
+      "step": 15877
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.787726491589181e-05,
+      "loss": 0.7258,
+      "step": 15878
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.786647447760754e-05,
+      "loss": 0.7592,
+      "step": 15879
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7855685789965458e-05,
+      "loss": 0.794,
+      "step": 15880
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7844898853227343e-05,
+      "loss": 0.9126,
+      "step": 15881
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7834113667654948e-05,
+      "loss": 0.8576,
+      "step": 15882
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7823330233510115e-05,
+      "loss": 0.7487,
+      "step": 15883
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7812548551054505e-05,
+      "loss": 0.8642,
+      "step": 15884
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7801768620549796e-05,
+      "loss": 0.7763,
+      "step": 15885
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.779099044225756e-05,
+      "loss": 0.8737,
+      "step": 15886
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7780214016439443e-05,
+      "loss": 0.9777,
+      "step": 15887
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7769439343356974e-05,
+      "loss": 0.7439,
+      "step": 15888
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7758666423271616e-05,
+      "loss": 0.9048,
+      "step": 15889
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7747895256444855e-05,
+      "loss": 0.8088,
+      "step": 15890
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7737125843138077e-05,
+      "loss": 0.8149,
+      "step": 15891
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7726358183612634e-05,
+      "loss": 0.8461,
+      "step": 15892
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.771559227812991e-05,
+      "loss": 0.9092,
+      "step": 15893
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7704828126951143e-05,
+      "loss": 0.8108,
+      "step": 15894
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.769406573033756e-05,
+      "loss": 0.8074,
+      "step": 15895
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7683305088550417e-05,
+      "loss": 0.7738,
+      "step": 15896
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7672546201850823e-05,
+      "loss": 0.9198,
+      "step": 15897
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7661789070499887e-05,
+      "loss": 0.7537,
+      "step": 15898
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.765103369475871e-05,
+      "loss": 0.9037,
+      "step": 15899
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7640280074888313e-05,
+      "loss": 0.9757,
+      "step": 15900
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7629528211149668e-05,
+      "loss": 0.7916,
+      "step": 15901
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7618778103803688e-05,
+      "loss": 0.918,
+      "step": 15902
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7608029753111362e-05,
+      "loss": 0.8054,
+      "step": 15903
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.759728315933344e-05,
+      "loss": 0.8452,
+      "step": 15904
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7586538322730803e-05,
+      "loss": 0.8677,
+      "step": 15905
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7575795243564216e-05,
+      "loss": 0.8834,
+      "step": 15906
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7565053922094388e-05,
+      "loss": 0.806,
+      "step": 15907
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.755431435858199e-05,
+      "loss": 1.0114,
+      "step": 15908
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7543576553287707e-05,
+      "loss": 0.6697,
+      "step": 15909
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7532840506472134e-05,
+      "loss": 0.8397,
+      "step": 15910
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7522106218395782e-05,
+      "loss": 0.8514,
+      "step": 15911
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.751137368931923e-05,
+      "loss": 0.8123,
+      "step": 15912
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.750064291950292e-05,
+      "loss": 1.1865,
+      "step": 15913
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7489913909207254e-05,
+      "loss": 1.1914,
+      "step": 15914
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.747918665869269e-05,
+      "loss": 0.907,
+      "step": 15915
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7468461168219527e-05,
+      "loss": 0.7051,
+      "step": 15916
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.745773743804807e-05,
+      "loss": 0.758,
+      "step": 15917
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7447015468438576e-05,
+      "loss": 0.7364,
+      "step": 15918
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.743629525965127e-05,
+      "loss": 0.6918,
+      "step": 15919
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7425576811946296e-05,
+      "loss": 0.903,
+      "step": 15920
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.741486012558383e-05,
+      "loss": 0.6835,
+      "step": 15921
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.740414520082395e-05,
+      "loss": 0.8723,
+      "step": 15922
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7393432037926658e-05,
+      "loss": 0.757,
+      "step": 15923
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7382720637152026e-05,
+      "loss": 0.8927,
+      "step": 15924
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7372010998759966e-05,
+      "loss": 0.7278,
+      "step": 15925
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7361303123010417e-05,
+      "loss": 0.8896,
+      "step": 15926
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7350597010163213e-05,
+      "loss": 0.887,
+      "step": 15927
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.733989266047825e-05,
+      "loss": 0.845,
+      "step": 15928
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7329190074215284e-05,
+      "loss": 0.7356,
+      "step": 15929
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7318489251634027e-05,
+      "loss": 0.7552,
+      "step": 15930
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7307790192994254e-05,
+      "loss": 1.1389,
+      "step": 15931
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7297092898555588e-05,
+      "loss": 1.0184,
+      "step": 15932
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.728639736857763e-05,
+      "loss": 0.7895,
+      "step": 15933
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.727570360331999e-05,
+      "loss": 0.8386,
+      "step": 15934
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7265011603042167e-05,
+      "loss": 1.281,
+      "step": 15935
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7254321368003644e-05,
+      "loss": 1.0883,
+      "step": 15936
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7243632898463912e-05,
+      "loss": 1.3284,
+      "step": 15937
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.723294619468235e-05,
+      "loss": 1.1818,
+      "step": 15938
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.722226125691829e-05,
+      "loss": 0.8407,
+      "step": 15939
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.721157808543111e-05,
+      "loss": 0.9603,
+      "step": 15940
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.720089668048006e-05,
+      "loss": 0.8061,
+      "step": 15941
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7190217042324362e-05,
+      "loss": 0.8565,
+      "step": 15942
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7179539171223166e-05,
+      "loss": 1.0076,
+      "step": 15943
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.71688630674357e-05,
+      "loss": 1.2524,
+      "step": 15944
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7158188731221023e-05,
+      "loss": 1.1696,
+      "step": 15945
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7147516162838182e-05,
+      "loss": 0.9157,
+      "step": 15946
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.713684536254626e-05,
+      "loss": 1.2802,
+      "step": 15947
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7126176330604158e-05,
+      "loss": 0.9576,
+      "step": 15948
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7115509067270807e-05,
+      "loss": 0.825,
+      "step": 15949
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.710484357280515e-05,
+      "loss": 0.6993,
+      "step": 15950
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7094179847466016e-05,
+      "loss": 0.9543,
+      "step": 15951
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7083517891512154e-05,
+      "loss": 0.9898,
+      "step": 15952
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7072857705202405e-05,
+      "loss": 0.8248,
+      "step": 15953
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7062199288795443e-05,
+      "loss": 0.8579,
+      "step": 15954
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7051542642549922e-05,
+      "loss": 0.9247,
+      "step": 15955
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.7040887766724532e-05,
+      "loss": 0.8065,
+      "step": 15956
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.7030234661577825e-05,
+      "loss": 0.7964,
+      "step": 15957
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.701958332736835e-05,
+      "loss": 0.8077,
+      "step": 15958
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.700893376435457e-05,
+      "loss": 0.7991,
+      "step": 15959
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6998285972795023e-05,
+      "loss": 0.8245,
+      "step": 15960
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6987639952948084e-05,
+      "loss": 0.957,
+      "step": 15961
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.697699570507213e-05,
+      "loss": 0.8308,
+      "step": 15962
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.696635322942548e-05,
+      "loss": 0.8539,
+      "step": 15963
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.695571252626643e-05,
+      "loss": 0.8931,
+      "step": 15964
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6945073595853197e-05,
+      "loss": 0.9366,
+      "step": 15965
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6934436438444043e-05,
+      "loss": 1.0295,
+      "step": 15966
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6923801054297083e-05,
+      "loss": 0.9548,
+      "step": 15967
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6913167443670418e-05,
+      "loss": 0.843,
+      "step": 15968
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6902535606822177e-05,
+      "loss": 0.8253,
+      "step": 15969
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6891905544010342e-05,
+      "loss": 0.725,
+      "step": 15970
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6881277255492886e-05,
+      "loss": 0.837,
+      "step": 15971
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6870650741527814e-05,
+      "loss": 0.9293,
+      "step": 15972
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6860026002372985e-05,
+      "loss": 0.7778,
+      "step": 15973
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6849403038286268e-05,
+      "loss": 0.8189,
+      "step": 15974
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.683878184952543e-05,
+      "loss": 1.3935,
+      "step": 15975
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6828162436348316e-05,
+      "loss": 1.5891,
+      "step": 15976
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6817544799012616e-05,
+      "loss": 1.1371,
+      "step": 15977
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6806928937776023e-05,
+      "loss": 0.8387,
+      "step": 15978
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6796314852896165e-05,
+      "loss": 0.7965,
+      "step": 15979
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.678570254463064e-05,
+      "loss": 0.9212,
+      "step": 15980
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6775092013236992e-05,
+      "loss": 0.9073,
+      "step": 15981
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6764483258972783e-05,
+      "loss": 1.2715,
+      "step": 15982
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6753876282095435e-05,
+      "loss": 1.0872,
+      "step": 15983
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6743271082862376e-05,
+      "loss": 0.8716,
+      "step": 15984
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.673266766153102e-05,
+      "loss": 0.8587,
+      "step": 15985
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6722066018358694e-05,
+      "loss": 0.8438,
+      "step": 15986
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.671146615360265e-05,
+      "loss": 0.7972,
+      "step": 15987
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.670086806752021e-05,
+      "loss": 0.9029,
+      "step": 15988
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6690271760368556e-05,
+      "loss": 1.205,
+      "step": 15989
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.667967723240482e-05,
+      "loss": 1.0025,
+      "step": 15990
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6669084483886187e-05,
+      "loss": 1.2564,
+      "step": 15991
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6658493515069737e-05,
+      "loss": 1.0645,
+      "step": 15992
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6647904326212426e-05,
+      "loss": 0.8345,
+      "step": 15993
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.663731691757132e-05,
+      "loss": 0.8006,
+      "step": 15994
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.662673128940335e-05,
+      "loss": 0.7682,
+      "step": 15995
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.661614744196539e-05,
+      "loss": 0.6811,
+      "step": 15996
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6605565375514374e-05,
+      "loss": 0.8398,
+      "step": 15997
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.659498509030708e-05,
+      "loss": 0.7634,
+      "step": 15998
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6584406586600296e-05,
+      "loss": 0.9149,
+      "step": 15999
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6573829864650713e-05,
+      "loss": 0.8238,
+      "step": 16000
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.656325492471511e-05,
+      "loss": 0.801,
+      "step": 16001
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.655268176705007e-05,
+      "loss": 0.764,
+      "step": 16002
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.654211039191219e-05,
+      "loss": 0.7879,
+      "step": 16003
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6531540799558087e-05,
+      "loss": 0.9292,
+      "step": 16004
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6520972990244252e-05,
+      "loss": 0.7741,
+      "step": 16005
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.651040696422715e-05,
+      "loss": 0.7755,
+      "step": 16006
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6499842721763234e-05,
+      "loss": 0.9173,
+      "step": 16007
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6489280263108863e-05,
+      "loss": 0.8335,
+      "step": 16008
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6478719588520372e-05,
+      "loss": 0.7469,
+      "step": 16009
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.646816069825413e-05,
+      "loss": 0.9823,
+      "step": 16010
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6457603592566337e-05,
+      "loss": 0.8558,
+      "step": 16011
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6447048271713205e-05,
+      "loss": 0.9283,
+      "step": 16012
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6436494735950956e-05,
+      "loss": 0.8898,
+      "step": 16013
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.642594298553569e-05,
+      "loss": 0.9352,
+      "step": 16014
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.641539302072349e-05,
+      "loss": 0.7906,
+      "step": 16015
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.640484484177036e-05,
+      "loss": 0.8138,
+      "step": 16016
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6394298448932385e-05,
+      "loss": 0.8712,
+      "step": 16017
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6383753842465466e-05,
+      "loss": 0.8618,
+      "step": 16018
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6373211022625487e-05,
+      "loss": 0.8639,
+      "step": 16019
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6362669989668386e-05,
+      "loss": 0.7883,
+      "step": 16020
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.635213074384999e-05,
+      "loss": 0.8048,
+      "step": 16021
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.634159328542598e-05,
+      "loss": 0.9158,
+      "step": 16022
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6331057614652187e-05,
+      "loss": 0.8379,
+      "step": 16023
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6320523731784285e-05,
+      "loss": 0.8198,
+      "step": 16024
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.630999163707788e-05,
+      "loss": 0.8795,
+      "step": 16025
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6299461330788654e-05,
+      "loss": 0.8066,
+      "step": 16026
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.628893281317215e-05,
+      "loss": 0.871,
+      "step": 16027
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.627840608448383e-05,
+      "loss": 0.8862,
+      "step": 16028
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6267881144979255e-05,
+      "loss": 0.726,
+      "step": 16029
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6257357994913834e-05,
+      "loss": 0.8849,
+      "step": 16030
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6246836634542938e-05,
+      "loss": 0.8236,
+      "step": 16031
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6236317064121896e-05,
+      "loss": 0.8878,
+      "step": 16032
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6225799283906073e-05,
+      "loss": 0.784,
+      "step": 16033
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6215283294150693e-05,
+      "loss": 0.7484,
+      "step": 16034
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.620476909511096e-05,
+      "loss": 0.8388,
+      "step": 16035
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.619425668704212e-05,
+      "loss": 0.845,
+      "step": 16036
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.618374607019921e-05,
+      "loss": 0.8646,
+      "step": 16037
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6173237244837347e-05,
+      "loss": 0.7906,
+      "step": 16038
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6162730211211594e-05,
+      "loss": 0.7752,
+      "step": 16039
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6152224969576953e-05,
+      "loss": 1.2178,
+      "step": 16040
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6141721520188345e-05,
+      "loss": 1.0696,
+      "step": 16041
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6131219863300727e-05,
+      "loss": 0.9155,
+      "step": 16042
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6120719999168952e-05,
+      "loss": 0.7638,
+      "step": 16043
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6110221928047817e-05,
+      "loss": 0.7364,
+      "step": 16044
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6099725650192165e-05,
+      "loss": 0.8701,
+      "step": 16045
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6089231165856697e-05,
+      "loss": 0.8866,
+      "step": 16046
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6078738475296105e-05,
+      "loss": 0.8729,
+      "step": 16047
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6068247578765027e-05,
+      "loss": 0.8165,
+      "step": 16048
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6057758476518125e-05,
+      "loss": 0.7201,
+      "step": 16049
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6047271168809927e-05,
+      "loss": 0.8411,
+      "step": 16050
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.603678565589496e-05,
+      "loss": 0.8294,
+      "step": 16051
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.60263019380277e-05,
+      "loss": 0.8664,
+      "step": 16052
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6015820015462557e-05,
+      "loss": 0.7892,
+      "step": 16053
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.6005339888453973e-05,
+      "loss": 0.7365,
+      "step": 16054
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5994861557256268e-05,
+      "loss": 0.7442,
+      "step": 16055
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.598438502212375e-05,
+      "loss": 0.8972,
+      "step": 16056
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5973910283310633e-05,
+      "loss": 0.8401,
+      "step": 16057
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5963437341071217e-05,
+      "loss": 0.8852,
+      "step": 16058
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5952966195659634e-05,
+      "loss": 0.752,
+      "step": 16059
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5942496847329978e-05,
+      "loss": 0.8134,
+      "step": 16060
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5932029296336402e-05,
+      "loss": 0.8954,
+      "step": 16061
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5921563542932915e-05,
+      "loss": 0.6522,
+      "step": 16062
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.591109958737349e-05,
+      "loss": 0.859,
+      "step": 16063
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5900637429912133e-05,
+      "loss": 0.9277,
+      "step": 16064
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5890177070802758e-05,
+      "loss": 0.8389,
+      "step": 16065
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5879718510299134e-05,
+      "loss": 0.9421,
+      "step": 16066
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5869261748655193e-05,
+      "loss": 0.8403,
+      "step": 16067
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.585880678612468e-05,
+      "loss": 0.7249,
+      "step": 16068
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5848353622961287e-05,
+      "loss": 0.7748,
+      "step": 16069
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.583790225941879e-05,
+      "loss": 0.871,
+      "step": 16070
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5827452695750788e-05,
+      "loss": 0.7391,
+      "step": 16071
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5817004932210887e-05,
+      "loss": 0.7542,
+      "step": 16072
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.580655896905263e-05,
+      "loss": 0.8573,
+      "step": 16073
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5796114806529594e-05,
+      "loss": 0.7861,
+      "step": 16074
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5785672444895215e-05,
+      "loss": 0.7899,
+      "step": 16075
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5775231884402894e-05,
+      "loss": 0.7774,
+      "step": 16076
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5764793125306085e-05,
+      "loss": 0.893,
+      "step": 16077
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5754356167858096e-05,
+      "loss": 0.8734,
+      "step": 16078
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.57439210123122e-05,
+      "loss": 0.9569,
+      "step": 16079
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.573348765892174e-05,
+      "loss": 0.764,
+      "step": 16080
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5723056107939835e-05,
+      "loss": 0.917,
+      "step": 16081
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5712626359619662e-05,
+      "loss": 0.7727,
+      "step": 16082
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5702198414214396e-05,
+      "loss": 0.8661,
+      "step": 16083
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.56917722719771e-05,
+      "loss": 0.8256,
+      "step": 16084
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.568134793316076e-05,
+      "loss": 0.7926,
+      "step": 16085
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.567092539801845e-05,
+      "loss": 0.8111,
+      "step": 16086
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5660504666803076e-05,
+      "loss": 0.8889,
+      "step": 16087
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.565008573976755e-05,
+      "loss": 0.8745,
+      "step": 16088
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5639668617164702e-05,
+      "loss": 0.9269,
+      "step": 16089
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5629253299247413e-05,
+      "loss": 0.7668,
+      "step": 16090
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5618839786268433e-05,
+      "loss": 0.8915,
+      "step": 16091
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.560842807848044e-05,
+      "loss": 0.7601,
+      "step": 16092
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5598018176136196e-05,
+      "loss": 1.2861,
+      "step": 16093
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5587610079488356e-05,
+      "loss": 1.1144,
+      "step": 16094
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5577203788789404e-05,
+      "loss": 0.7243,
+      "step": 16095
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5566799304292e-05,
+      "loss": 0.9709,
+      "step": 16096
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5556396626248613e-05,
+      "loss": 0.6814,
+      "step": 16097
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.55459957549117e-05,
+      "loss": 0.9548,
+      "step": 16098
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5535596690533726e-05,
+      "loss": 0.7928,
+      "step": 16099
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.552519943336705e-05,
+      "loss": 0.9019,
+      "step": 16100
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5514803983663983e-05,
+      "loss": 0.9021,
+      "step": 16101
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5504410341676865e-05,
+      "loss": 0.8522,
+      "step": 16102
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5494018507657924e-05,
+      "loss": 0.9252,
+      "step": 16103
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5483628481859355e-05,
+      "loss": 0.7511,
+      "step": 16104
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5473240264533293e-05,
+      "loss": 0.861,
+      "step": 16105
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5462853855931924e-05,
+      "loss": 0.7136,
+      "step": 16106
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5452469256307266e-05,
+      "loss": 0.8857,
+      "step": 16107
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.544208646591134e-05,
+      "loss": 0.8618,
+      "step": 16108
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5431705484996215e-05,
+      "loss": 0.8336,
+      "step": 16109
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.542132631381374e-05,
+      "loss": 0.81,
+      "step": 16110
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.541094895261582e-05,
+      "loss": 0.8475,
+      "step": 16111
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5400573401654348e-05,
+      "loss": 0.7716,
+      "step": 16112
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5390199661181125e-05,
+      "loss": 0.9344,
+      "step": 16113
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537982773144787e-05,
+      "loss": 0.8224,
+      "step": 16114
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5369457612706372e-05,
+      "loss": 0.8122,
+      "step": 16115
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5359089305208284e-05,
+      "loss": 1.0164,
+      "step": 16116
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5348722809205195e-05,
+      "loss": 0.9734,
+      "step": 16117
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.533835812494877e-05,
+      "loss": 0.7806,
+      "step": 16118
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5327995252690506e-05,
+      "loss": 0.8194,
+      "step": 16119
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5317634192681916e-05,
+      "loss": 0.8591,
+      "step": 16120
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.530727494517443e-05,
+      "loss": 0.7623,
+      "step": 16121
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5296917510419514e-05,
+      "loss": 0.8039,
+      "step": 16122
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5286561888668512e-05,
+      "loss": 0.8207,
+      "step": 16123
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.527620808017276e-05,
+      "loss": 0.7278,
+      "step": 16124
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5265856085183516e-05,
+      "loss": 0.7103,
+      "step": 16125
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5255505903951992e-05,
+      "loss": 0.8918,
+      "step": 16126
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5245157536729457e-05,
+      "loss": 0.9099,
+      "step": 16127
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.523481098376702e-05,
+      "loss": 0.825,
+      "step": 16128
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.522446624531579e-05,
+      "loss": 0.8425,
+      "step": 16129
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.521412332162679e-05,
+      "loss": 0.8146,
+      "step": 16130
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.520378221295111e-05,
+      "loss": 0.8069,
+      "step": 16131
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.519344291953969e-05,
+      "loss": 0.8972,
+      "step": 16132
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.518310544164342e-05,
+      "loss": 0.9284,
+      "step": 16133
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.517276977951325e-05,
+      "loss": 0.8128,
+      "step": 16134
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5162435933399998e-05,
+      "loss": 0.8859,
+      "step": 16135
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5152103903554423e-05,
+      "loss": 1.2483,
+      "step": 16136
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.514177369022733e-05,
+      "loss": 1.2169,
+      "step": 16137
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.513144529366942e-05,
+      "loss": 0.7946,
+      "step": 16138
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.512111871413134e-05,
+      "loss": 1.3035,
+      "step": 16139
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.511079395186372e-05,
+      "loss": 1.0052,
+      "step": 16140
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5100471007117122e-05,
+      "loss": 0.7867,
+      "step": 16141
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5090149880142065e-05,
+      "loss": 0.9402,
+      "step": 16142
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5079830571189087e-05,
+      "loss": 0.801,
+      "step": 16143
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.506951308050861e-05,
+      "loss": 0.7967,
+      "step": 16144
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.505919740835103e-05,
+      "loss": 0.7882,
+      "step": 16145
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5048883554966673e-05,
+      "loss": 0.8403,
+      "step": 16146
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5038571520605913e-05,
+      "loss": 0.8763,
+      "step": 16147
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5028261305518976e-05,
+      "loss": 0.8377,
+      "step": 16148
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5017952909956066e-05,
+      "loss": 0.8455,
+      "step": 16149
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5007646334167423e-05,
+      "loss": 0.8358,
+      "step": 16150
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4997341578403154e-05,
+      "loss": 0.756,
+      "step": 16151
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4987038642913308e-05,
+      "loss": 0.8525,
+      "step": 16152
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4976737527948035e-05,
+      "loss": 1.2181,
+      "step": 16153
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.496643823375724e-05,
+      "loss": 1.1142,
+      "step": 16154
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4956140760590886e-05,
+      "loss": 0.8947,
+      "step": 16155
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.494584510869894e-05,
+      "loss": 0.8362,
+      "step": 16156
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4935551278331247e-05,
+      "loss": 0.7961,
+      "step": 16157
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.492525926973761e-05,
+      "loss": 0.824,
+      "step": 16158
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.491496908316785e-05,
+      "loss": 0.7968,
+      "step": 16159
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4904680718871696e-05,
+      "loss": 0.8793,
+      "step": 16160
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4894394177098835e-05,
+      "loss": 0.8285,
+      "step": 16161
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.488410945809887e-05,
+      "loss": 0.7918,
+      "step": 16162
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4873826562121493e-05,
+      "loss": 0.8116,
+      "step": 16163
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.4863545489416206e-05,
+      "loss": 0.8355,
+      "step": 16164
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.485326624023252e-05,
+      "loss": 0.886,
+      "step": 16165
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4842988814819955e-05,
+      "loss": 1.0002,
+      "step": 16166
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4832713213427904e-05,
+      "loss": 0.8141,
+      "step": 16167
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4822439436305768e-05,
+      "loss": 1.0428,
+      "step": 16168
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4812167483702863e-05,
+      "loss": 0.8484,
+      "step": 16169
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4801897355868498e-05,
+      "loss": 0.7121,
+      "step": 16170
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.479162905305189e-05,
+      "loss": 1.0043,
+      "step": 16171
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4781362575502308e-05,
+      "loss": 0.838,
+      "step": 16172
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4771097923468878e-05,
+      "loss": 0.7715,
+      "step": 16173
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4760835097200698e-05,
+      "loss": 1.2877,
+      "step": 16174
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4750574096946887e-05,
+      "loss": 1.1912,
+      "step": 16175
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4740314922956453e-05,
+      "loss": 0.7575,
+      "step": 16176
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4730057575478384e-05,
+      "loss": 0.9361,
+      "step": 16177
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4719802054761577e-05,
+      "loss": 0.9076,
+      "step": 16178
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4709548361055003e-05,
+      "loss": 0.7598,
+      "step": 16179
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4699296494607483e-05,
+      "loss": 0.8559,
+      "step": 16180
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4689046455667774e-05,
+      "loss": 0.643,
+      "step": 16181
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4678798244484723e-05,
+      "loss": 1.3531,
+      "step": 16182
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4668551861307044e-05,
+      "loss": 1.2104,
+      "step": 16183
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.465830730638331e-05,
+      "loss": 1.2474,
+      "step": 16184
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.464806457996225e-05,
+      "loss": 1.1005,
+      "step": 16185
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4637823682292404e-05,
+      "loss": 0.8052,
+      "step": 16186
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4627584613622313e-05,
+      "loss": 0.6706,
+      "step": 16187
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4617347374200505e-05,
+      "loss": 0.8206,
+      "step": 16188
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.460711196427542e-05,
+      "loss": 0.7585,
+      "step": 16189
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4596878384095433e-05,
+      "loss": 0.9135,
+      "step": 16190
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.458664663390896e-05,
+      "loss": 0.8606,
+      "step": 16191
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.457641671396429e-05,
+      "loss": 0.7649,
+      "step": 16192
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.456618862450968e-05,
+      "loss": 0.8757,
+      "step": 16193
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4555962365793417e-05,
+      "loss": 0.8442,
+      "step": 16194
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.454573793806364e-05,
+      "loss": 0.9533,
+      "step": 16195
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4535515341568506e-05,
+      "loss": 0.8232,
+      "step": 16196
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4525294576556114e-05,
+      "loss": 0.8461,
+      "step": 16197
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.451507564327451e-05,
+      "loss": 0.8921,
+      "step": 16198
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4504858541971665e-05,
+      "loss": 0.8575,
+      "step": 16199
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4494643272895623e-05,
+      "loss": 1.0023,
+      "step": 16200
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4484429836294255e-05,
+      "loss": 0.788,
+      "step": 16201
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4474218232415435e-05,
+      "loss": 1.3312,
+      "step": 16202
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4464008461506982e-05,
+      "loss": 0.9481,
+      "step": 16203
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4453800523816728e-05,
+      "loss": 0.8783,
+      "step": 16204
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4443594419592374e-05,
+      "loss": 0.7973,
+      "step": 16205
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4433390149081615e-05,
+      "loss": 0.7584,
+      "step": 16206
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4423187712532136e-05,
+      "loss": 0.7782,
+      "step": 16207
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.441298711019152e-05,
+      "loss": 0.9755,
+      "step": 16208
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4402788342307315e-05,
+      "loss": 0.7858,
+      "step": 16209
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4392591409127084e-05,
+      "loss": 0.7666,
+      "step": 16210
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.438239631089829e-05,
+      "loss": 0.7162,
+      "step": 16211
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4372203047868337e-05,
+      "loss": 0.7131,
+      "step": 16212
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.436201162028463e-05,
+      "loss": 0.7243,
+      "step": 16213
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.43518220283945e-05,
+      "loss": 0.8763,
+      "step": 16214
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4341634272445214e-05,
+      "loss": 0.8856,
+      "step": 16215
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4331448352684082e-05,
+      "loss": 0.8714,
+      "step": 16216
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4321264269358292e-05,
+      "loss": 0.81,
+      "step": 16217
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4311082022714993e-05,
+      "loss": 0.7391,
+      "step": 16218
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.430090161300127e-05,
+      "loss": 0.8213,
+      "step": 16219
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.429072304046427e-05,
+      "loss": 0.8162,
+      "step": 16220
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.428054630535099e-05,
+      "loss": 1.2737,
+      "step": 16221
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4270371407908375e-05,
+      "loss": 1.0252,
+      "step": 16222
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4260198348383435e-05,
+      "loss": 0.7521,
+      "step": 16223
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4250027127023013e-05,
+      "loss": 0.8407,
+      "step": 16224
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.423985774407396e-05,
+      "loss": 0.9362,
+      "step": 16225
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4229690199783118e-05,
+      "loss": 0.7944,
+      "step": 16226
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.421952449439725e-05,
+      "loss": 0.7628,
+      "step": 16227
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.420936062816299e-05,
+      "loss": 0.7632,
+      "step": 16228
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4199198601327112e-05,
+      "loss": 0.8624,
+      "step": 16229
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.418903841413618e-05,
+      "loss": 0.8881,
+      "step": 16230
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4178880066836773e-05,
+      "loss": 0.7823,
+      "step": 16231
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.416872355967549e-05,
+      "loss": 0.8705,
+      "step": 16232
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.415856889289877e-05,
+      "loss": 0.8767,
+      "step": 16233
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4148416066753075e-05,
+      "loss": 0.7891,
+      "step": 16234
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.413826508148479e-05,
+      "loss": 0.8378,
+      "step": 16235
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.412811593734031e-05,
+      "loss": 0.8333,
+      "step": 16236
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4117968634565945e-05,
+      "loss": 0.6643,
+      "step": 16237
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4107823173407918e-05,
+      "loss": 0.8194,
+      "step": 16238
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.409767955411252e-05,
+      "loss": 0.8913,
+      "step": 16239
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4087537776925894e-05,
+      "loss": 0.8086,
+      "step": 16240
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4077397842094185e-05,
+      "loss": 0.8557,
+      "step": 16241
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.406725974986348e-05,
+      "loss": 0.7553,
+      "step": 16242
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4057123500479815e-05,
+      "loss": 0.7866,
+      "step": 16243
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4046989094189188e-05,
+      "loss": 0.8799,
+      "step": 16244
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4036856531237594e-05,
+      "loss": 0.8035,
+      "step": 16245
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.402672581187091e-05,
+      "loss": 0.8571,
+      "step": 16246
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.401659693633499e-05,
+      "loss": 0.8634,
+      "step": 16247
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.4006469904875706e-05,
+      "loss": 0.8335,
+      "step": 16248
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3996344717738816e-05,
+      "loss": 0.7525,
+      "step": 16249
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3986221375170037e-05,
+      "loss": 0.9844,
+      "step": 16250
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.397609987741505e-05,
+      "loss": 0.6733,
+      "step": 16251
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.396598022471953e-05,
+      "loss": 0.7938,
+      "step": 16252
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3955862417329066e-05,
+      "loss": 0.8412,
+      "step": 16253
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.394574645548917e-05,
+      "loss": 0.6963,
+      "step": 16254
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3935632339445424e-05,
+      "loss": 1.0148,
+      "step": 16255
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3925520069443252e-05,
+      "loss": 1.0226,
+      "step": 16256
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3915409645728072e-05,
+      "loss": 0.7952,
+      "step": 16257
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3905301068545273e-05,
+      "loss": 0.872,
+      "step": 16258
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.389519433814016e-05,
+      "loss": 0.9778,
+      "step": 16259
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3885089454758013e-05,
+      "loss": 0.7699,
+      "step": 16260
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.387498641864412e-05,
+      "loss": 0.8932,
+      "step": 16261
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3864885230043643e-05,
+      "loss": 0.9041,
+      "step": 16262
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3854785889201713e-05,
+      "loss": 0.8294,
+      "step": 16263
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.384468839636349e-05,
+      "loss": 0.8502,
+      "step": 16264
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.383459275177401e-05,
+      "loss": 0.8276,
+      "step": 16265
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.382449895567824e-05,
+      "loss": 0.8474,
+      "step": 16266
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3814407008321237e-05,
+      "loss": 0.8266,
+      "step": 16267
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.3804316909947877e-05,
+      "loss": 0.7715,
+      "step": 16268
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3794228660803053e-05,
+      "loss": 0.8555,
+      "step": 16269
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.378414226113156e-05,
+      "loss": 0.9038,
+      "step": 16270
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.37740577111783e-05,
+      "loss": 0.7648,
+      "step": 16271
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.376397501118788e-05,
+      "loss": 0.856,
+      "step": 16272
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.37538941614051e-05,
+      "loss": 0.7504,
+      "step": 16273
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.374381516207458e-05,
+      "loss": 0.8352,
+      "step": 16274
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3733738013440942e-05,
+      "loss": 0.8424,
+      "step": 16275
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.372366271574872e-05,
+      "loss": 0.9059,
+      "step": 16276
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3713589269242486e-05,
+      "loss": 0.8181,
+      "step": 16277
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3703517674166697e-05,
+      "loss": 0.7801,
+      "step": 16278
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3693447930765757e-05,
+      "loss": 0.7303,
+      "step": 16279
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.368338003928411e-05,
+      "loss": 0.803,
+      "step": 16280
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3673313999966072e-05,
+      "loss": 0.69,
+      "step": 16281
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.366324981305591e-05,
+      "loss": 0.8372,
+      "step": 16282
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3653187478797923e-05,
+      "loss": 0.8484,
+      "step": 16283
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3643126997436304e-05,
+      "loss": 0.7745,
+      "step": 16284
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3633068369215206e-05,
+      "loss": 0.7462,
+      "step": 16285
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.362301159437875e-05,
+      "loss": 0.8711,
+      "step": 16286
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3612956673171017e-05,
+      "loss": 0.7468,
+      "step": 16287
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.360290360583599e-05,
+      "loss": 0.9141,
+      "step": 16288
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3592852392617726e-05,
+      "loss": 0.907,
+      "step": 16289
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3582803033760113e-05,
+      "loss": 1.2563,
+      "step": 16290
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3572755529507074e-05,
+      "loss": 1.0662,
+      "step": 16291
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3562709880102396e-05,
+      "loss": 0.8752,
+      "step": 16292
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3552666085789953e-05,
+      "loss": 0.7137,
+      "step": 16293
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3542624146813476e-05,
+      "loss": 0.856,
+      "step": 16294
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.353258406341664e-05,
+      "loss": 0.7912,
+      "step": 16295
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3522545835843178e-05,
+      "loss": 0.6436,
+      "step": 16296
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3512509464336686e-05,
+      "loss": 1.1846,
+      "step": 16297
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.350247494914071e-05,
+      "loss": 1.0565,
+      "step": 16298
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.349244229049884e-05,
+      "loss": 0.8695,
+      "step": 16299
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3482411488654555e-05,
+      "loss": 0.9006,
+      "step": 16300
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3472382543851212e-05,
+      "loss": 1.2221,
+      "step": 16301
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3462355456332307e-05,
+      "loss": 0.9975,
+      "step": 16302
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3452330226341157e-05,
+      "loss": 0.9508,
+      "step": 16303
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.344230685412103e-05,
+      "loss": 0.7338,
+      "step": 16304
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3432285339915272e-05,
+      "loss": 0.8558,
+      "step": 16305
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.342226568396704e-05,
+      "loss": 0.9474,
+      "step": 16306
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3412247886519524e-05,
+      "loss": 0.8507,
+      "step": 16307
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.340223194781581e-05,
+      "loss": 0.8874,
+      "step": 16308
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3392217868099052e-05,
+      "loss": 0.8564,
+      "step": 16309
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.338220564761224e-05,
+      "loss": 0.9479,
+      "step": 16310
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3372195286598364e-05,
+      "loss": 0.806,
+      "step": 16311
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.33621867853004e-05,
+      "loss": 0.809,
+      "step": 16312
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3352180143961234e-05,
+      "loss": 0.8729,
+      "step": 16313
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3342175362823683e-05,
+      "loss": 0.8541,
+      "step": 16314
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.333217244213065e-05,
+      "loss": 0.9319,
+      "step": 16315
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3322171382124812e-05,
+      "loss": 0.8682,
+      "step": 16316
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3312172183048896e-05,
+      "loss": 0.724,
+      "step": 16317
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.330217484514563e-05,
+      "loss": 0.7201,
+      "step": 16318
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.329217936865762e-05,
+      "loss": 0.6751,
+      "step": 16319
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.328218575382741e-05,
+      "loss": 0.8377,
+      "step": 16320
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.327219400089762e-05,
+      "loss": 0.9634,
+      "step": 16321
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3262204110110685e-05,
+      "loss": 0.6785,
+      "step": 16322
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.325221608170908e-05,
+      "loss": 0.6576,
+      "step": 16323
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3242229915935166e-05,
+      "loss": 0.875,
+      "step": 16324
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.323224561303137e-05,
+      "loss": 0.856,
+      "step": 16325
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3222263173239968e-05,
+      "loss": 0.8403,
+      "step": 16326
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.321228259680319e-05,
+      "loss": 0.849,
+      "step": 16327
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.320230388396335e-05,
+      "loss": 0.8907,
+      "step": 16328
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3192327034962558e-05,
+      "loss": 0.8823,
+      "step": 16329
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3182352050042976e-05,
+      "loss": 0.9988,
+      "step": 16330
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.317237892944668e-05,
+      "loss": 0.647,
+      "step": 16331
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.316240767341571e-05,
+      "loss": 1.1854,
+      "step": 16332
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3152438282192045e-05,
+      "loss": 1.1488,
+      "step": 16333
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3142470756017687e-05,
+      "loss": 0.8784,
+      "step": 16334
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3132505095134505e-05,
+      "loss": 0.869,
+      "step": 16335
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.312254129978434e-05,
+      "loss": 0.845,
+      "step": 16336
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3112579370209074e-05,
+      "loss": 0.8255,
+      "step": 16337
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.310261930665044e-05,
+      "loss": 0.9493,
+      "step": 16338
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3092661109350134e-05,
+      "loss": 0.9845,
+      "step": 16339
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.308270477854989e-05,
+      "loss": 0.9577,
+      "step": 16340
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.307275031449132e-05,
+      "loss": 1.3753,
+      "step": 16341
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.306279771741602e-05,
+      "loss": 1.0379,
+      "step": 16342
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.305284698756549e-05,
+      "loss": 0.9186,
+      "step": 16343
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3042898125181333e-05,
+      "loss": 0.7294,
+      "step": 16344
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3032951130504864e-05,
+      "loss": 0.8152,
+      "step": 16345
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3023006003777604e-05,
+      "loss": 0.8091,
+      "step": 16346
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3013062745240865e-05,
+      "loss": 0.8986,
+      "step": 16347
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.3003121355135982e-05,
+      "loss": 0.9366,
+      "step": 16348
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2993181833704193e-05,
+      "loss": 0.9277,
+      "step": 16349
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.298324418118677e-05,
+      "loss": 0.8417,
+      "step": 16350
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2973308397824878e-05,
+      "loss": 0.8635,
+      "step": 16351
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2963374483859623e-05,
+      "loss": 0.8311,
+      "step": 16352
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2953442439532157e-05,
+      "loss": 0.8186,
+      "step": 16353
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.294351226508349e-05,
+      "loss": 0.9525,
+      "step": 16354
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.293358396075459e-05,
+      "loss": 0.7598,
+      "step": 16355
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.292365752678648e-05,
+      "loss": 0.9971,
+      "step": 16356
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2913732963420033e-05,
+      "loss": 0.8354,
+      "step": 16357
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.290381027089612e-05,
+      "loss": 0.8911,
+      "step": 16358
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.289388944945555e-05,
+      "loss": 0.9266,
+      "step": 16359
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2883970499339103e-05,
+      "loss": 0.899,
+      "step": 16360
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2874053420787465e-05,
+      "loss": 0.8365,
+      "step": 16361
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2864138214041398e-05,
+      "loss": 1.331,
+      "step": 16362
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.285422487934149e-05,
+      "loss": 1.192,
+      "step": 16363
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2844313416928343e-05,
+      "loss": 0.7668,
+      "step": 16364
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2834403827042473e-05,
+      "loss": 0.8013,
+      "step": 16365
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2824496109924433e-05,
+      "loss": 0.7433,
+      "step": 16366
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.281459026581465e-05,
+      "loss": 0.9501,
+      "step": 16367
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2804686294953515e-05,
+      "loss": 0.7567,
+      "step": 16368
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2794784197581443e-05,
+      "loss": 0.82,
+      "step": 16369
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.278488397393872e-05,
+      "loss": 0.8213,
+      "step": 16370
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2774985624265586e-05,
+      "loss": 0.8589,
+      "step": 16371
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.2765089148802343e-05,
+      "loss": 0.9288,
+      "step": 16372
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.275519454778914e-05,
+      "loss": 0.843,
+      "step": 16373
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.274530182146609e-05,
+      "loss": 0.8094,
+      "step": 16374
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2735410970073312e-05,
+      "loss": 0.7622,
+      "step": 16375
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.272552199385085e-05,
+      "loss": 0.8067,
+      "step": 16376
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.271563489303866e-05,
+      "loss": 1.3529,
+      "step": 16377
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2705749667876764e-05,
+      "loss": 1.1613,
+      "step": 16378
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.269586631860504e-05,
+      "loss": 0.8134,
+      "step": 16379
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.268598484546335e-05,
+      "loss": 0.7921,
+      "step": 16380
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.267610524869148e-05,
+      "loss": 0.8264,
+      "step": 16381
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.266622752852927e-05,
+      "loss": 0.8574,
+      "step": 16382
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2656351685216425e-05,
+      "loss": 0.8095,
+      "step": 16383
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2646477718992575e-05,
+      "loss": 0.7142,
+      "step": 16384
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2636605630097428e-05,
+      "loss": 0.9522,
+      "step": 16385
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.262673541877054e-05,
+      "loss": 0.8269,
+      "step": 16386
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2616867085251427e-05,
+      "loss": 1.3308,
+      "step": 16387
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2607000629779685e-05,
+      "loss": 1.1448,
+      "step": 16388
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2597136052594657e-05,
+      "loss": 0.8905,
+      "step": 16389
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.258727335393578e-05,
+      "loss": 0.7498,
+      "step": 16390
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2577412534042463e-05,
+      "loss": 0.7598,
+      "step": 16391
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.256755359315399e-05,
+      "loss": 0.9759,
+      "step": 16392
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2557696531509598e-05,
+      "loss": 0.7583,
+      "step": 16393
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2547841349348585e-05,
+      "loss": 0.7907,
+      "step": 16394
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2537988046910084e-05,
+      "loss": 0.7738,
+      "step": 16395
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2528136624433228e-05,
+      "loss": 0.9255,
+      "step": 16396
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2518287082157143e-05,
+      "loss": 0.8228,
+      "step": 16397
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2508439420320838e-05,
+      "loss": 1.37,
+      "step": 16398
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2498593639163323e-05,
+      "loss": 1.0902,
+      "step": 16399
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.248874973892352e-05,
+      "loss": 0.7522,
+      "step": 16400
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.247890771984039e-05,
+      "loss": 0.8529,
+      "step": 16401
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.246906758215277e-05,
+      "loss": 0.8008,
+      "step": 16402
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.245922932609946e-05,
+      "loss": 0.7518,
+      "step": 16403
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2449392951919245e-05,
+      "loss": 0.7396,
+      "step": 16404
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.243955845985084e-05,
+      "loss": 0.8025,
+      "step": 16405
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.242972585013291e-05,
+      "loss": 0.7973,
+      "step": 16406
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.241989512300412e-05,
+      "loss": 0.9603,
+      "step": 16407
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2410066278703036e-05,
+      "loss": 0.7823,
+      "step": 16408
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2400239317468175e-05,
+      "loss": 0.8186,
+      "step": 16409
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2390414239538092e-05,
+      "loss": 0.9184,
+      "step": 16410
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2380591045151212e-05,
+      "loss": 0.7674,
+      "step": 16411
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2370769734545892e-05,
+      "loss": 0.8188,
+      "step": 16412
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.236095030796056e-05,
+      "loss": 0.8483,
+      "step": 16413
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2351132765633485e-05,
+      "loss": 0.6585,
+      "step": 16414
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2341317107802963e-05,
+      "loss": 0.8255,
+      "step": 16415
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.233150333470716e-05,
+      "loss": 0.8291,
+      "step": 16416
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2321691446584314e-05,
+      "loss": 0.8225,
+      "step": 16417
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2311881443672534e-05,
+      "loss": 0.9247,
+      "step": 16418
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2302073326209895e-05,
+      "loss": 0.7295,
+      "step": 16419
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.229226709443444e-05,
+      "loss": 0.753,
+      "step": 16420
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2282462748584153e-05,
+      "loss": 0.7716,
+      "step": 16421
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2272660288896963e-05,
+      "loss": 0.8996,
+      "step": 16422
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2262859715610816e-05,
+      "loss": 0.6897,
+      "step": 16423
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2253061028963552e-05,
+      "loss": 0.8678,
+      "step": 16424
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.224326422919293e-05,
+      "loss": 0.8086,
+      "step": 16425
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2233469316536793e-05,
+      "loss": 0.8506,
+      "step": 16426
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.222367629123282e-05,
+      "loss": 0.8904,
+      "step": 16427
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2213885153518653e-05,
+      "loss": 0.9868,
+      "step": 16428
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2204095903631983e-05,
+      "loss": 0.8168,
+      "step": 16429
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2194308541810338e-05,
+      "loss": 0.8008,
+      "step": 16430
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2184523068291286e-05,
+      "loss": 1.3554,
+      "step": 16431
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.217473948331229e-05,
+      "loss": 1.2079,
+      "step": 16432
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2164957787110795e-05,
+      "loss": 0.8425,
+      "step": 16433
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.215517797992418e-05,
+      "loss": 0.8074,
+      "step": 16434
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2145400061989842e-05,
+      "loss": 0.7858,
+      "step": 16435
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2135624033545067e-05,
+      "loss": 0.8248,
+      "step": 16436
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.212584989482711e-05,
+      "loss": 0.7928,
+      "step": 16437
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.211607764607315e-05,
+      "loss": 0.8196,
+      "step": 16438
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2106307287520412e-05,
+      "loss": 0.685,
+      "step": 16439
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2096538819406e-05,
+      "loss": 0.6589,
+      "step": 16440
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.208677224196696e-05,
+      "loss": 0.8827,
+      "step": 16441
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2077007555440356e-05,
+      "loss": 0.7769,
+      "step": 16442
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2067244760063167e-05,
+      "loss": 0.8187,
+      "step": 16443
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2057483856072304e-05,
+      "loss": 0.779,
+      "step": 16444
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.20477248437047e-05,
+      "loss": 0.81,
+      "step": 16445
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2037967723197172e-05,
+      "loss": 0.8215,
+      "step": 16446
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2028212494786537e-05,
+      "loss": 0.8003,
+      "step": 16447
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.2018459158709536e-05,
+      "loss": 0.7563,
+      "step": 16448
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.200870771520288e-05,
+      "loss": 0.8067,
+      "step": 16449
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1998958164503203e-05,
+      "loss": 0.8722,
+      "step": 16450
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.198921050684718e-05,
+      "loss": 0.8605,
+      "step": 16451
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1979464742471355e-05,
+      "loss": 0.8736,
+      "step": 16452
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1969720871612243e-05,
+      "loss": 0.8246,
+      "step": 16453
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1959978894506304e-05,
+      "loss": 0.7271,
+      "step": 16454
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1950238811390034e-05,
+      "loss": 0.7593,
+      "step": 16455
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.194050062249977e-05,
+      "loss": 0.9514,
+      "step": 16456
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.193076432807184e-05,
+      "loss": 0.751,
+      "step": 16457
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1921029928342586e-05,
+      "loss": 0.8439,
+      "step": 16458
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1911297423548237e-05,
+      "loss": 0.8396,
+      "step": 16459
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1901566813924968e-05,
+      "loss": 0.7111,
+      "step": 16460
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1891838099708985e-05,
+      "loss": 0.9305,
+      "step": 16461
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1882111281136397e-05,
+      "loss": 0.8061,
+      "step": 16462
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1872386358443187e-05,
+      "loss": 0.7894,
+      "step": 16463
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1862663331865464e-05,
+      "loss": 0.8234,
+      "step": 16464
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.185294220163917e-05,
+      "loss": 0.9027,
+      "step": 16465
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1843222968000197e-05,
+      "loss": 0.9309,
+      "step": 16466
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.183350563118449e-05,
+      "loss": 0.7906,
+      "step": 16467
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1823790191427862e-05,
+      "loss": 0.8375,
+      "step": 16468
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1814076648966052e-05,
+      "loss": 0.7779,
+      "step": 16469
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1804365004034877e-05,
+      "loss": 0.8916,
+      "step": 16470
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.179465525686999e-05,
+      "loss": 0.9068,
+      "step": 16471
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.178494740770706e-05,
+      "loss": 0.8801,
+      "step": 16472
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1775241456781646e-05,
+      "loss": 0.9151,
+      "step": 16473
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1765537404329373e-05,
+      "loss": 0.799,
+      "step": 16474
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.175583525058572e-05,
+      "loss": 0.7507,
+      "step": 16475
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1746134995786127e-05,
+      "loss": 0.8603,
+      "step": 16476
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.173643664016609e-05,
+      "loss": 0.9619,
+      "step": 16477
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1726740183960915e-05,
+      "loss": 0.8306,
+      "step": 16478
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.171704562740593e-05,
+      "loss": 0.9706,
+      "step": 16479
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.170735297073645e-05,
+      "loss": 0.951,
+      "step": 16480
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1697662214187707e-05,
+      "loss": 0.9843,
+      "step": 16481
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1687973357994852e-05,
+      "loss": 0.7804,
+      "step": 16482
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.167828640239308e-05,
+      "loss": 0.8769,
+      "step": 16483
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1668601347617467e-05,
+      "loss": 0.8066,
+      "step": 16484
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1658918193903033e-05,
+      "loss": 0.7546,
+      "step": 16485
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.164923694148484e-05,
+      "loss": 0.9052,
+      "step": 16486
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1639557590597813e-05,
+      "loss": 0.9311,
+      "step": 16487
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1629880141476877e-05,
+      "loss": 0.7565,
+      "step": 16488
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1620204594356874e-05,
+      "loss": 0.8892,
+      "step": 16489
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1610530949472652e-05,
+      "loss": 0.9282,
+      "step": 16490
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.160085920705899e-05,
+      "loss": 0.7787,
+      "step": 16491
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1591189367350594e-05,
+      "loss": 0.797,
+      "step": 16492
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1581521430582152e-05,
+      "loss": 0.855,
+      "step": 16493
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1571855396988306e-05,
+      "loss": 1.2373,
+      "step": 16494
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1562191266803612e-05,
+      "loss": 1.0039,
+      "step": 16495
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.155252904026267e-05,
+      "loss": 0.9253,
+      "step": 16496
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1542868717599953e-05,
+      "loss": 0.7949,
+      "step": 16497
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1533210299049878e-05,
+      "loss": 0.8841,
+      "step": 16498
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1523553784846905e-05,
+      "loss": 0.8251,
+      "step": 16499
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.151389917522536e-05,
+      "loss": 0.7664,
+      "step": 16500
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1504246470419543e-05,
+      "loss": 0.8779,
+      "step": 16501
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1494595670663763e-05,
+      "loss": 0.7939,
+      "step": 16502
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.148494677619223e-05,
+      "loss": 0.8546,
+      "step": 16503
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.147529978723909e-05,
+      "loss": 0.8027,
+      "step": 16504
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1465654704038452e-05,
+      "loss": 0.8626,
+      "step": 16505
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1456011526824494e-05,
+      "loss": 0.743,
+      "step": 16506
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1446370255831126e-05,
+      "loss": 0.7767,
+      "step": 16507
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1436730891292424e-05,
+      "loss": 0.7804,
+      "step": 16508
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.14270934334423e-05,
+      "loss": 0.7737,
+      "step": 16509
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1417457882514658e-05,
+      "loss": 0.7484,
+      "step": 16510
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.140782423874331e-05,
+      "loss": 0.8918,
+      "step": 16511
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1398192502362115e-05,
+      "loss": 0.8141,
+      "step": 16512
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1388562673604818e-05,
+      "loss": 0.9026,
+      "step": 16513
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.137893475270508e-05,
+      "loss": 0.843,
+      "step": 16514
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1369308739896633e-05,
+      "loss": 1.2624,
+      "step": 16515
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.135968463541307e-05,
+      "loss": 1.1313,
+      "step": 16516
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1350062439487927e-05,
+      "loss": 0.7673,
+      "step": 16517
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1340442152354788e-05,
+      "loss": 0.8459,
+      "step": 16518
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.133082377424711e-05,
+      "loss": 0.762,
+      "step": 16519
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1321207305398326e-05,
+      "loss": 0.8146,
+      "step": 16520
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1311592746041808e-05,
+      "loss": 0.8092,
+      "step": 16521
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1301980096410912e-05,
+      "loss": 0.8481,
+      "step": 16522
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1292369356738895e-05,
+      "loss": 0.7856,
+      "step": 16523
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1282760527259073e-05,
+      "loss": 1.2617,
+      "step": 16524
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1273153608204598e-05,
+      "loss": 1.0353,
+      "step": 16525
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1263548599808613e-05,
+      "loss": 0.9321,
+      "step": 16526
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1253945502304273e-05,
+      "loss": 0.7745,
+      "step": 16527
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1244344315924624e-05,
+      "loss": 0.7889,
+      "step": 16528
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.123474504090267e-05,
+      "loss": 0.8059,
+      "step": 16529
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1225147677471357e-05,
+      "loss": 0.9581,
+      "step": 16530
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1215552225863666e-05,
+      "loss": 0.9764,
+      "step": 16531
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1205958686312445e-05,
+      "loss": 0.7655,
+      "step": 16532
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1196367059050494e-05,
+      "loss": 0.7137,
+      "step": 16533
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.118677734431065e-05,
+      "loss": 0.8278,
+      "step": 16534
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.117718954232566e-05,
+      "loss": 0.733,
+      "step": 16535
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1167603653328115e-05,
+      "loss": 0.7229,
+      "step": 16536
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1158019677550757e-05,
+      "loss": 0.8901,
+      "step": 16537
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1148437615226146e-05,
+      "loss": 0.8132,
+      "step": 16538
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1138857466586826e-05,
+      "loss": 0.6607,
+      "step": 16539
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.112927923186533e-05,
+      "loss": 0.8239,
+      "step": 16540
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1119702911294104e-05,
+      "loss": 0.7338,
+      "step": 16541
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.111012850510553e-05,
+      "loss": 0.8734,
+      "step": 16542
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1100556013532035e-05,
+      "loss": 0.8452,
+      "step": 16543
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.10909854368059e-05,
+      "loss": 0.8786,
+      "step": 16544
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.10814167751594e-05,
+      "loss": 0.9237,
+      "step": 16545
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1071850028824745e-05,
+      "loss": 0.8948,
+      "step": 16546
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.106228519803416e-05,
+      "loss": 0.7044,
+      "step": 16547
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1052722283019745e-05,
+      "loss": 0.8105,
+      "step": 16548
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1043161284013568e-05,
+      "loss": 0.8147,
+      "step": 16549
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.103360220124776e-05,
+      "loss": 0.8187,
+      "step": 16550
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.102404503495421e-05,
+      "loss": 0.7174,
+      "step": 16551
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.101448978536489e-05,
+      "loss": 1.3206,
+      "step": 16552
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.1004936452711732e-05,
+      "loss": 1.1386,
+      "step": 16553
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0995385037226588e-05,
+      "loss": 1.2787,
+      "step": 16554
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.098583553914122e-05,
+      "loss": 1.0984,
+      "step": 16555
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0976287958687447e-05,
+      "loss": 0.8787,
+      "step": 16556
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0966742296096954e-05,
+      "loss": 0.8115,
+      "step": 16557
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.095719855160139e-05,
+      "loss": 0.753,
+      "step": 16558
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.094765672543243e-05,
+      "loss": 0.9254,
+      "step": 16559
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0938116817821606e-05,
+      "loss": 0.8354,
+      "step": 16560
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0928578829000478e-05,
+      "loss": 0.8478,
+      "step": 16561
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0919042759200468e-05,
+      "loss": 0.7683,
+      "step": 16562
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0909508608653094e-05,
+      "loss": 0.8085,
+      "step": 16563
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.089997637758969e-05,
+      "loss": 0.8074,
+      "step": 16564
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.089044606624162e-05,
+      "loss": 0.802,
+      "step": 16565
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0880917674840173e-05,
+      "loss": 0.867,
+      "step": 16566
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0871391203616587e-05,
+      "loss": 0.7133,
+      "step": 16567
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0861866652802053e-05,
+      "loss": 0.6171,
+      "step": 16568
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0852344022627768e-05,
+      "loss": 1.0496,
+      "step": 16569
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0842823313324832e-05,
+      "loss": 0.8976,
+      "step": 16570
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0833304525124252e-05,
+      "loss": 1.332,
+      "step": 16571
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0823787658257133e-05,
+      "loss": 1.0737,
+      "step": 16572
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0814272712954386e-05,
+      "loss": 0.7838,
+      "step": 16573
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.080475968944693e-05,
+      "loss": 0.8539,
+      "step": 16574
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0795248587965676e-05,
+      "loss": 0.7566,
+      "step": 16575
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0785739408741434e-05,
+      "loss": 0.6939,
+      "step": 16576
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0776232152004992e-05,
+      "loss": 0.8682,
+      "step": 16577
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.076672681798706e-05,
+      "loss": 0.9341,
+      "step": 16578
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0757223406918404e-05,
+      "loss": 0.8196,
+      "step": 16579
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.074772191902955e-05,
+      "loss": 0.7924,
+      "step": 16580
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0738222354551184e-05,
+      "loss": 0.7005,
+      "step": 16581
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0728724713713833e-05,
+      "loss": 0.8877,
+      "step": 16582
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0719228996747985e-05,
+      "loss": 0.8931,
+      "step": 16583
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0709735203884085e-05,
+      "loss": 0.8371,
+      "step": 16584
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0700243335352588e-05,
+      "loss": 0.8285,
+      "step": 16585
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.069075339138383e-05,
+      "loss": 0.8246,
+      "step": 16586
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0681265372208102e-05,
+      "loss": 1.398,
+      "step": 16587
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0671779278055726e-05,
+      "loss": 1.0957,
+      "step": 16588
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0662295109156892e-05,
+      "loss": 0.9048,
+      "step": 16589
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0652812865741756e-05,
+      "loss": 0.856,
+      "step": 16590
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0643332548040505e-05,
+      "loss": 0.7838,
+      "step": 16591
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0633854156283183e-05,
+      "loss": 0.9013,
+      "step": 16592
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0624377690699836e-05,
+      "loss": 0.7764,
+      "step": 16593
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0614903151520448e-05,
+      "loss": 0.9235,
+      "step": 16594
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.060543053897496e-05,
+      "loss": 0.6763,
+      "step": 16595
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0595959853293234e-05,
+      "loss": 0.7994,
+      "step": 16596
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0586491094705174e-05,
+      "loss": 0.8672,
+      "step": 16597
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.057702426344057e-05,
+      "loss": 0.7792,
+      "step": 16598
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0567559359729137e-05,
+      "loss": 0.7927,
+      "step": 16599
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0558096383800652e-05,
+      "loss": 1.249,
+      "step": 16600
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0548635335884726e-05,
+      "loss": 0.9807,
+      "step": 16601
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0539176216210987e-05,
+      "loss": 0.9176,
+      "step": 16602
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0529719025008976e-05,
+      "loss": 0.8784,
+      "step": 16603
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0520263762508264e-05,
+      "loss": 0.8809,
+      "step": 16604
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.051081042893829e-05,
+      "loss": 0.7132,
+      "step": 16605
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0501359024528476e-05,
+      "loss": 0.7987,
+      "step": 16606
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0491909549508248e-05,
+      "loss": 0.7672,
+      "step": 16607
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0482462004106895e-05,
+      "loss": 0.9897,
+      "step": 16608
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0473016388553734e-05,
+      "loss": 0.7073,
+      "step": 16609
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0463572703077982e-05,
+      "loss": 0.8459,
+      "step": 16610
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0454130947908835e-05,
+      "loss": 0.8395,
+      "step": 16611
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0444691123275427e-05,
+      "loss": 0.9261,
+      "step": 16612
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0435253229406892e-05,
+      "loss": 0.8037,
+      "step": 16613
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0425817266532276e-05,
+      "loss": 0.7809,
+      "step": 16614
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0416383234880542e-05,
+      "loss": 0.7559,
+      "step": 16615
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0406951134680707e-05,
+      "loss": 0.8911,
+      "step": 16616
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0397520966161654e-05,
+      "loss": 0.6997,
+      "step": 16617
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0388092729552256e-05,
+      "loss": 0.8216,
+      "step": 16618
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0378666425081305e-05,
+      "loss": 0.7478,
+      "step": 16619
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.036924205297761e-05,
+      "loss": 0.9568,
+      "step": 16620
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0359819613469876e-05,
+      "loss": 1.1972,
+      "step": 16621
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0350399106786766e-05,
+      "loss": 1.0851,
+      "step": 16622
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0340980533156963e-05,
+      "loss": 0.8531,
+      "step": 16623
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0331563892809026e-05,
+      "loss": 1.2707,
+      "step": 16624
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.032214918597144e-05,
+      "loss": 1.5283,
+      "step": 16625
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0312736412872757e-05,
+      "loss": 1.1378,
+      "step": 16626
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0303325573741407e-05,
+      "loss": 1.3088,
+      "step": 16627
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0293916668805746e-05,
+      "loss": 1.1058,
+      "step": 16628
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0284509698294185e-05,
+      "loss": 0.9032,
+      "step": 16629
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0275104662434997e-05,
+      "loss": 0.8587,
+      "step": 16630
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0265701561456407e-05,
+      "loss": 0.9394,
+      "step": 16631
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.025630039558668e-05,
+      "loss": 0.8708,
+      "step": 16632
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0246901165053955e-05,
+      "loss": 0.7951,
+      "step": 16633
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0237503870086337e-05,
+      "loss": 0.7047,
+      "step": 16634
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0228108510911882e-05,
+      "loss": 0.8876,
+      "step": 16635
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0218715087758655e-05,
+      "loss": 0.8091,
+      "step": 16636
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.020932360085459e-05,
+      "loss": 0.798,
+      "step": 16637
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0199934050427628e-05,
+      "loss": 1.3799,
+      "step": 16638
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.019054643670565e-05,
+      "loss": 0.9348,
+      "step": 16639
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0181160759916472e-05,
+      "loss": 0.7862,
+      "step": 16640
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0171777020287875e-05,
+      "loss": 0.7114,
+      "step": 16641
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0162395218047638e-05,
+      "loss": 0.8356,
+      "step": 16642
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0153015353423423e-05,
+      "loss": 0.8376,
+      "step": 16643
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0143637426642858e-05,
+      "loss": 0.752,
+      "step": 16644
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.013426143793359e-05,
+      "loss": 0.901,
+      "step": 16645
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.012488738752313e-05,
+      "loss": 0.8277,
+      "step": 16646
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0115515275638963e-05,
+      "loss": 0.7726,
+      "step": 16647
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0106145102508612e-05,
+      "loss": 0.8586,
+      "step": 16648
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0096776868359446e-05,
+      "loss": 1.0042,
+      "step": 16649
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.008741057341882e-05,
+      "loss": 0.8756,
+      "step": 16650
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0078046217914038e-05,
+      "loss": 1.35,
+      "step": 16651
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0068683802072418e-05,
+      "loss": 1.1389,
+      "step": 16652
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0059323326121138e-05,
+      "loss": 0.7458,
+      "step": 16653
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.004996479028739e-05,
+      "loss": 0.83,
+      "step": 16654
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0040608194798304e-05,
+      "loss": 0.7808,
+      "step": 16655
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0031253539880936e-05,
+      "loss": 0.7169,
+      "step": 16656
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.002190082576232e-05,
+      "loss": 0.8599,
+      "step": 16657
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0012550052669466e-05,
+      "loss": 0.8794,
+      "step": 16658
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0003201220829314e-05,
+      "loss": 0.7336,
+      "step": 16659
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9993854330468708e-05,
+      "loss": 0.9287,
+      "step": 16660
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9984509381814554e-05,
+      "loss": 0.7333,
+      "step": 16661
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9975166375093625e-05,
+      "loss": 1.2487,
+      "step": 16662
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9965825310532625e-05,
+      "loss": 1.018,
+      "step": 16663
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9956486188358338e-05,
+      "loss": 0.8236,
+      "step": 16664
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9947149008797384e-05,
+      "loss": 0.6313,
+      "step": 16665
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9937813772076352e-05,
+      "loss": 0.9046,
+      "step": 16666
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9928480478421796e-05,
+      "loss": 0.7787,
+      "step": 16667
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.991914912806031e-05,
+      "loss": 0.8398,
+      "step": 16668
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9909819721218247e-05,
+      "loss": 0.8423,
+      "step": 16669
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9900492258122115e-05,
+      "loss": 0.8805,
+      "step": 16670
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9891166738998246e-05,
+      "loss": 0.7725,
+      "step": 16671
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9881843164072956e-05,
+      "loss": 0.7256,
+      "step": 16672
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9872521533572563e-05,
+      "loss": 0.8116,
+      "step": 16673
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.986320184772328e-05,
+      "loss": 0.8773,
+      "step": 16674
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9853884106751274e-05,
+      "loss": 0.8939,
+      "step": 16675
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9844568310882683e-05,
+      "loss": 0.8391,
+      "step": 16676
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9835254460343624e-05,
+      "loss": 0.7964,
+      "step": 16677
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9825942555360134e-05,
+      "loss": 0.9265,
+      "step": 16678
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9816632596158157e-05,
+      "loss": 1.2614,
+      "step": 16679
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9807324582963716e-05,
+      "loss": 1.5902,
+      "step": 16680
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.979801851600267e-05,
+      "loss": 1.1668,
+      "step": 16681
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9788714395500883e-05,
+      "loss": 0.9777,
+      "step": 16682
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9779412221684157e-05,
+      "loss": 0.7177,
+      "step": 16683
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.977011199477825e-05,
+      "loss": 0.8092,
+      "step": 16684
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.976081371500884e-05,
+      "loss": 0.7729,
+      "step": 16685
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9751517382601657e-05,
+      "loss": 0.9258,
+      "step": 16686
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.974222299778229e-05,
+      "loss": 0.8705,
+      "step": 16687
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9732930560776276e-05,
+      "loss": 0.7954,
+      "step": 16688
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.972364007180919e-05,
+      "loss": 0.7503,
+      "step": 16689
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9714351531106478e-05,
+      "loss": 0.7383,
+      "step": 16690
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.970506493889358e-05,
+      "loss": 0.841,
+      "step": 16691
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.969578029539585e-05,
+      "loss": 0.7822,
+      "step": 16692
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.968649760083866e-05,
+      "loss": 0.8741,
+      "step": 16693
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9677216855447277e-05,
+      "loss": 0.7751,
+      "step": 16694
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.966793805944691e-05,
+      "loss": 0.7619,
+      "step": 16695
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9658661213062812e-05,
+      "loss": 0.8166,
+      "step": 16696
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9649386316520126e-05,
+      "loss": 0.8093,
+      "step": 16697
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9640113370043867e-05,
+      "loss": 0.8529,
+      "step": 16698
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9630842373859147e-05,
+      "loss": 0.7984,
+      "step": 16699
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9621573328190977e-05,
+      "loss": 0.7725,
+      "step": 16700
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9612306233264256e-05,
+      "loss": 0.9048,
+      "step": 16701
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.960304108930395e-05,
+      "loss": 0.8409,
+      "step": 16702
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9593777896534904e-05,
+      "loss": 0.8411,
+      "step": 16703
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9584516655181895e-05,
+      "loss": 0.8185,
+      "step": 16704
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9575257365469734e-05,
+      "loss": 0.9525,
+      "step": 16705
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9566000027623132e-05,
+      "loss": 0.8499,
+      "step": 16706
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9556744641866743e-05,
+      "loss": 0.9061,
+      "step": 16707
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9547491208425173e-05,
+      "loss": 0.8379,
+      "step": 16708
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9538239727523045e-05,
+      "loss": 0.8899,
+      "step": 16709
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9528990199384866e-05,
+      "loss": 0.868,
+      "step": 16710
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9519742624235093e-05,
+      "loss": 0.6817,
+      "step": 16711
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.951049700229822e-05,
+      "loss": 0.755,
+      "step": 16712
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9501253333798585e-05,
+      "loss": 0.8584,
+      "step": 16713
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9492011618960503e-05,
+      "loss": 0.6803,
+      "step": 16714
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.948277185800833e-05,
+      "loss": 0.8636,
+      "step": 16715
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9473534051166274e-05,
+      "loss": 0.9,
+      "step": 16716
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9464298198658514e-05,
+      "loss": 0.8706,
+      "step": 16717
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9455064300709247e-05,
+      "loss": 1.0244,
+      "step": 16718
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9445832357542548e-05,
+      "loss": 1.3652,
+      "step": 16719
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.943660236938245e-05,
+      "loss": 1.1459,
+      "step": 16720
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9427374336453008e-05,
+      "loss": 0.8281,
+      "step": 16721
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9418148258978163e-05,
+      "loss": 0.8134,
+      "step": 16722
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9408924137181816e-05,
+      "loss": 0.9368,
+      "step": 16723
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9399701971287808e-05,
+      "loss": 0.9384,
+      "step": 16724
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9390481761520006e-05,
+      "loss": 0.9666,
+      "step": 16725
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9381263508102165e-05,
+      "loss": 0.8713,
+      "step": 16726
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9372047211257995e-05,
+      "loss": 0.7593,
+      "step": 16727
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.936283287121118e-05,
+      "loss": 0.7158,
+      "step": 16728
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.935362048818532e-05,
+      "loss": 0.9036,
+      "step": 16729
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.934441006240404e-05,
+      "loss": 0.7701,
+      "step": 16730
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9335201594090847e-05,
+      "loss": 0.8958,
+      "step": 16731
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9325995083469216e-05,
+      "loss": 0.7164,
+      "step": 16732
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9316790530762585e-05,
+      "loss": 0.8672,
+      "step": 16733
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9307587936194382e-05,
+      "loss": 0.7689,
+      "step": 16734
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.929838729998792e-05,
+      "loss": 0.8015,
+      "step": 16735
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9289188622366462e-05,
+      "loss": 0.734,
+      "step": 16736
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.927999190355332e-05,
+      "loss": 0.8204,
+      "step": 16737
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9270797143771656e-05,
+      "loss": 0.7745,
+      "step": 16738
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.926160434324461e-05,
+      "loss": 0.8808,
+      "step": 16739
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9252413502195322e-05,
+      "loss": 0.805,
+      "step": 16740
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9243224620846868e-05,
+      "loss": 1.296,
+      "step": 16741
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.923403769942216e-05,
+      "loss": 1.5528,
+      "step": 16742
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9224852738144238e-05,
+      "loss": 1.047,
+      "step": 16743
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9215669737236008e-05,
+      "loss": 0.8255,
+      "step": 16744
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9206488696920298e-05,
+      "loss": 0.7806,
+      "step": 16745
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.919730961741998e-05,
+      "loss": 0.9095,
+      "step": 16746
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.91881324989578e-05,
+      "loss": 0.8292,
+      "step": 16747
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9178957341756486e-05,
+      "loss": 0.776,
+      "step": 16748
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9169784146038682e-05,
+      "loss": 0.8828,
+      "step": 16749
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9160612912027066e-05,
+      "loss": 1.2322,
+      "step": 16750
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.91514436399442e-05,
+      "loss": 1.0204,
+      "step": 16751
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9142276330012597e-05,
+      "loss": 0.8992,
+      "step": 16752
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.913311098245478e-05,
+      "loss": 0.7857,
+      "step": 16753
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9123947597493175e-05,
+      "loss": 0.8282,
+      "step": 16754
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.911478617535013e-05,
+      "loss": 0.7736,
+      "step": 16755
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9105626716248082e-05,
+      "loss": 0.8651,
+      "step": 16756
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.909646922040924e-05,
+      "loss": 0.7764,
+      "step": 16757
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9087313688055852e-05,
+      "loss": 0.8871,
+      "step": 16758
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9078160119410172e-05,
+      "loss": 0.8139,
+      "step": 16759
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.906900851469432e-05,
+      "loss": 0.8011,
+      "step": 16760
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9059858874130387e-05,
+      "loss": 1.2655,
+      "step": 16761
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9050711197940474e-05,
+      "loss": 1.0829,
+      "step": 16762
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9041565486346557e-05,
+      "loss": 0.7262,
+      "step": 16763
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9032421739570618e-05,
+      "loss": 0.7978,
+      "step": 16764
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9023279957834527e-05,
+      "loss": 0.8423,
+      "step": 16765
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9014140141360205e-05,
+      "loss": 0.7498,
+      "step": 16766
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9005002290369444e-05,
+      "loss": 0.8522,
+      "step": 16767
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8995866405083996e-05,
+      "loss": 0.7659,
+      "step": 16768
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8986732485725633e-05,
+      "loss": 0.6893,
+      "step": 16769
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8977600532516003e-05,
+      "loss": 0.7671,
+      "step": 16770
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8968470545676732e-05,
+      "loss": 0.8922,
+      "step": 16771
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8959342525429402e-05,
+      "loss": 0.7866,
+      "step": 16772
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.895021647199554e-05,
+      "loss": 0.748,
+      "step": 16773
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8941092385596617e-05,
+      "loss": 0.825,
+      "step": 16774
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.89319702664541e-05,
+      "loss": 1.3515,
+      "step": 16775
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.892285011478938e-05,
+      "loss": 1.1341,
+      "step": 16776
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8913731930823753e-05,
+      "loss": 1.3587,
+      "step": 16777
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8904615714778574e-05,
+      "loss": 1.2246,
+      "step": 16778
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8895501466875055e-05,
+      "loss": 0.861,
+      "step": 16779
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8886389187334397e-05,
+      "loss": 1.2999,
+      "step": 16780
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8877278876377726e-05,
+      "loss": 1.1121,
+      "step": 16781
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8868170534226192e-05,
+      "loss": 0.8612,
+      "step": 16782
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8859064161100827e-05,
+      "loss": 0.8772,
+      "step": 16783
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.884995975722261e-05,
+      "loss": 0.7613,
+      "step": 16784
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8840857322812578e-05,
+      "loss": 0.8866,
+      "step": 16785
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8831756858091563e-05,
+      "loss": 0.933,
+      "step": 16786
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.8822658363280444e-05,
+      "loss": 1.1669,
+      "step": 16787
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.881356183860006e-05,
+      "loss": 1.1278,
+      "step": 16788
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8804467284271178e-05,
+      "loss": 0.6673,
+      "step": 16789
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8795374700514477e-05,
+      "loss": 0.8198,
+      "step": 16790
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8786284087550687e-05,
+      "loss": 0.8546,
+      "step": 16791
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8777195445600405e-05,
+      "loss": 0.7087,
+      "step": 16792
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8768108774884184e-05,
+      "loss": 0.9102,
+      "step": 16793
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.87590240756226e-05,
+      "loss": 0.7379,
+      "step": 16794
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8749941348036114e-05,
+      "loss": 0.7462,
+      "step": 16795
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8740860592345156e-05,
+      "loss": 0.7445,
+      "step": 16796
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8731781808770088e-05,
+      "loss": 0.984,
+      "step": 16797
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.872270499753128e-05,
+      "loss": 0.7162,
+      "step": 16798
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8713630158849027e-05,
+      "loss": 0.6486,
+      "step": 16799
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.870455729294355e-05,
+      "loss": 1.2516,
+      "step": 16800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8695486400035046e-05,
+      "loss": 1.1126,
+      "step": 16801
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.868641748034363e-05,
+      "loss": 0.8113,
+      "step": 16802
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8677350534089467e-05,
+      "loss": 0.9088,
+      "step": 16803
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8668285561492572e-05,
+      "loss": 0.7906,
+      "step": 16804
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.865922256277295e-05,
+      "loss": 1.2834,
+      "step": 16805
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.865016153815051e-05,
+      "loss": 1.0161,
+      "step": 16806
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.864110248784524e-05,
+      "loss": 0.8499,
+      "step": 16807
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.863204541207696e-05,
+      "loss": 0.8727,
+      "step": 16808
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8622990311065447e-05,
+      "loss": 0.9109,
+      "step": 16809
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8613937185030518e-05,
+      "loss": 0.8419,
+      "step": 16810
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.860488603419187e-05,
+      "loss": 0.989,
+      "step": 16811
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8595836858769132e-05,
+      "loss": 0.6905,
+      "step": 16812
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8586789658981983e-05,
+      "loss": 0.718,
+      "step": 16813
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.857774443504997e-05,
+      "loss": 1.2206,
+      "step": 16814
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8568701187192618e-05,
+      "loss": 1.1575,
+      "step": 16815
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.855965991562938e-05,
+      "loss": 0.7654,
+      "step": 16816
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8550620620579705e-05,
+      "loss": 0.8652,
+      "step": 16817
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8541583302262944e-05,
+      "loss": 0.8359,
+      "step": 16818
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8532547960898473e-05,
+      "loss": 0.9014,
+      "step": 16819
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8523514596705548e-05,
+      "loss": 0.9178,
+      "step": 16820
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.851448320990341e-05,
+      "loss": 0.8136,
+      "step": 16821
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8505453800711215e-05,
+      "loss": 0.8602,
+      "step": 16822
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8496426369348153e-05,
+      "loss": 0.7651,
+      "step": 16823
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8487400916033293e-05,
+      "loss": 0.7787,
+      "step": 16824
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.847837744098565e-05,
+      "loss": 0.7364,
+      "step": 16825
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.846935594442427e-05,
+      "loss": 0.7776,
+      "step": 16826
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.846033642656808e-05,
+      "loss": 0.8617,
+      "step": 16827
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.845131888763595e-05,
+      "loss": 0.7731,
+      "step": 16828
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.84423033278468e-05,
+      "loss": 0.9619,
+      "step": 16829
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8433289747419357e-05,
+      "loss": 0.8,
+      "step": 16830
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8424278146572393e-05,
+      "loss": 0.7524,
+      "step": 16831
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8415268525524654e-05,
+      "loss": 0.849,
+      "step": 16832
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8406260884494762e-05,
+      "loss": 0.8693,
+      "step": 16833
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8397255223701328e-05,
+      "loss": 0.8916,
+      "step": 16834
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8388251543362946e-05,
+      "loss": 0.7189,
+      "step": 16835
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.83792498436981e-05,
+      "loss": 0.8149,
+      "step": 16836
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.837025012492527e-05,
+      "loss": 0.8515,
+      "step": 16837
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8361252387262852e-05,
+      "loss": 0.6938,
+      "step": 16838
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8352256630929255e-05,
+      "loss": 0.7643,
+      "step": 16839
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8343262856142784e-05,
+      "loss": 0.8896,
+      "step": 16840
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8334271063121687e-05,
+      "loss": 0.8122,
+      "step": 16841
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8325281252084236e-05,
+      "loss": 0.739,
+      "step": 16842
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8316293423248588e-05,
+      "loss": 0.8665,
+      "step": 16843
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8307307576832865e-05,
+      "loss": 0.7769,
+      "step": 16844
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8298323713055154e-05,
+      "loss": 0.8997,
+      "step": 16845
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.82893418321335e-05,
+      "loss": 0.8462,
+      "step": 16846
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.828036193428584e-05,
+      "loss": 0.7462,
+      "step": 16847
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8271384019730188e-05,
+      "loss": 0.9053,
+      "step": 16848
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.826240808868438e-05,
+      "loss": 0.8756,
+      "step": 16849
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8253434141366255e-05,
+      "loss": 0.8026,
+      "step": 16850
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8244462177993637e-05,
+      "loss": 0.7743,
+      "step": 16851
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8235492198784267e-05,
+      "loss": 0.8411,
+      "step": 16852
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.822652420395583e-05,
+      "loss": 0.6635,
+      "step": 16853
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8217558193725947e-05,
+      "loss": 0.9132,
+      "step": 16854
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8208594168312264e-05,
+      "loss": 0.8295,
+      "step": 16855
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.819963212793233e-05,
+      "loss": 0.8347,
+      "step": 16856
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8190672072803605e-05,
+      "loss": 0.8719,
+      "step": 16857
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.818171400314359e-05,
+      "loss": 0.7288,
+      "step": 16858
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8172757919169714e-05,
+      "loss": 0.9165,
+      "step": 16859
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8163803821099233e-05,
+      "loss": 0.8229,
+      "step": 16860
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8154851709149567e-05,
+      "loss": 0.8978,
+      "step": 16861
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.814590158353793e-05,
+      "loss": 0.8998,
+      "step": 16862
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.813695344448151e-05,
+      "loss": 1.012,
+      "step": 16863
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8128007292197536e-05,
+      "loss": 0.8828,
+      "step": 16864
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8119063126903102e-05,
+      "loss": 0.8138,
+      "step": 16865
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8110120948815244e-05,
+      "loss": 0.8761,
+      "step": 16866
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8101180758151036e-05,
+      "loss": 0.7167,
+      "step": 16867
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8092242555127437e-05,
+      "loss": 0.7124,
+      "step": 16868
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8083306339961325e-05,
+      "loss": 0.849,
+      "step": 16869
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8074372112869652e-05,
+      "loss": 0.8149,
+      "step": 16870
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.806543987406921e-05,
+      "loss": 0.8548,
+      "step": 16871
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8056509623776774e-05,
+      "loss": 0.8011,
+      "step": 16872
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8047581362209087e-05,
+      "loss": 0.9084,
+      "step": 16873
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8038655089582836e-05,
+      "loss": 0.8535,
+      "step": 16874
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8029730806114608e-05,
+      "loss": 0.8282,
+      "step": 16875
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.802080851202107e-05,
+      "loss": 0.7841,
+      "step": 16876
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8011888207518725e-05,
+      "loss": 0.6725,
+      "step": 16877
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8002969892824062e-05,
+      "loss": 0.8628,
+      "step": 16878
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.79940535681535e-05,
+      "loss": 0.8296,
+      "step": 16879
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.798513923372348e-05,
+      "loss": 0.8853,
+      "step": 16880
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7976226889750323e-05,
+      "loss": 0.8144,
+      "step": 16881
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7967316536450306e-05,
+      "loss": 0.762,
+      "step": 16882
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.795840817403973e-05,
+      "loss": 0.7309,
+      "step": 16883
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7949501802734782e-05,
+      "loss": 0.838,
+      "step": 16884
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.794059742275156e-05,
+      "loss": 0.736,
+      "step": 16885
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.793169503430624e-05,
+      "loss": 0.7796,
+      "step": 16886
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7922794637614846e-05,
+      "loss": 0.6061,
+      "step": 16887
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.791389623289339e-05,
+      "loss": 0.9258,
+      "step": 16888
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7904999820357826e-05,
+      "loss": 0.8178,
+      "step": 16889
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7896105400224074e-05,
+      "loss": 0.8049,
+      "step": 16890
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7887212972707968e-05,
+      "loss": 0.8963,
+      "step": 16891
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.7878322538025372e-05,
+      "loss": 1.3077,
+      "step": 16892
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7869434096392022e-05,
+      "loss": 1.1094,
+      "step": 16893
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.786054764802364e-05,
+      "loss": 0.8354,
+      "step": 16894
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.785166319313588e-05,
+      "loss": 0.6957,
+      "step": 16895
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.78427807319444e-05,
+      "loss": 1.2125,
+      "step": 16896
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7833900264664737e-05,
+      "loss": 1.0377,
+      "step": 16897
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7825021791512418e-05,
+      "loss": 0.8728,
+      "step": 16898
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7816145312702948e-05,
+      "loss": 1.2336,
+      "step": 16899
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7807270828451737e-05,
+      "loss": 1.068,
+      "step": 16900
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7798398338974142e-05,
+      "loss": 1.0087,
+      "step": 16901
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7789527844485543e-05,
+      "loss": 1.1919,
+      "step": 16902
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7780659345201222e-05,
+      "loss": 1.0772,
+      "step": 16903
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.777179284133632e-05,
+      "loss": 1.2098,
+      "step": 16904
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.776292833310612e-05,
+      "loss": 1.113,
+      "step": 16905
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7754065820725728e-05,
+      "loss": 0.9629,
+      "step": 16906
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7745205304410205e-05,
+      "loss": 0.8542,
+      "step": 16907
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7736346784374648e-05,
+      "loss": 0.6755,
+      "step": 16908
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7727490260834013e-05,
+      "loss": 1.0419,
+      "step": 16909
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.771863573400324e-05,
+      "loss": 0.8453,
+      "step": 16910
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7709783204097208e-05,
+      "loss": 0.9179,
+      "step": 16911
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7700932671330815e-05,
+      "loss": 0.7945,
+      "step": 16912
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.769208413591883e-05,
+      "loss": 0.813,
+      "step": 16913
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7683237598075976e-05,
+      "loss": 0.7469,
+      "step": 16914
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7674393058017002e-05,
+      "loss": 1.1471,
+      "step": 16915
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7665550515956543e-05,
+      "loss": 1.0923,
+      "step": 16916
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7656709972109174e-05,
+      "loss": 0.7625,
+      "step": 16917
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.764787142668952e-05,
+      "loss": 0.7545,
+      "step": 16918
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.763903487991202e-05,
+      "loss": 0.7045,
+      "step": 16919
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.763020033199112e-05,
+      "loss": 0.788,
+      "step": 16920
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7621367783141296e-05,
+      "loss": 0.9158,
+      "step": 16921
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7612537233576876e-05,
+      "loss": 0.9492,
+      "step": 16922
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7603708683512153e-05,
+      "loss": 1.3363,
+      "step": 16923
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7594882133161427e-05,
+      "loss": 0.997,
+      "step": 16924
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.758605758273891e-05,
+      "loss": 0.7928,
+      "step": 16925
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.757723503245875e-05,
+      "loss": 0.8568,
+      "step": 16926
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7568414482535057e-05,
+      "loss": 0.7175,
+      "step": 16927
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7559595933181927e-05,
+      "loss": 0.8334,
+      "step": 16928
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.755077938461338e-05,
+      "loss": 0.7539,
+      "step": 16929
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7541964837043356e-05,
+      "loss": 0.8551,
+      "step": 16930
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.753315229068584e-05,
+      "loss": 0.8554,
+      "step": 16931
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7524341745754657e-05,
+      "loss": 0.8403,
+      "step": 16932
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.751553320246365e-05,
+      "loss": 0.7512,
+      "step": 16933
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7506726661026607e-05,
+      "loss": 0.8419,
+      "step": 16934
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.749792212165724e-05,
+      "loss": 0.9749,
+      "step": 16935
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7489119584569214e-05,
+      "loss": 0.8841,
+      "step": 16936
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7480319049976212e-05,
+      "loss": 0.6808,
+      "step": 16937
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.74715205180918e-05,
+      "loss": 1.3295,
+      "step": 16938
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.746272398912947e-05,
+      "loss": 1.0948,
+      "step": 16939
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7453929463302775e-05,
+      "loss": 1.2194,
+      "step": 16940
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7445136940825124e-05,
+      "loss": 1.4347,
+      "step": 16941
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7436346421909876e-05,
+      "loss": 1.1573,
+      "step": 16942
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7427557906770442e-05,
+      "loss": 0.9902,
+      "step": 16943
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.741877139562006e-05,
+      "loss": 0.732,
+      "step": 16944
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7409986888671992e-05,
+      "loss": 0.9028,
+      "step": 16945
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7401204386139413e-05,
+      "loss": 0.6962,
+      "step": 16946
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7392423888235533e-05,
+      "loss": 1.2268,
+      "step": 16947
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7383645395173352e-05,
+      "loss": 1.0191,
+      "step": 16948
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7374868907166e-05,
+      "loss": 0.8507,
+      "step": 16949
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7366094424426448e-05,
+      "loss": 0.7085,
+      "step": 16950
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.735732194716765e-05,
+      "loss": 0.8648,
+      "step": 16951
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7348551475602482e-05,
+      "loss": 0.8979,
+      "step": 16952
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7339783009943855e-05,
+      "loss": 0.7615,
+      "step": 16953
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7331016550404534e-05,
+      "loss": 0.8644,
+      "step": 16954
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.732225209719727e-05,
+      "loss": 0.7662,
+      "step": 16955
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7313489650534808e-05,
+      "loss": 0.8442,
+      "step": 16956
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.73047292106298e-05,
+      "loss": 0.7177,
+      "step": 16957
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7295970777694813e-05,
+      "loss": 0.8641,
+      "step": 16958
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7287214351942472e-05,
+      "loss": 0.7437,
+      "step": 16959
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7278459933585256e-05,
+      "loss": 0.9391,
+      "step": 16960
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7269707522835645e-05,
+      "loss": 1.2121,
+      "step": 16961
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.726095711990604e-05,
+      "loss": 1.1279,
+      "step": 16962
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.725220872500881e-05,
+      "loss": 0.7994,
+      "step": 16963
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7243462338356253e-05,
+      "loss": 0.8724,
+      "step": 16964
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7234717960160695e-05,
+      "loss": 0.8587,
+      "step": 16965
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7225975590634325e-05,
+      "loss": 0.8459,
+      "step": 16966
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7217235229989324e-05,
+      "loss": 0.863,
+      "step": 16967
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7208496878437773e-05,
+      "loss": 0.893,
+      "step": 16968
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.719976053619181e-05,
+      "loss": 0.7289,
+      "step": 16969
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7191026203463446e-05,
+      "loss": 0.8693,
+      "step": 16970
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7182293880464617e-05,
+      "loss": 0.8792,
+      "step": 16971
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7173563567407303e-05,
+      "loss": 0.7554,
+      "step": 16972
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7164835264503366e-05,
+      "loss": 0.8405,
+      "step": 16973
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7156108971964603e-05,
+      "loss": 0.8044,
+      "step": 16974
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7147384690002865e-05,
+      "loss": 0.8568,
+      "step": 16975
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7138662418829866e-05,
+      "loss": 0.8805,
+      "step": 16976
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7129942158657218e-05,
+      "loss": 0.8248,
+      "step": 16977
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.712122390969665e-05,
+      "loss": 0.792,
+      "step": 16978
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7112507672159707e-05,
+      "loss": 0.8467,
+      "step": 16979
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.71037934462579e-05,
+      "loss": 0.7585,
+      "step": 16980
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.709508123220278e-05,
+      "loss": 0.852,
+      "step": 16981
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.708637103020576e-05,
+      "loss": 0.8187,
+      "step": 16982
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7077662840478237e-05,
+      "loss": 0.8449,
+      "step": 16983
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7068956663231516e-05,
+      "loss": 0.8013,
+      "step": 16984
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7060252498676955e-05,
+      "loss": 0.8082,
+      "step": 16985
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7051550347025757e-05,
+      "loss": 0.7885,
+      "step": 16986
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7042850208489102e-05,
+      "loss": 1.1704,
+      "step": 16987
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7034152083278198e-05,
+      "loss": 1.212,
+      "step": 16988
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.702545597160411e-05,
+      "loss": 0.7656,
+      "step": 16989
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7016761873677866e-05,
+      "loss": 0.9636,
+      "step": 16990
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.7008069789710534e-05,
+      "loss": 0.8897,
+      "step": 16991
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.6999379719913e-05,
+      "loss": 0.905,
+      "step": 16992
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.699069166449616e-05,
+      "loss": 0.8068,
+      "step": 16993
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.6982005623670926e-05,
+      "loss": 0.7951,
+      "step": 16994
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.6973321597648073e-05,
+      "loss": 0.8864,
+      "step": 16995
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.696463958663833e-05,
+      "loss": 0.7676,
+      "step": 16996
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6955959590852467e-05,
+      "loss": 0.9153,
+      "step": 16997
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.694728161050111e-05,
+      "loss": 0.7313,
+      "step": 16998
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.693860564579487e-05,
+      "loss": 0.7647,
+      "step": 16999
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.692993169694428e-05,
+      "loss": 0.86,
+      "step": 17000
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6921259764159912e-05,
+      "loss": 0.8858,
+      "step": 17001
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6912589847652184e-05,
+      "loss": 1.3158,
+      "step": 17002
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6903921947631506e-05,
+      "loss": 1.0576,
+      "step": 17003
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6895256064308283e-05,
+      "loss": 0.8289,
+      "step": 17004
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6886592197892816e-05,
+      "loss": 1.3184,
+      "step": 17005
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.687793034859536e-05,
+      "loss": 1.1621,
+      "step": 17006
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6869270516626136e-05,
+      "loss": 0.7885,
+      "step": 17007
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.686061270219531e-05,
+      "loss": 0.9192,
+      "step": 17008
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6851956905513e-05,
+      "loss": 0.8522,
+      "step": 17009
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.68433031267893e-05,
+      "loss": 0.7528,
+      "step": 17010
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6834651366234223e-05,
+      "loss": 1.1182,
+      "step": 17011
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6826001624057708e-05,
+      "loss": 1.1821,
+      "step": 17012
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6817353900469736e-05,
+      "loss": 0.7391,
+      "step": 17013
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6808708195680155e-05,
+      "loss": 1.2601,
+      "step": 17014
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6800064509898784e-05,
+      "loss": 0.9867,
+      "step": 17015
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6791422843335415e-05,
+      "loss": 0.8147,
+      "step": 17016
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.678278319619978e-05,
+      "loss": 0.7169,
+      "step": 17017
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.677414556870155e-05,
+      "loss": 0.8026,
+      "step": 17018
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.676550996105033e-05,
+      "loss": 0.8209,
+      "step": 17019
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6756876373455776e-05,
+      "loss": 0.9221,
+      "step": 17020
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.674824480612732e-05,
+      "loss": 0.8771,
+      "step": 17021
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.673961525927453e-05,
+      "loss": 0.778,
+      "step": 17022
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6730987733106806e-05,
+      "loss": 0.6712,
+      "step": 17023
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6722362227833533e-05,
+      "loss": 0.8799,
+      "step": 17024
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.671373874366403e-05,
+      "loss": 0.8028,
+      "step": 17025
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6705117280807624e-05,
+      "loss": 0.9564,
+      "step": 17026
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6696497839473534e-05,
+      "loss": 0.8046,
+      "step": 17027
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6687880419870917e-05,
+      "loss": 0.748,
+      "step": 17028
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.667926502220898e-05,
+      "loss": 0.7472,
+      "step": 17029
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.667065164669678e-05,
+      "loss": 0.7879,
+      "step": 17030
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6662040293543324e-05,
+      "loss": 0.7628,
+      "step": 17031
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.665343096295766e-05,
+      "loss": 0.8285,
+      "step": 17032
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6644823655148722e-05,
+      "loss": 1.1153,
+      "step": 17033
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6636218370325397e-05,
+      "loss": 1.0933,
+      "step": 17034
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6627615108696514e-05,
+      "loss": 0.8655,
+      "step": 17035
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.661901387047089e-05,
+      "loss": 0.788,
+      "step": 17036
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6610414655857243e-05,
+      "loss": 0.8965,
+      "step": 17037
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6601817465064318e-05,
+      "loss": 0.7918,
+      "step": 17038
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6593222298300727e-05,
+      "loss": 0.8179,
+      "step": 17039
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6584629155775098e-05,
+      "loss": 0.7617,
+      "step": 17040
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.657603803769594e-05,
+      "loss": 0.9429,
+      "step": 17041
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6567448944271802e-05,
+      "loss": 0.7113,
+      "step": 17042
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.655886187571112e-05,
+      "loss": 0.91,
+      "step": 17043
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6550276832222255e-05,
+      "loss": 0.9753,
+      "step": 17044
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6541693814013637e-05,
+      "loss": 1.3051,
+      "step": 17045
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.653311282129354e-05,
+      "loss": 1.004,
+      "step": 17046
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6524533854270184e-05,
+      "loss": 0.8469,
+      "step": 17047
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.651595691315183e-05,
+      "loss": 0.7677,
+      "step": 17048
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6507381998146622e-05,
+      "loss": 0.7763,
+      "step": 17049
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6498809109462644e-05,
+      "loss": 0.8006,
+      "step": 17050
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6490238247307977e-05,
+      "loss": 0.916,
+      "step": 17051
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.648166941189063e-05,
+      "loss": 0.6248,
+      "step": 17052
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6473102603418532e-05,
+      "loss": 0.6962,
+      "step": 17053
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6464537822099647e-05,
+      "loss": 0.8703,
+      "step": 17054
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6455975068141803e-05,
+      "loss": 0.647,
+      "step": 17055
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.644741434175283e-05,
+      "loss": 0.9392,
+      "step": 17056
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.643885564314046e-05,
+      "loss": 0.8611,
+      "step": 17057
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6430298972512458e-05,
+      "loss": 0.7344,
+      "step": 17058
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.642174433007646e-05,
+      "loss": 0.8628,
+      "step": 17059
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6413191716040077e-05,
+      "loss": 0.7834,
+      "step": 17060
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.64046411306109e-05,
+      "loss": 0.8387,
+      "step": 17061
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6396092573996436e-05,
+      "loss": 0.7673,
+      "step": 17062
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6387546046404122e-05,
+      "loss": 0.8951,
+      "step": 17063
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6379001548041474e-05,
+      "loss": 0.8289,
+      "step": 17064
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6370459079115754e-05,
+      "loss": 0.8509,
+      "step": 17065
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6361918639834305e-05,
+      "loss": 0.8729,
+      "step": 17066
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6353380230404448e-05,
+      "loss": 0.746,
+      "step": 17067
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6344843851033376e-05,
+      "loss": 0.7357,
+      "step": 17068
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6336309501928237e-05,
+      "loss": 1.3775,
+      "step": 17069
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.63277771832962e-05,
+      "loss": 1.0705,
+      "step": 17070
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6319246895344332e-05,
+      "loss": 1.395,
+      "step": 17071
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.631071863827962e-05,
+      "loss": 1.0617,
+      "step": 17072
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6302192412309092e-05,
+      "loss": 0.8491,
+      "step": 17073
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6293668217639646e-05,
+      "loss": 0.7232,
+      "step": 17074
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.628514605447816e-05,
+      "loss": 0.7944,
+      "step": 17075
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.627662592303145e-05,
+      "loss": 0.7475,
+      "step": 17076
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.626810782350634e-05,
+      "loss": 0.9073,
+      "step": 17077
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.625959175610953e-05,
+      "loss": 0.9194,
+      "step": 17078
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6251077721047704e-05,
+      "loss": 1.278,
+      "step": 17079
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6242565718527482e-05,
+      "loss": 1.2906,
+      "step": 17080
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6234055748755463e-05,
+      "loss": 0.8319,
+      "step": 17081
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6225547811938137e-05,
+      "loss": 0.7785,
+      "step": 17082
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.621704190828206e-05,
+      "loss": 0.95,
+      "step": 17083
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6208538037993615e-05,
+      "loss": 0.8531,
+      "step": 17084
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6200036201279178e-05,
+      "loss": 0.9652,
+      "step": 17085
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6191536398345132e-05,
+      "loss": 0.8745,
+      "step": 17086
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.618303862939773e-05,
+      "loss": 0.9352,
+      "step": 17087
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6174542894643186e-05,
+      "loss": 0.8448,
+      "step": 17088
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6166049194287747e-05,
+      "loss": 0.9149,
+      "step": 17089
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6157557528537515e-05,
+      "loss": 0.8949,
+      "step": 17090
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.614906789759858e-05,
+      "loss": 0.8972,
+      "step": 17091
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6140580301676955e-05,
+      "loss": 1.2628,
+      "step": 17092
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6132094740978687e-05,
+      "loss": 1.1603,
+      "step": 17093
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.612361121570968e-05,
+      "loss": 0.7847,
+      "step": 17094
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.611512972607584e-05,
+      "loss": 0.7513,
+      "step": 17095
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6106650272282986e-05,
+      "loss": 0.6905,
+      "step": 17096
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6098172854536932e-05,
+      "loss": 0.8446,
+      "step": 17097
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6089697473043376e-05,
+      "loss": 0.8939,
+      "step": 17098
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6081224128008078e-05,
+      "loss": 0.7151,
+      "step": 17099
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6072752819636637e-05,
+      "loss": 0.7905,
+      "step": 17100
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.6064283548134653e-05,
+      "loss": 0.6992,
+      "step": 17101
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.6055816313707685e-05,
+      "loss": 0.7035,
+      "step": 17102
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.6047351116561228e-05,
+      "loss": 1.331,
+      "step": 17103
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.6038887956900694e-05,
+      "loss": 1.1389,
+      "step": 17104
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.6030426834931533e-05,
+      "loss": 0.8473,
+      "step": 17105
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.6021967750859058e-05,
+      "loss": 0.8613,
+      "step": 17106
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.601351070488858e-05,
+      "loss": 0.9134,
+      "step": 17107
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.600505569722531e-05,
+      "loss": 0.9007,
+      "step": 17108
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5996602728074526e-05,
+      "loss": 0.8175,
+      "step": 17109
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5988151797641284e-05,
+      "loss": 0.7664,
+      "step": 17110
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5979702906130745e-05,
+      "loss": 0.7765,
+      "step": 17111
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.597125605374794e-05,
+      "loss": 0.91,
+      "step": 17112
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5962811240697872e-05,
+      "loss": 0.9143,
+      "step": 17113
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5954368467185464e-05,
+      "loss": 1.2565,
+      "step": 17114
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5945927733415668e-05,
+      "loss": 0.9953,
+      "step": 17115
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5937489039593313e-05,
+      "loss": 0.8346,
+      "step": 17116
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.592905238592318e-05,
+      "loss": 0.6944,
+      "step": 17117
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.592061777261007e-05,
+      "loss": 0.7368,
+      "step": 17118
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5912185199858653e-05,
+      "loss": 0.8035,
+      "step": 17119
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5903754667873572e-05,
+      "loss": 0.8553,
+      "step": 17120
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.589532617685946e-05,
+      "loss": 0.88,
+      "step": 17121
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5886899727020877e-05,
+      "loss": 1.1902,
+      "step": 17122
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5878475318562304e-05,
+      "loss": 1.0343,
+      "step": 17123
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.587005295168821e-05,
+      "loss": 0.7883,
+      "step": 17124
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5861632626602985e-05,
+      "loss": 0.7924,
+      "step": 17125
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.585321434351098e-05,
+      "loss": 0.8685,
+      "step": 17126
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5844798102616543e-05,
+      "loss": 0.8759,
+      "step": 17127
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.583638390412392e-05,
+      "loss": 0.7334,
+      "step": 17128
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.582797174823729e-05,
+      "loss": 1.0,
+      "step": 17129
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.581956163516082e-05,
+      "loss": 0.7117,
+      "step": 17130
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5811153565098648e-05,
+      "loss": 1.1381,
+      "step": 17131
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5802747538254803e-05,
+      "loss": 1.1609,
+      "step": 17132
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5794343554833292e-05,
+      "loss": 0.8012,
+      "step": 17133
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5785941615038104e-05,
+      "loss": 0.9208,
+      "step": 17134
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.577754171907314e-05,
+      "loss": 1.3247,
+      "step": 17135
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5769143867142233e-05,
+      "loss": 1.113,
+      "step": 17136
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.576074805944924e-05,
+      "loss": 0.7973,
+      "step": 17137
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5752354296197914e-05,
+      "loss": 0.8283,
+      "step": 17138
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.574396257759191e-05,
+      "loss": 0.9584,
+      "step": 17139
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5735572903834962e-05,
+      "loss": 0.8597,
+      "step": 17140
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.572718527513064e-05,
+      "loss": 0.7867,
+      "step": 17141
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5718799691682507e-05,
+      "loss": 0.8614,
+      "step": 17142
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5710416153694106e-05,
+      "loss": 0.8542,
+      "step": 17143
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.57020346613689e-05,
+      "loss": 0.9667,
+      "step": 17144
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5693655214910252e-05,
+      "loss": 0.8881,
+      "step": 17145
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5685277814521594e-05,
+      "loss": 0.7762,
+      "step": 17146
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5676902460406206e-05,
+      "loss": 0.8789,
+      "step": 17147
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5668529152767363e-05,
+      "loss": 1.275,
+      "step": 17148
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.566015789180826e-05,
+      "loss": 1.1194,
+      "step": 17149
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5651788677732083e-05,
+      "loss": 0.5981,
+      "step": 17150
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5643421510741952e-05,
+      "loss": 0.8033,
+      "step": 17151
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5635056391040904e-05,
+      "loss": 0.8637,
+      "step": 17152
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5626693318832032e-05,
+      "loss": 0.8119,
+      "step": 17153
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5618332294318216e-05,
+      "loss": 0.775,
+      "step": 17154
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5609973317702388e-05,
+      "loss": 0.8211,
+      "step": 17155
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5601616389187445e-05,
+      "loss": 0.7324,
+      "step": 17156
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5593261508976208e-05,
+      "loss": 0.886,
+      "step": 17157
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5584908677271403e-05,
+      "loss": 0.7641,
+      "step": 17158
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.55765578942758e-05,
+      "loss": 0.8071,
+      "step": 17159
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.556820916019206e-05,
+      "loss": 0.8503,
+      "step": 17160
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5559862475222752e-05,
+      "loss": 0.8733,
+      "step": 17161
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5551517839570507e-05,
+      "loss": 0.8687,
+      "step": 17162
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5543175253437813e-05,
+      "loss": 0.7251,
+      "step": 17163
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5534834717027158e-05,
+      "loss": 0.7164,
+      "step": 17164
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.552649623054092e-05,
+      "loss": 0.8941,
+      "step": 17165
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.551815979418153e-05,
+      "loss": 0.7824,
+      "step": 17166
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.550982540815128e-05,
+      "loss": 0.7573,
+      "step": 17167
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.550149307265244e-05,
+      "loss": 1.2047,
+      "step": 17168
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5493162787887227e-05,
+      "loss": 1.0566,
+      "step": 17169
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5484834554057825e-05,
+      "loss": 0.8932,
+      "step": 17170
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5476508371366317e-05,
+      "loss": 0.9375,
+      "step": 17171
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5468184240014837e-05,
+      "loss": 0.8716,
+      "step": 17172
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5459862160205372e-05,
+      "loss": 0.7776,
+      "step": 17173
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5451542132139885e-05,
+      "loss": 0.7028,
+      "step": 17174
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.544322415602032e-05,
+      "loss": 0.7583,
+      "step": 17175
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.543490823204855e-05,
+      "loss": 0.87,
+      "step": 17176
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5426594360426372e-05,
+      "loss": 0.939,
+      "step": 17177
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.54182825413556e-05,
+      "loss": 1.1913,
+      "step": 17178
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.540997277503793e-05,
+      "loss": 1.222,
+      "step": 17179
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.540166506167504e-05,
+      "loss": 0.7226,
+      "step": 17180
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5393359401468543e-05,
+      "loss": 1.0074,
+      "step": 17181
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5385055794620063e-05,
+      "loss": 0.8562,
+      "step": 17182
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5376754241331036e-05,
+      "loss": 0.9304,
+      "step": 17183
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5368454741803016e-05,
+      "loss": 0.8364,
+      "step": 17184
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5360157296237388e-05,
+      "loss": 0.6878,
+      "step": 17185
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.535186190483554e-05,
+      "loss": 0.7917,
+      "step": 17186
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.534356856779877e-05,
+      "loss": 0.8672,
+      "step": 17187
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5335277285328407e-05,
+      "loss": 0.8229,
+      "step": 17188
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5326988057625635e-05,
+      "loss": 0.7455,
+      "step": 17189
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5318700884891613e-05,
+      "loss": 0.8764,
+      "step": 17190
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5310415767327522e-05,
+      "loss": 0.8645,
+      "step": 17191
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5302132705134408e-05,
+      "loss": 0.8152,
+      "step": 17192
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5293851698513274e-05,
+      "loss": 0.736,
+      "step": 17193
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5285572747665132e-05,
+      "loss": 0.8619,
+      "step": 17194
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5277295852790898e-05,
+      "loss": 0.9111,
+      "step": 17195
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5269021014091433e-05,
+      "loss": 0.6512,
+      "step": 17196
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5260748231767587e-05,
+      "loss": 0.8054,
+      "step": 17197
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5252477506020112e-05,
+      "loss": 0.8661,
+      "step": 17198
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5244208837049733e-05,
+      "loss": 0.8505,
+      "step": 17199
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5235942225057153e-05,
+      "loss": 0.7697,
+      "step": 17200
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5227677670242979e-05,
+      "loss": 0.8357,
+      "step": 17201
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5219415172807794e-05,
+      "loss": 0.8185,
+      "step": 17202
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5211154732952094e-05,
+      "loss": 0.7345,
+      "step": 17203
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.5202896350876406e-05,
+      "loss": 0.9045,
+      "step": 17204
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5194640026781137e-05,
+      "loss": 0.8131,
+      "step": 17205
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.518638576086664e-05,
+      "loss": 0.8445,
+      "step": 17206
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5178133553333273e-05,
+      "loss": 0.9391,
+      "step": 17207
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5169883404381314e-05,
+      "loss": 0.7693,
+      "step": 17208
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5161635314210942e-05,
+      "loss": 0.8657,
+      "step": 17209
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5153389283022401e-05,
+      "loss": 0.8306,
+      "step": 17210
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5145145311015807e-05,
+      "loss": 0.7709,
+      "step": 17211
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5136903398391166e-05,
+      "loss": 0.8663,
+      "step": 17212
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5128663545348575e-05,
+      "loss": 0.8867,
+      "step": 17213
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5120425752087986e-05,
+      "loss": 0.8272,
+      "step": 17214
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5112190018809314e-05,
+      "loss": 0.7215,
+      "step": 17215
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5103956345712478e-05,
+      "loss": 0.7826,
+      "step": 17216
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5095724732997263e-05,
+      "loss": 0.7742,
+      "step": 17217
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5087495180863454e-05,
+      "loss": 1.2555,
+      "step": 17218
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.50792676895108e-05,
+      "loss": 1.0233,
+      "step": 17219
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5071042259138956e-05,
+      "loss": 0.7341,
+      "step": 17220
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5062818889947572e-05,
+      "loss": 0.743,
+      "step": 17221
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5054597582136176e-05,
+      "loss": 0.8645,
+      "step": 17222
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5046378335904343e-05,
+      "loss": 0.744,
+      "step": 17223
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5038161151451547e-05,
+      "loss": 0.8314,
+      "step": 17224
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.502994602897717e-05,
+      "loss": 0.7979,
+      "step": 17225
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5021732968680668e-05,
+      "loss": 0.851,
+      "step": 17226
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5013521970761291e-05,
+      "loss": 0.762,
+      "step": 17227
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.5005313035418333e-05,
+      "loss": 0.6706,
+      "step": 17228
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.499710616285105e-05,
+      "loss": 0.9025,
+      "step": 17229
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4988901353258588e-05,
+      "loss": 0.807,
+      "step": 17230
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4980698606840082e-05,
+      "loss": 0.7767,
+      "step": 17231
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4972497923794627e-05,
+      "loss": 0.8994,
+      "step": 17232
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4964299304321227e-05,
+      "loss": 0.9186,
+      "step": 17233
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.495610274861885e-05,
+      "loss": 0.9074,
+      "step": 17234
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4947908256886456e-05,
+      "loss": 0.7161,
+      "step": 17235
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4939715829322897e-05,
+      "loss": 0.9489,
+      "step": 17236
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4931525466127016e-05,
+      "loss": 0.85,
+      "step": 17237
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4923337167497543e-05,
+      "loss": 0.7772,
+      "step": 17238
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.491515093363326e-05,
+      "loss": 0.8033,
+      "step": 17239
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4906966764732832e-05,
+      "loss": 0.8497,
+      "step": 17240
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4898784660994857e-05,
+      "loss": 0.838,
+      "step": 17241
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4890604622617931e-05,
+      "loss": 0.8091,
+      "step": 17242
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4882426649800574e-05,
+      "loss": 0.7137,
+      "step": 17243
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4874250742741247e-05,
+      "loss": 0.8179,
+      "step": 17244
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4866076901638392e-05,
+      "loss": 0.8313,
+      "step": 17245
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4857905126690397e-05,
+      "loss": 0.979,
+      "step": 17246
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4849735418095534e-05,
+      "loss": 0.831,
+      "step": 17247
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4841567776052145e-05,
+      "loss": 0.7848,
+      "step": 17248
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4833402200758428e-05,
+      "loss": 0.8846,
+      "step": 17249
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4825238692412513e-05,
+      "loss": 0.7668,
+      "step": 17250
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4817077251212597e-05,
+      "loss": 0.86,
+      "step": 17251
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4808917877356721e-05,
+      "loss": 0.7397,
+      "step": 17252
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4800760571042894e-05,
+      "loss": 0.8214,
+      "step": 17253
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.479260533246909e-05,
+      "loss": 0.8163,
+      "step": 17254
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4784452161833274e-05,
+      "loss": 0.9506,
+      "step": 17255
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4776301059333274e-05,
+      "loss": 0.8088,
+      "step": 17256
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4768152025166937e-05,
+      "loss": 0.8546,
+      "step": 17257
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4760005059532033e-05,
+      "loss": 0.7897,
+      "step": 17258
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4751860162626274e-05,
+      "loss": 0.8583,
+      "step": 17259
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.474371733464731e-05,
+      "loss": 0.9248,
+      "step": 17260
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.473557657579282e-05,
+      "loss": 0.7256,
+      "step": 17261
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4727437886260343e-05,
+      "loss": 0.7822,
+      "step": 17262
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4719301266247376e-05,
+      "loss": 0.7944,
+      "step": 17263
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4711166715951452e-05,
+      "loss": 0.8894,
+      "step": 17264
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4703034235569957e-05,
+      "loss": 0.7098,
+      "step": 17265
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4694903825300232e-05,
+      "loss": 0.8464,
+      "step": 17266
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4686775485339665e-05,
+      "loss": 0.8295,
+      "step": 17267
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4678649215885475e-05,
+      "loss": 0.8008,
+      "step": 17268
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4670525017134917e-05,
+      "loss": 0.8934,
+      "step": 17269
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4662402889285132e-05,
+      "loss": 0.8158,
+      "step": 17270
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.465428283253325e-05,
+      "loss": 0.6058,
+      "step": 17271
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4646164847076316e-05,
+      "loss": 0.9214,
+      "step": 17272
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4638048933111392e-05,
+      "loss": 0.8761,
+      "step": 17273
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4629935090835433e-05,
+      "loss": 0.8425,
+      "step": 17274
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4621823320445328e-05,
+      "loss": 0.7671,
+      "step": 17275
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4613713622138003e-05,
+      "loss": 0.8202,
+      "step": 17276
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4605605996110227e-05,
+      "loss": 0.8424,
+      "step": 17277
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4597500442558798e-05,
+      "loss": 0.7344,
+      "step": 17278
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.458939696168039e-05,
+      "loss": 0.7228,
+      "step": 17279
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4581295553671726e-05,
+      "loss": 0.9077,
+      "step": 17280
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.457319621872939e-05,
+      "loss": 0.9675,
+      "step": 17281
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4565098957049939e-05,
+      "loss": 0.7747,
+      "step": 17282
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4557003768829924e-05,
+      "loss": 0.7639,
+      "step": 17283
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.454891065426579e-05,
+      "loss": 0.7246,
+      "step": 17284
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4540819613553958e-05,
+      "loss": 1.3764,
+      "step": 17285
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4532730646890791e-05,
+      "loss": 1.0123,
+      "step": 17286
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.45246437544726e-05,
+      "loss": 0.8057,
+      "step": 17287
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4516558936495628e-05,
+      "loss": 0.7296,
+      "step": 17288
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4508476193156139e-05,
+      "loss": 1.3109,
+      "step": 17289
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4500395524650268e-05,
+      "loss": 1.0625,
+      "step": 17290
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.449231693117411e-05,
+      "loss": 0.7557,
+      "step": 17291
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4484240412923778e-05,
+      "loss": 0.7755,
+      "step": 17292
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4476165970095257e-05,
+      "loss": 0.8461,
+      "step": 17293
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4468093602884503e-05,
+      "loss": 0.875,
+      "step": 17294
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4460023311487413e-05,
+      "loss": 0.7993,
+      "step": 17295
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4451955096099901e-05,
+      "loss": 0.8195,
+      "step": 17296
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.444388895691774e-05,
+      "loss": 0.7958,
+      "step": 17297
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4435824894136674e-05,
+      "loss": 0.8347,
+      "step": 17298
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4427762907952458e-05,
+      "loss": 0.8097,
+      "step": 17299
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4419702998560757e-05,
+      "loss": 0.8669,
+      "step": 17300
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4411645166157118e-05,
+      "loss": 1.2818,
+      "step": 17301
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4403589410937157e-05,
+      "loss": 1.1368,
+      "step": 17302
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4395535733096355e-05,
+      "loss": 0.7707,
+      "step": 17303
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4387484132830164e-05,
+      "loss": 0.9244,
+      "step": 17304
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.437943461033403e-05,
+      "loss": 0.7507,
+      "step": 17305
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4371387165803285e-05,
+      "loss": 0.8699,
+      "step": 17306
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4363341799433217e-05,
+      "loss": 0.7445,
+      "step": 17307
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.4355298511419124e-05,
+      "loss": 1.1976,
+      "step": 17308
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4347257301956197e-05,
+      "loss": 1.0591,
+      "step": 17309
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4339218171239588e-05,
+      "loss": 0.9507,
+      "step": 17310
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4331181119464376e-05,
+      "loss": 0.6882,
+      "step": 17311
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4323146146825661e-05,
+      "loss": 0.8201,
+      "step": 17312
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4315113253518442e-05,
+      "loss": 1.0277,
+      "step": 17313
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.430708243973765e-05,
+      "loss": 0.7367,
+      "step": 17314
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.42990537056782e-05,
+      "loss": 0.7825,
+      "step": 17315
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4291027051534945e-05,
+      "loss": 0.9364,
+      "step": 17316
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.428300247750266e-05,
+      "loss": 0.818,
+      "step": 17317
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.427497998377615e-05,
+      "loss": 0.7222,
+      "step": 17318
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4266959570550087e-05,
+      "loss": 0.8884,
+      "step": 17319
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4258941238019096e-05,
+      "loss": 1.1994,
+      "step": 17320
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4250924986377834e-05,
+      "loss": 1.0814,
+      "step": 17321
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4242910815820831e-05,
+      "loss": 0.7312,
+      "step": 17322
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4234898726542545e-05,
+      "loss": 0.8092,
+      "step": 17323
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4226888718737486e-05,
+      "loss": 0.8072,
+      "step": 17324
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.421888079260003e-05,
+      "loss": 0.8246,
+      "step": 17325
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4210874948324515e-05,
+      "loss": 0.8332,
+      "step": 17326
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4202871186105227e-05,
+      "loss": 0.9098,
+      "step": 17327
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4194869506136454e-05,
+      "loss": 0.8507,
+      "step": 17328
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4186869908612366e-05,
+      "loss": 0.8538,
+      "step": 17329
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4178872393727117e-05,
+      "loss": 0.8252,
+      "step": 17330
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4170876961674795e-05,
+      "loss": 0.9008,
+      "step": 17331
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4162883612649447e-05,
+      "loss": 0.8304,
+      "step": 17332
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4154892346845039e-05,
+      "loss": 0.8466,
+      "step": 17333
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4146903164455572e-05,
+      "loss": 0.985,
+      "step": 17334
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4138916065674911e-05,
+      "loss": 0.7779,
+      "step": 17335
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4130931050696872e-05,
+      "loss": 0.8717,
+      "step": 17336
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4122948119715306e-05,
+      "loss": 0.8543,
+      "step": 17337
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4114967272923907e-05,
+      "loss": 0.9319,
+      "step": 17338
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4106988510516372e-05,
+      "loss": 1.3726,
+      "step": 17339
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4099011832686359e-05,
+      "loss": 1.0119,
+      "step": 17340
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4091037239627458e-05,
+      "loss": 0.8076,
+      "step": 17341
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4083064731533202e-05,
+      "loss": 0.8986,
+      "step": 17342
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4075094308597048e-05,
+      "loss": 0.7089,
+      "step": 17343
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4067125971012507e-05,
+      "loss": 0.9514,
+      "step": 17344
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.405915971897288e-05,
+      "loss": 0.6811,
+      "step": 17345
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.405119555267156e-05,
+      "loss": 0.8216,
+      "step": 17346
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.404323347230182e-05,
+      "loss": 0.9182,
+      "step": 17347
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4035273478056876e-05,
+      "loss": 0.7088,
+      "step": 17348
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.402731557012995e-05,
+      "loss": 0.8281,
+      "step": 17349
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4019359748714167e-05,
+      "loss": 0.741,
+      "step": 17350
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.4011406014002592e-05,
+      "loss": 0.8783,
+      "step": 17351
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.400345436618824e-05,
+      "loss": 1.4242,
+      "step": 17352
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3995504805464144e-05,
+      "loss": 1.0866,
+      "step": 17353
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3987557332023216e-05,
+      "loss": 0.9283,
+      "step": 17354
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3979611946058301e-05,
+      "loss": 0.8462,
+      "step": 17355
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3971668647762293e-05,
+      "loss": 0.8881,
+      "step": 17356
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3963727437327934e-05,
+      "loss": 0.8479,
+      "step": 17357
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3955788314947948e-05,
+      "loss": 0.7776,
+      "step": 17358
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3947851280815039e-05,
+      "loss": 1.2982,
+      "step": 17359
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3939916335121805e-05,
+      "loss": 0.8787,
+      "step": 17360
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3931983478060828e-05,
+      "loss": 0.7614,
+      "step": 17361
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3924052709824653e-05,
+      "loss": 0.7113,
+      "step": 17362
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3916124030605748e-05,
+      "loss": 0.7919,
+      "step": 17363
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3908197440596515e-05,
+      "loss": 0.8615,
+      "step": 17364
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3900272939989367e-05,
+      "loss": 0.8864,
+      "step": 17365
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3892350528976617e-05,
+      "loss": 0.7655,
+      "step": 17366
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.388443020775052e-05,
+      "loss": 0.7452,
+      "step": 17367
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3876511976503292e-05,
+      "loss": 0.825,
+      "step": 17368
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3868595835427133e-05,
+      "loss": 0.8723,
+      "step": 17369
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3860681784714157e-05,
+      "loss": 0.8075,
+      "step": 17370
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.38527698245564e-05,
+      "loss": 0.8329,
+      "step": 17371
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3844859955145928e-05,
+      "loss": 0.9569,
+      "step": 17372
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3836952176674722e-05,
+      "loss": 0.8042,
+      "step": 17373
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3829046489334618e-05,
+      "loss": 0.8678,
+      "step": 17374
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3821142893317552e-05,
+      "loss": 0.8726,
+      "step": 17375
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3813241388815312e-05,
+      "loss": 0.7699,
+      "step": 17376
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3805341976019658e-05,
+      "loss": 0.8181,
+      "step": 17377
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3797444655122339e-05,
+      "loss": 0.8519,
+      "step": 17378
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3789549426314996e-05,
+      "loss": 0.7277,
+      "step": 17379
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3781656289789213e-05,
+      "loss": 0.7926,
+      "step": 17380
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3773765245736614e-05,
+      "loss": 0.8156,
+      "step": 17381
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3765876294348678e-05,
+      "loss": 0.7543,
+      "step": 17382
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3757989435816865e-05,
+      "loss": 0.7486,
+      "step": 17383
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3750104670332564e-05,
+      "loss": 0.8055,
+      "step": 17384
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3742221998087168e-05,
+      "loss": 0.7999,
+      "step": 17385
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.373434141927199e-05,
+      "loss": 0.962,
+      "step": 17386
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3726462934078243e-05,
+      "loss": 1.0101,
+      "step": 17387
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3718586542697199e-05,
+      "loss": 0.8469,
+      "step": 17388
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3710712245319956e-05,
+      "loss": 0.9403,
+      "step": 17389
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.370284004213762e-05,
+      "loss": 0.8167,
+      "step": 17390
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3694969933341295e-05,
+      "loss": 0.808,
+      "step": 17391
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3687101919121947e-05,
+      "loss": 0.8764,
+      "step": 17392
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3679235999670504e-05,
+      "loss": 0.8435,
+      "step": 17393
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3671372175177932e-05,
+      "loss": 0.7637,
+      "step": 17394
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3663510445835048e-05,
+      "loss": 0.8468,
+      "step": 17395
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3655650811832632e-05,
+      "loss": 0.8043,
+      "step": 17396
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3647793273361465e-05,
+      "loss": 0.7958,
+      "step": 17397
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3639937830612249e-05,
+      "loss": 0.9161,
+      "step": 17398
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.363208448377561e-05,
+      "loss": 0.9705,
+      "step": 17399
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3624233233042128e-05,
+      "loss": 0.8614,
+      "step": 17400
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3616384078602396e-05,
+      "loss": 0.7822,
+      "step": 17401
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3608537020646872e-05,
+      "loss": 0.7313,
+      "step": 17402
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3600692059366027e-05,
+      "loss": 0.7848,
+      "step": 17403
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.359284919495024e-05,
+      "loss": 0.759,
+      "step": 17404
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3585008427589818e-05,
+      "loss": 0.8317,
+      "step": 17405
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3577169757475117e-05,
+      "loss": 0.8519,
+      "step": 17406
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.356933318479635e-05,
+      "loss": 0.7892,
+      "step": 17407
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3561498709743703e-05,
+      "loss": 0.8591,
+      "step": 17408
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3553666332507287e-05,
+      "loss": 0.8835,
+      "step": 17409
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.354583605327725e-05,
+      "loss": 0.5851,
+      "step": 17410
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3538007872243586e-05,
+      "loss": 0.7805,
+      "step": 17411
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.3530181789596274e-05,
+      "loss": 0.9886,
+      "step": 17412
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3522357805525287e-05,
+      "loss": 0.7068,
+      "step": 17413
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3514535920220495e-05,
+      "loss": 0.8418,
+      "step": 17414
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3506716133871688e-05,
+      "loss": 1.2573,
+      "step": 17415
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3498898446668717e-05,
+      "loss": 1.0368,
+      "step": 17416
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3491082858801295e-05,
+      "loss": 0.8299,
+      "step": 17417
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.348326937045905e-05,
+      "loss": 0.9146,
+      "step": 17418
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3475457981831663e-05,
+      "loss": 0.8085,
+      "step": 17419
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3467648693108703e-05,
+      "loss": 0.8691,
+      "step": 17420
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3459841504479664e-05,
+      "loss": 0.8845,
+      "step": 17421
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3452036416134085e-05,
+      "loss": 0.8582,
+      "step": 17422
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3444233428261343e-05,
+      "loss": 0.7606,
+      "step": 17423
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3436432541050836e-05,
+      "loss": 0.7878,
+      "step": 17424
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3428633754691844e-05,
+      "loss": 0.7312,
+      "step": 17425
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3420837069373704e-05,
+      "loss": 0.8882,
+      "step": 17426
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.34130424852856e-05,
+      "loss": 0.8914,
+      "step": 17427
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3405250002616686e-05,
+      "loss": 0.8081,
+      "step": 17428
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.339745962155613e-05,
+      "loss": 0.948,
+      "step": 17429
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3389671342292975e-05,
+      "loss": 0.8853,
+      "step": 17430
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3381885165016216e-05,
+      "loss": 0.8426,
+      "step": 17431
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.337410108991488e-05,
+      "loss": 0.86,
+      "step": 17432
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3366319117177828e-05,
+      "loss": 0.7687,
+      "step": 17433
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3358539246993907e-05,
+      "loss": 0.7431,
+      "step": 17434
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3350761479551999e-05,
+      "loss": 0.8931,
+      "step": 17435
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3342985815040821e-05,
+      "loss": 0.8727,
+      "step": 17436
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3335212253649076e-05,
+      "loss": 1.3093,
+      "step": 17437
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.332744079556546e-05,
+      "loss": 1.1409,
+      "step": 17438
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3319671440978566e-05,
+      "loss": 0.8578,
+      "step": 17439
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.331190419007694e-05,
+      "loss": 0.7505,
+      "step": 17440
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3304139043049091e-05,
+      "loss": 0.8251,
+      "step": 17441
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3296376000083488e-05,
+      "loss": 0.745,
+      "step": 17442
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3288615061368548e-05,
+      "loss": 0.7609,
+      "step": 17443
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3280856227092576e-05,
+      "loss": 1.2688,
+      "step": 17444
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3273099497443931e-05,
+      "loss": 0.9952,
+      "step": 17445
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.326534487261084e-05,
+      "loss": 0.8668,
+      "step": 17446
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.32575923527815e-05,
+      "loss": 0.8306,
+      "step": 17447
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.324984193814407e-05,
+      "loss": 0.7564,
+      "step": 17448
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3242093628886643e-05,
+      "loss": 0.8083,
+      "step": 17449
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3234347425197236e-05,
+      "loss": 0.805,
+      "step": 17450
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.322660332726391e-05,
+      "loss": 0.8341,
+      "step": 17451
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3218861335274568e-05,
+      "loss": 0.8262,
+      "step": 17452
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3211121449417085e-05,
+      "loss": 1.0411,
+      "step": 17453
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3203383669879355e-05,
+      "loss": 1.0569,
+      "step": 17454
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3195647996849148e-05,
+      "loss": 0.8343,
+      "step": 17455
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3187914430514192e-05,
+      "loss": 0.7025,
+      "step": 17456
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3180182971062171e-05,
+      "loss": 0.7486,
+      "step": 17457
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3172453618680758e-05,
+      "loss": 0.9169,
+      "step": 17458
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3164726373557524e-05,
+      "loss": 0.8942,
+      "step": 17459
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3157001235879984e-05,
+      "loss": 0.6487,
+      "step": 17460
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.314927820583568e-05,
+      "loss": 0.7942,
+      "step": 17461
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3141557283611983e-05,
+      "loss": 0.8155,
+      "step": 17462
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3133838469396276e-05,
+      "loss": 0.6352,
+      "step": 17463
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3126121763375942e-05,
+      "loss": 0.8588,
+      "step": 17464
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3118407165738222e-05,
+      "loss": 1.2127,
+      "step": 17465
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3110694676670344e-05,
+      "loss": 1.2216,
+      "step": 17466
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3102984296359521e-05,
+      "loss": 0.7844,
+      "step": 17467
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3095276024992853e-05,
+      "loss": 0.7939,
+      "step": 17468
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3087569862757399e-05,
+      "loss": 0.9037,
+      "step": 17469
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.307986580984022e-05,
+      "loss": 0.7561,
+      "step": 17470
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3072163866428278e-05,
+      "loss": 0.6586,
+      "step": 17471
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.306446403270849e-05,
+      "loss": 0.808,
+      "step": 17472
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3056766308867707e-05,
+      "loss": 0.9033,
+      "step": 17473
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3049070695092792e-05,
+      "loss": 0.8893,
+      "step": 17474
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3041377191570491e-05,
+      "loss": 0.8036,
+      "step": 17475
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3033685798487516e-05,
+      "loss": 0.826,
+      "step": 17476
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3025996516030548e-05,
+      "loss": 0.7728,
+      "step": 17477
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.301830934438616e-05,
+      "loss": 0.7827,
+      "step": 17478
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.3010624283740968e-05,
+      "loss": 0.8677,
+      "step": 17479
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.300294133428146e-05,
+      "loss": 0.9647,
+      "step": 17480
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2995260496194106e-05,
+      "loss": 0.9299,
+      "step": 17481
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2987581769665292e-05,
+      "loss": 0.9371,
+      "step": 17482
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.297990515488141e-05,
+      "loss": 0.8859,
+      "step": 17483
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2972230652028749e-05,
+      "loss": 0.6916,
+      "step": 17484
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2964558261293558e-05,
+      "loss": 0.9919,
+      "step": 17485
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2956887982862065e-05,
+      "loss": 0.825,
+      "step": 17486
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2949219816920411e-05,
+      "loss": 0.8583,
+      "step": 17487
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.294155376365468e-05,
+      "loss": 0.6525,
+      "step": 17488
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2933889823250966e-05,
+      "loss": 0.7138,
+      "step": 17489
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2926227995895234e-05,
+      "loss": 0.8486,
+      "step": 17490
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2918568281773446e-05,
+      "loss": 0.7211,
+      "step": 17491
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2910910681071498e-05,
+      "loss": 0.8432,
+      "step": 17492
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2903255193975228e-05,
+      "loss": 0.7806,
+      "step": 17493
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2895601820670422e-05,
+      "loss": 0.7888,
+      "step": 17494
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2887950561342853e-05,
+      "loss": 0.7599,
+      "step": 17495
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2880301416178197e-05,
+      "loss": 0.7358,
+      "step": 17496
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2872654385362093e-05,
+      "loss": 0.715,
+      "step": 17497
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2865009469080113e-05,
+      "loss": 0.8395,
+      "step": 17498
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2857366667517835e-05,
+      "loss": 0.8031,
+      "step": 17499
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2849725980860716e-05,
+      "loss": 0.6294,
+      "step": 17500
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.284208740929419e-05,
+      "loss": 0.8302,
+      "step": 17501
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2834450953003674e-05,
+      "loss": 0.8422,
+      "step": 17502
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2826816612174475e-05,
+      "loss": 0.6236,
+      "step": 17503
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2819184386991844e-05,
+      "loss": 1.1519,
+      "step": 17504
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2811554277641102e-05,
+      "loss": 1.1558,
+      "step": 17505
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2803926284307343e-05,
+      "loss": 0.9584,
+      "step": 17506
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2796300407175698e-05,
+      "loss": 0.8007,
+      "step": 17507
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2788676646431298e-05,
+      "loss": 0.8275,
+      "step": 17508
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2781055002259124e-05,
+      "loss": 1.2764,
+      "step": 17509
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2773435474844142e-05,
+      "loss": 0.9304,
+      "step": 17510
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2765818064371316e-05,
+      "loss": 0.8007,
+      "step": 17511
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2758202771025496e-05,
+      "loss": 0.802,
+      "step": 17512
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2750589594991502e-05,
+      "loss": 0.8102,
+      "step": 17513
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2742978536454065e-05,
+      "loss": 0.7003,
+      "step": 17514
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2735369595597968e-05,
+      "loss": 0.8609,
+      "step": 17515
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.2727762772607831e-05,
+      "loss": 0.8789,
+      "step": 17516
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2720158067668263e-05,
+      "loss": 0.8618,
+      "step": 17517
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2712555480963862e-05,
+      "loss": 0.8206,
+      "step": 17518
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.270495501267912e-05,
+      "loss": 0.8175,
+      "step": 17519
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2697356662998495e-05,
+      "loss": 0.8858,
+      "step": 17520
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2689760432106401e-05,
+      "loss": 0.8685,
+      "step": 17521
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2682166320187183e-05,
+      "loss": 1.2492,
+      "step": 17522
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2674574327425125e-05,
+      "loss": 1.0901,
+      "step": 17523
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2666984454004537e-05,
+      "loss": 0.7567,
+      "step": 17524
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2659396700109582e-05,
+      "loss": 0.8215,
+      "step": 17525
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2651811065924413e-05,
+      "loss": 0.8318,
+      "step": 17526
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2644227551633148e-05,
+      "loss": 0.8605,
+      "step": 17527
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2636646157419817e-05,
+      "loss": 0.785,
+      "step": 17528
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2629066883468431e-05,
+      "loss": 0.777,
+      "step": 17529
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.262148972996291e-05,
+      "loss": 0.8491,
+      "step": 17530
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2613914697087182e-05,
+      "loss": 0.8176,
+      "step": 17531
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2606341785025067e-05,
+      "loss": 0.8371,
+      "step": 17532
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2598770993960341e-05,
+      "loss": 0.6605,
+      "step": 17533
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2591202324076779e-05,
+      "loss": 0.7598,
+      "step": 17534
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2583635775558078e-05,
+      "loss": 0.6607,
+      "step": 17535
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2576071348587804e-05,
+      "loss": 0.9151,
+      "step": 17536
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2568509043349598e-05,
+      "loss": 0.684,
+      "step": 17537
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2560948860026988e-05,
+      "loss": 0.7451,
+      "step": 17538
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2553390798803422e-05,
+      "loss": 0.9452,
+      "step": 17539
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2545834859862382e-05,
+      "loss": 0.7661,
+      "step": 17540
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2538281043387223e-05,
+      "loss": 0.8824,
+      "step": 17541
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2530729349561244e-05,
+      "loss": 0.8333,
+      "step": 17542
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2523179778567762e-05,
+      "loss": 0.8649,
+      "step": 17543
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2515632330589988e-05,
+      "loss": 0.754,
+      "step": 17544
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2508087005811075e-05,
+      "loss": 0.9703,
+      "step": 17545
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2500543804414177e-05,
+      "loss": 0.7581,
+      "step": 17546
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2493002726582347e-05,
+      "loss": 0.8522,
+      "step": 17547
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2485463772498606e-05,
+      "loss": 0.7086,
+      "step": 17548
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2477926942345886e-05,
+      "loss": 0.768,
+      "step": 17549
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2470392236307193e-05,
+      "loss": 0.91,
+      "step": 17550
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2462859654565273e-05,
+      "loss": 0.9666,
+      "step": 17551
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2455329197303023e-05,
+      "loss": 0.7764,
+      "step": 17552
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2447800864703174e-05,
+      "loss": 0.9546,
+      "step": 17553
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2440274656948436e-05,
+      "loss": 0.7759,
+      "step": 17554
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2432750574221452e-05,
+      "loss": 0.9054,
+      "step": 17555
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2425228616704854e-05,
+      "loss": 0.749,
+      "step": 17556
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2417708784581184e-05,
+      "loss": 0.7899,
+      "step": 17557
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2410191078032918e-05,
+      "loss": 0.7079,
+      "step": 17558
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2402675497242566e-05,
+      "loss": 0.7039,
+      "step": 17559
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2395162042392483e-05,
+      "loss": 1.1759,
+      "step": 17560
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2387650713665011e-05,
+      "loss": 1.0925,
+      "step": 17561
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2380141511242482e-05,
+      "loss": 0.9556,
+      "step": 17562
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2372634435307118e-05,
+      "loss": 0.8872,
+      "step": 17563
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2365129486041116e-05,
+      "loss": 0.832,
+      "step": 17564
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.235762666362662e-05,
+      "loss": 0.7981,
+      "step": 17565
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2350125968245707e-05,
+      "loss": 0.8081,
+      "step": 17566
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2342627400080408e-05,
+      "loss": 0.761,
+      "step": 17567
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2335130959312747e-05,
+      "loss": 0.9227,
+      "step": 17568
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2327636646124629e-05,
+      "loss": 0.8173,
+      "step": 17569
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2320144460697947e-05,
+      "loss": 0.7987,
+      "step": 17570
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2312654403214496e-05,
+      "loss": 0.8754,
+      "step": 17571
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2305166473856122e-05,
+      "loss": 0.8412,
+      "step": 17572
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2297680672804512e-05,
+      "loss": 0.8994,
+      "step": 17573
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2290197000241343e-05,
+      "loss": 0.9289,
+      "step": 17574
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2282715456348248e-05,
+      "loss": 0.8488,
+      "step": 17575
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2275236041306814e-05,
+      "loss": 0.9185,
+      "step": 17576
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.226775875529852e-05,
+      "loss": 0.7802,
+      "step": 17577
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2260283598504885e-05,
+      "loss": 0.8447,
+      "step": 17578
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2252810571107321e-05,
+      "loss": 0.7751,
+      "step": 17579
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2245339673287148e-05,
+      "loss": 0.9682,
+      "step": 17580
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2237870905225723e-05,
+      "loss": 0.8527,
+      "step": 17581
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2230404267104301e-05,
+      "loss": 0.7961,
+      "step": 17582
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2222939759104057e-05,
+      "loss": 0.8084,
+      "step": 17583
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2215477381406214e-05,
+      "loss": 0.7305,
+      "step": 17584
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.220801713419185e-05,
+      "loss": 0.885,
+      "step": 17585
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2200559017642022e-05,
+      "loss": 0.8697,
+      "step": 17586
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2193103031937703e-05,
+      "loss": 0.843,
+      "step": 17587
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2185649177259906e-05,
+      "loss": 0.7798,
+      "step": 17588
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2178197453789487e-05,
+      "loss": 0.873,
+      "step": 17589
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2170747861707299e-05,
+      "loss": 0.7671,
+      "step": 17590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2163300401194178e-05,
+      "loss": 0.7817,
+      "step": 17591
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2155855072430833e-05,
+      "loss": 0.7965,
+      "step": 17592
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2148411875597943e-05,
+      "loss": 0.7683,
+      "step": 17593
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2140970810876218e-05,
+      "loss": 0.9446,
+      "step": 17594
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2133531878446192e-05,
+      "loss": 0.7605,
+      "step": 17595
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2126095078488397e-05,
+      "loss": 1.0682,
+      "step": 17596
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2118660411183357e-05,
+      "loss": 0.9488,
+      "step": 17597
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2111227876711483e-05,
+      "loss": 0.8295,
+      "step": 17598
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2103797475253154e-05,
+      "loss": 0.8489,
+      "step": 17599
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2096369206988733e-05,
+      "loss": 0.9387,
+      "step": 17600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.208894307209849e-05,
+      "loss": 0.8534,
+      "step": 17601
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2081519070762636e-05,
+      "loss": 0.7793,
+      "step": 17602
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2074097203161339e-05,
+      "loss": 0.8061,
+      "step": 17603
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2066677469474763e-05,
+      "loss": 0.7374,
+      "step": 17604
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2059259869882955e-05,
+      "loss": 0.7181,
+      "step": 17605
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2051844404565927e-05,
+      "loss": 1.0012,
+      "step": 17606
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2044431073703677e-05,
+      "loss": 0.8324,
+      "step": 17607
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2037019877476108e-05,
+      "loss": 0.8364,
+      "step": 17608
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2029610816063086e-05,
+      "loss": 0.8674,
+      "step": 17609
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2022203889644423e-05,
+      "loss": 0.9126,
+      "step": 17610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2014799098399886e-05,
+      "loss": 1.2723,
+      "step": 17611
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2007396442509155e-05,
+      "loss": 1.1435,
+      "step": 17612
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.199999592215194e-05,
+      "loss": 0.6553,
+      "step": 17613
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.199259753750782e-05,
+      "loss": 0.837,
+      "step": 17614
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1985201288756331e-05,
+      "loss": 0.8829,
+      "step": 17615
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1977807176077028e-05,
+      "loss": 1.2919,
+      "step": 17616
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.197041519964932e-05,
+      "loss": 1.0555,
+      "step": 17617
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1963025359652602e-05,
+      "loss": 0.8966,
+      "step": 17618
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1955637656266261e-05,
+      "loss": 0.9458,
+      "step": 17619
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1948252089669565e-05,
+      "loss": 0.8782,
+      "step": 17620
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.194086866004176e-05,
+      "loss": 0.8053,
+      "step": 17621
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1933487367562024e-05,
+      "loss": 0.7602,
+      "step": 17622
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1926108212409549e-05,
+      "loss": 0.8868,
+      "step": 17623
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1918731194763356e-05,
+      "loss": 0.862,
+      "step": 17624
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1911356314802524e-05,
+      "loss": 0.8384,
+      "step": 17625
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1903983572706023e-05,
+      "loss": 1.1944,
+      "step": 17626
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1896612968652797e-05,
+      "loss": 0.9937,
+      "step": 17627
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1889244502821695e-05,
+      "loss": 0.8284,
+      "step": 17628
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.188187817539158e-05,
+      "loss": 0.7117,
+      "step": 17629
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1874513986541225e-05,
+      "loss": 0.8608,
+      "step": 17630
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1867151936449328e-05,
+      "loss": 0.8644,
+      "step": 17631
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1859792025294602e-05,
+      "loss": 0.8521,
+      "step": 17632
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1852434253255651e-05,
+      "loss": 0.7858,
+      "step": 17633
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1845078620511007e-05,
+      "loss": 0.7231,
+      "step": 17634
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1837725127239252e-05,
+      "loss": 1.3519,
+      "step": 17635
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1830373773618818e-05,
+      "loss": 1.1346,
+      "step": 17636
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1823024559828121e-05,
+      "loss": 0.8392,
+      "step": 17637
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1815677486045518e-05,
+      "loss": 0.6852,
+      "step": 17638
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1808332552449331e-05,
+      "loss": 0.8514,
+      "step": 17639
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1800989759217785e-05,
+      "loss": 0.7903,
+      "step": 17640
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1793649106529125e-05,
+      "loss": 0.8365,
+      "step": 17641
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1786310594561501e-05,
+      "loss": 0.8036,
+      "step": 17642
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.177897422349299e-05,
+      "loss": 0.7545,
+      "step": 17643
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1771639993501648e-05,
+      "loss": 1.1529,
+      "step": 17644
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1764307904765493e-05,
+      "loss": 0.996,
+      "step": 17645
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1756977957462457e-05,
+      "loss": 0.8458,
+      "step": 17646
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.174965015177042e-05,
+      "loss": 1.2428,
+      "step": 17647
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.174232448786725e-05,
+      "loss": 1.0217,
+      "step": 17648
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1735000965930742e-05,
+      "loss": 0.9398,
+      "step": 17649
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1727679586138585e-05,
+      "loss": 0.7216,
+      "step": 17650
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1720360348668524e-05,
+      "loss": 0.7523,
+      "step": 17651
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1713043253698186e-05,
+      "loss": 0.7807,
+      "step": 17652
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1705728301405084e-05,
+      "loss": 0.8771,
+      "step": 17653
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.169841549196683e-05,
+      "loss": 0.8915,
+      "step": 17654
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1691104825560862e-05,
+      "loss": 0.8342,
+      "step": 17655
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1683796302364603e-05,
+      "loss": 0.6499,
+      "step": 17656
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1676489922555444e-05,
+      "loss": 0.8728,
+      "step": 17657
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1669185686310713e-05,
+      "loss": 0.8376,
+      "step": 17658
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1661883593807665e-05,
+      "loss": 0.8008,
+      "step": 17659
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1654583645223494e-05,
+      "loss": 0.8675,
+      "step": 17660
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1647285840735412e-05,
+      "loss": 0.8182,
+      "step": 17661
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1639990180520522e-05,
+      "loss": 1.2143,
+      "step": 17662
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1632696664755848e-05,
+      "loss": 0.9847,
+      "step": 17663
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1625405293618453e-05,
+      "loss": 0.7644,
+      "step": 17664
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1618116067285267e-05,
+      "loss": 0.8471,
+      "step": 17665
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1610828985933187e-05,
+      "loss": 0.7803,
+      "step": 17666
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1603544049739112e-05,
+      "loss": 0.8646,
+      "step": 17667
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1596261258879792e-05,
+      "loss": 0.8486,
+      "step": 17668
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1588980613531975e-05,
+      "loss": 0.7332,
+      "step": 17669
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1581702113872394e-05,
+      "loss": 0.8257,
+      "step": 17670
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1574425760077679e-05,
+      "loss": 0.8668,
+      "step": 17671
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1567151552324407e-05,
+      "loss": 0.8772,
+      "step": 17672
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.155987949078915e-05,
+      "loss": 0.7937,
+      "step": 17673
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1552609575648388e-05,
+      "loss": 0.8063,
+      "step": 17674
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1545341807078535e-05,
+      "loss": 0.8975,
+      "step": 17675
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1538076185255985e-05,
+      "loss": 0.8278,
+      "step": 17676
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1530812710357098e-05,
+      "loss": 0.7557,
+      "step": 17677
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.152355138255814e-05,
+      "loss": 0.7928,
+      "step": 17678
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1516292202035307e-05,
+      "loss": 0.7614,
+      "step": 17679
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1509035168964821e-05,
+      "loss": 0.8068,
+      "step": 17680
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1501780283522789e-05,
+      "loss": 0.7269,
+      "step": 17681
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.149452754588528e-05,
+      "loss": 0.7369,
+      "step": 17682
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1487276956228322e-05,
+      "loss": 0.7728,
+      "step": 17683
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.148002851472788e-05,
+      "loss": 0.9031,
+      "step": 17684
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1472782221559842e-05,
+      "loss": 0.8838,
+      "step": 17685
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1465538076900117e-05,
+      "loss": 0.8808,
+      "step": 17686
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1458296080924492e-05,
+      "loss": 0.8042,
+      "step": 17687
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1451056233808711e-05,
+      "loss": 0.8478,
+      "step": 17688
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1443818535728523e-05,
+      "loss": 1.2162,
+      "step": 17689
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1436582986859568e-05,
+      "loss": 1.0635,
+      "step": 17690
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1429349587377413e-05,
+      "loss": 0.7972,
+      "step": 17691
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1422118337457667e-05,
+      "loss": 0.8212,
+      "step": 17692
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1414889237275783e-05,
+      "loss": 0.817,
+      "step": 17693
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1407662287007236e-05,
+      "loss": 0.7049,
+      "step": 17694
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1400437486827387e-05,
+      "loss": 0.9045,
+      "step": 17695
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1393214836911615e-05,
+      "loss": 0.6512,
+      "step": 17696
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1385994337435191e-05,
+      "loss": 0.7381,
+      "step": 17697
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1378775988573364e-05,
+      "loss": 0.7866,
+      "step": 17698
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1371559790501307e-05,
+      "loss": 0.886,
+      "step": 17699
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1364345743394155e-05,
+      "loss": 0.7984,
+      "step": 17700
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.135713384742697e-05,
+      "loss": 0.6284,
+      "step": 17701
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1349924102774822e-05,
+      "loss": 0.9247,
+      "step": 17702
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1342716509612671e-05,
+      "loss": 0.805,
+      "step": 17703
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1335511068115424e-05,
+      "loss": 0.7809,
+      "step": 17704
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.132830777845798e-05,
+      "loss": 0.8375,
+      "step": 17705
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1321106640815148e-05,
+      "loss": 0.8475,
+      "step": 17706
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1313907655361666e-05,
+      "loss": 0.9927,
+      "step": 17707
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1306710822272304e-05,
+      "loss": 0.8682,
+      "step": 17708
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.12995161417217e-05,
+      "loss": 0.8831,
+      "step": 17709
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1292323613884459e-05,
+      "loss": 0.8089,
+      "step": 17710
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1285133238935153e-05,
+      "loss": 0.7143,
+      "step": 17711
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1277945017048275e-05,
+      "loss": 1.2079,
+      "step": 17712
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1270758948398263e-05,
+      "loss": 1.1247,
+      "step": 17713
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1263575033159568e-05,
+      "loss": 0.7867,
+      "step": 17714
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1256393271506505e-05,
+      "loss": 0.6693,
+      "step": 17715
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1249213663613378e-05,
+      "loss": 0.7427,
+      "step": 17716
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1242036209654416e-05,
+      "loss": 0.8347,
+      "step": 17717
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1234860909803857e-05,
+      "loss": 0.9278,
+      "step": 17718
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1227687764235806e-05,
+      "loss": 1.2016,
+      "step": 17719
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1220516773124334e-05,
+      "loss": 0.9753,
+      "step": 17720
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1213347936643538e-05,
+      "loss": 0.8355,
+      "step": 17721
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1206181254967352e-05,
+      "loss": 0.6993,
+      "step": 17722
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1199016728269707e-05,
+      "loss": 1.3293,
+      "step": 17723
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1191854356724518e-05,
+      "loss": 1.0444,
+      "step": 17724
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1184694140505592e-05,
+      "loss": 0.7466,
+      "step": 17725
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1177536079786698e-05,
+      "loss": 0.7403,
+      "step": 17726
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1170380174741556e-05,
+      "loss": 0.7089,
+      "step": 17727
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.116322642554386e-05,
+      "loss": 0.9015,
+      "step": 17728
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.115607483236718e-05,
+      "loss": 0.803,
+      "step": 17729
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1148925395385135e-05,
+      "loss": 0.7962,
+      "step": 17730
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1141778114771206e-05,
+      "loss": 0.9478,
+      "step": 17731
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1134632990698879e-05,
+      "loss": 0.9671,
+      "step": 17732
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1127490023341513e-05,
+      "loss": 0.847,
+      "step": 17733
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1120349212872515e-05,
+      "loss": 1.2826,
+      "step": 17734
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1113210559465182e-05,
+      "loss": 0.9965,
+      "step": 17735
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1106074063292727e-05,
+      "loss": 0.8403,
+      "step": 17736
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1098939724528401e-05,
+      "loss": 0.9915,
+      "step": 17737
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1091807543345322e-05,
+      "loss": 0.9031,
+      "step": 17738
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1084677519916564e-05,
+      "loss": 0.942,
+      "step": 17739
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.107754965441522e-05,
+      "loss": 0.9412,
+      "step": 17740
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1070423947014263e-05,
+      "loss": 0.7272,
+      "step": 17741
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.106330039788659e-05,
+      "loss": 0.7614,
+      "step": 17742
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1056179007205125e-05,
+      "loss": 0.8351,
+      "step": 17743
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.10490597751427e-05,
+      "loss": 0.7148,
+      "step": 17744
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1041942701872054e-05,
+      "loss": 0.7972,
+      "step": 17745
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1034827787565971e-05,
+      "loss": 0.7022,
+      "step": 17746
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1027715032397102e-05,
+      "loss": 0.7228,
+      "step": 17747
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1020604436538052e-05,
+      "loss": 0.8159,
+      "step": 17748
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1013496000161417e-05,
+      "loss": 0.8987,
+      "step": 17749
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1006389723439714e-05,
+      "loss": 0.8347,
+      "step": 17750
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0999285606545395e-05,
+      "loss": 0.8619,
+      "step": 17751
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0992183649650867e-05,
+      "loss": 0.8737,
+      "step": 17752
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0985083852928525e-05,
+      "loss": 0.7709,
+      "step": 17753
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.097798621655064e-05,
+      "loss": 0.6433,
+      "step": 17754
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0970890740689488e-05,
+      "loss": 0.7997,
+      "step": 17755
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0963797425517276e-05,
+      "loss": 1.3116,
+      "step": 17756
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.095670627120614e-05,
+      "loss": 1.3748,
+      "step": 17757
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0949617277928158e-05,
+      "loss": 1.0143,
+      "step": 17758
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0942530445855425e-05,
+      "loss": 0.7869,
+      "step": 17759
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0935445775159925e-05,
+      "loss": 0.7243,
+      "step": 17760
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0928363266013552e-05,
+      "loss": 0.9355,
+      "step": 17761
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.092128291858825e-05,
+      "loss": 0.8571,
+      "step": 17762
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0914204733055844e-05,
+      "loss": 0.8763,
+      "step": 17763
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0907128709588076e-05,
+      "loss": 0.8234,
+      "step": 17764
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0900054848356733e-05,
+      "loss": 0.7951,
+      "step": 17765
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0892983149533476e-05,
+      "loss": 0.8962,
+      "step": 17766
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0885913613289922e-05,
+      "loss": 0.8485,
+      "step": 17767
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0878846239797623e-05,
+      "loss": 0.8778,
+      "step": 17768
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0871781029228157e-05,
+      "loss": 0.754,
+      "step": 17769
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0864717981752947e-05,
+      "loss": 0.8565,
+      "step": 17770
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0857657097543438e-05,
+      "loss": 0.8892,
+      "step": 17771
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.085059837677097e-05,
+      "loss": 0.8434,
+      "step": 17772
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0843541819606873e-05,
+      "loss": 0.8853,
+      "step": 17773
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0836487426222364e-05,
+      "loss": 0.7492,
+      "step": 17774
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0829435196788717e-05,
+      "loss": 0.9465,
+      "step": 17775
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0822385131477042e-05,
+      "loss": 0.7917,
+      "step": 17776
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0815337230458433e-05,
+      "loss": 1.0067,
+      "step": 17777
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0808291493903976e-05,
+      "loss": 0.9577,
+      "step": 17778
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0801247921984648e-05,
+      "loss": 0.8775,
+      "step": 17779
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0794206514871363e-05,
+      "loss": 0.9893,
+      "step": 17780
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0787167272735077e-05,
+      "loss": 0.8218,
+      "step": 17781
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0780130195746574e-05,
+      "loss": 0.8196,
+      "step": 17782
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0773095284076673e-05,
+      "loss": 0.7809,
+      "step": 17783
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.076606253789607e-05,
+      "loss": 0.7802,
+      "step": 17784
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0759031957375499e-05,
+      "loss": 0.9099,
+      "step": 17785
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0752003542685528e-05,
+      "loss": 0.7463,
+      "step": 17786
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0744977293996783e-05,
+      "loss": 0.7388,
+      "step": 17787
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0737953211479767e-05,
+      "loss": 0.6609,
+      "step": 17788
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0730931295304946e-05,
+      "loss": 0.7349,
+      "step": 17789
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0723911545642728e-05,
+      "loss": 0.9297,
+      "step": 17790
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.071689396266351e-05,
+      "loss": 0.7291,
+      "step": 17791
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0709878546537588e-05,
+      "loss": 0.7326,
+      "step": 17792
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0702865297435216e-05,
+      "loss": 0.8722,
+      "step": 17793
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0695854215526613e-05,
+      "loss": 0.7373,
+      "step": 17794
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0688845300981943e-05,
+      "loss": 0.6795,
+      "step": 17795
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0681838553971268e-05,
+      "loss": 1.1842,
+      "step": 17796
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0674833974664688e-05,
+      "loss": 0.9833,
+      "step": 17797
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0667831563232178e-05,
+      "loss": 0.835,
+      "step": 17798
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0660831319843689e-05,
+      "loss": 0.7029,
+      "step": 17799
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0653833244669109e-05,
+      "loss": 0.7966,
+      "step": 17800
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0646837337878269e-05,
+      "loss": 0.7788,
+      "step": 17801
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0639843599640941e-05,
+      "loss": 0.9196,
+      "step": 17802
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0632852030126905e-05,
+      "loss": 0.7971,
+      "step": 17803
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0625862629505812e-05,
+      "loss": 0.8508,
+      "step": 17804
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0618875397947303e-05,
+      "loss": 0.9197,
+      "step": 17805
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0611890335620934e-05,
+      "loss": 0.8242,
+      "step": 17806
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0604907442696266e-05,
+      "loss": 0.7792,
+      "step": 17807
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0597926719342744e-05,
+      "loss": 0.7914,
+      "step": 17808
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0590948165729764e-05,
+      "loss": 1.2923,
+      "step": 17809
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0583971782026748e-05,
+      "loss": 1.1169,
+      "step": 17810
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.057699756840297e-05,
+      "loss": 0.8772,
+      "step": 17811
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0570025525027682e-05,
+      "loss": 0.6929,
+      "step": 17812
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0563055652070131e-05,
+      "loss": 0.8453,
+      "step": 17813
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0556087949699466e-05,
+      "loss": 0.7807,
+      "step": 17814
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0549122418084745e-05,
+      "loss": 0.7493,
+      "step": 17815
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0542159057395063e-05,
+      "loss": 0.6941,
+      "step": 17816
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0535197867799395e-05,
+      "loss": 0.8429,
+      "step": 17817
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0528238849466675e-05,
+      "loss": 0.734,
+      "step": 17818
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0521282002565835e-05,
+      "loss": 0.9857,
+      "step": 17819
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0514327327265683e-05,
+      "loss": 0.8473,
+      "step": 17820
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0507374823734995e-05,
+      "loss": 0.7769,
+      "step": 17821
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0500424492142547e-05,
+      "loss": 1.2562,
+      "step": 17822
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0493476332656982e-05,
+      "loss": 1.0598,
+      "step": 17823
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0486530345446954e-05,
+      "loss": 1.2439,
+      "step": 17824
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0479586530680995e-05,
+      "loss": 1.0206,
+      "step": 17825
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0472644888527683e-05,
+      "loss": 0.8023,
+      "step": 17826
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.0465705419155458e-05,
+      "loss": 0.831,
+      "step": 17827
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.045876812273272e-05,
+      "loss": 0.781,
+      "step": 17828
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0451832999427901e-05,
+      "loss": 0.6301,
+      "step": 17829
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0444900049409245e-05,
+      "loss": 0.8797,
+      "step": 17830
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0437969272845016e-05,
+      "loss": 0.9241,
+      "step": 17831
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0431040669903459e-05,
+      "loss": 0.8472,
+      "step": 17832
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0424114240752702e-05,
+      "loss": 0.8326,
+      "step": 17833
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0417189985560826e-05,
+      "loss": 0.7858,
+      "step": 17834
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0410267904495929e-05,
+      "loss": 0.9886,
+      "step": 17835
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0403347997725976e-05,
+      "loss": 0.7379,
+      "step": 17836
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0396430265418888e-05,
+      "loss": 0.8161,
+      "step": 17837
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0389514707742608e-05,
+      "loss": 0.8673,
+      "step": 17838
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0382601324864937e-05,
+      "loss": 0.7375,
+      "step": 17839
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0375690116953673e-05,
+      "loss": 0.8406,
+      "step": 17840
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0368781084176526e-05,
+      "loss": 0.7785,
+      "step": 17841
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0361874226701196e-05,
+      "loss": 0.9052,
+      "step": 17842
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0354969544695315e-05,
+      "loss": 0.8587,
+      "step": 17843
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0348067038326436e-05,
+      "loss": 0.8762,
+      "step": 17844
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0341166707762085e-05,
+      "loss": 0.9066,
+      "step": 17845
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0334268553169734e-05,
+      "loss": 0.8121,
+      "step": 17846
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0327372574716776e-05,
+      "loss": 0.9346,
+      "step": 17847
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0320478772570618e-05,
+      "loss": 1.406,
+      "step": 17848
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0313587146898529e-05,
+      "loss": 1.0866,
+      "step": 17849
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0306697697867763e-05,
+      "loss": 0.9343,
+      "step": 17850
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0299810425645562e-05,
+      "loss": 0.7913,
+      "step": 17851
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.029292533039905e-05,
+      "loss": 0.9162,
+      "step": 17852
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0286042412295304e-05,
+      "loss": 0.8774,
+      "step": 17853
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0279161671501413e-05,
+      "loss": 0.8852,
+      "step": 17854
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0272283108184344e-05,
+      "loss": 0.7569,
+      "step": 17855
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.026540672251105e-05,
+      "loss": 0.9154,
+      "step": 17856
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0258532514648378e-05,
+      "loss": 0.7421,
+      "step": 17857
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0251660484763237e-05,
+      "loss": 0.9128,
+      "step": 17858
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0244790633022316e-05,
+      "loss": 0.795,
+      "step": 17859
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0237922959592417e-05,
+      "loss": 0.7717,
+      "step": 17860
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.023105746464017e-05,
+      "loss": 0.8441,
+      "step": 17861
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0224194148332223e-05,
+      "loss": 0.857,
+      "step": 17862
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0217333010835117e-05,
+      "loss": 0.794,
+      "step": 17863
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.02104740523154e-05,
+      "loss": 0.7077,
+      "step": 17864
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0203617272939525e-05,
+      "loss": 0.9014,
+      "step": 17865
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0196762672873873e-05,
+      "loss": 0.8814,
+      "step": 17866
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0189910252284851e-05,
+      "loss": 0.8323,
+      "step": 17867
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.018306001133873e-05,
+      "loss": 0.8116,
+      "step": 17868
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0176211950201765e-05,
+      "loss": 0.7846,
+      "step": 17869
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0169366069040165e-05,
+      "loss": 0.7864,
+      "step": 17870
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.016252236802009e-05,
+      "loss": 1.3248,
+      "step": 17871
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0155680847307603e-05,
+      "loss": 0.9289,
+      "step": 17872
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0148841507068751e-05,
+      "loss": 0.7342,
+      "step": 17873
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0142004347469535e-05,
+      "loss": 1.3082,
+      "step": 17874
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0135169368675857e-05,
+      "loss": 1.1964,
+      "step": 17875
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0128336570853636e-05,
+      "loss": 0.8394,
+      "step": 17876
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0121505954168697e-05,
+      "loss": 0.8374,
+      "step": 17877
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0114677518786798e-05,
+      "loss": 0.7504,
+      "step": 17878
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0107851264873646e-05,
+      "loss": 0.8026,
+      "step": 17879
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0101027192594947e-05,
+      "loss": 0.7763,
+      "step": 17880
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0094205302116311e-05,
+      "loss": 0.8866,
+      "step": 17881
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0087385593603271e-05,
+      "loss": 0.8689,
+      "step": 17882
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0080568067221374e-05,
+      "loss": 0.7809,
+      "step": 17883
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0073752723136076e-05,
+      "loss": 0.8014,
+      "step": 17884
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0066939561512745e-05,
+      "loss": 0.6574,
+      "step": 17885
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.006012858251677e-05,
+      "loss": 0.7964,
+      "step": 17886
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0053319786313442e-05,
+      "loss": 0.7957,
+      "step": 17887
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0046513173068006e-05,
+      "loss": 0.7988,
+      "step": 17888
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0039708742945642e-05,
+      "loss": 0.8466,
+      "step": 17889
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0032906496111505e-05,
+      "loss": 0.8924,
+      "step": 17890
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0026106432730652e-05,
+      "loss": 0.8056,
+      "step": 17891
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0019308552968166e-05,
+      "loss": 0.7424,
+      "step": 17892
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0012512856988999e-05,
+      "loss": 0.8338,
+      "step": 17893
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.0005719344958054e-05,
+      "loss": 0.8146,
+      "step": 17894
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.998928017040265e-06,
+      "loss": 0.7248,
+      "step": 17895
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.992138873400415e-06,
+      "loss": 0.7698,
+      "step": 17896
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.985351914203289e-06,
+      "loss": 0.8236,
+      "step": 17897
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.97856713961357e-06,
+      "loss": 0.7752,
+      "step": 17898
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.97178454979597e-06,
+      "loss": 0.9174,
+      "step": 17899
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.965004144915068e-06,
+      "loss": 0.8317,
+      "step": 17900
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.958225925135411e-06,
+      "loss": 0.8473,
+      "step": 17901
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.951449890621556e-06,
+      "loss": 0.8754,
+      "step": 17902
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.944676041537904e-06,
+      "loss": 0.8164,
+      "step": 17903
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.937904378048846e-06,
+      "loss": 0.8124,
+      "step": 17904
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.931134900318762e-06,
+      "loss": 0.7994,
+      "step": 17905
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.92436760851192e-06,
+      "loss": 0.8067,
+      "step": 17906
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.917602502792556e-06,
+      "loss": 0.7499,
+      "step": 17907
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.910839583324882e-06,
+      "loss": 0.8469,
+      "step": 17908
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.904078850273013e-06,
+      "loss": 0.731,
+      "step": 17909
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.897320303801017e-06,
+      "loss": 0.7575,
+      "step": 17910
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.89056394407295e-06,
+      "loss": 0.9724,
+      "step": 17911
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.883809771252762e-06,
+      "loss": 0.9635,
+      "step": 17912
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.877057785504384e-06,
+      "loss": 0.7976,
+      "step": 17913
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.870307986991655e-06,
+      "loss": 1.312,
+      "step": 17914
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.863560375878445e-06,
+      "loss": 1.0778,
+      "step": 17915
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.856814952328464e-06,
+      "loss": 0.9108,
+      "step": 17916
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.85007171650545e-06,
+      "loss": 0.7265,
+      "step": 17917
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.843330668573048e-06,
+      "loss": 0.7974,
+      "step": 17918
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.83659180869485e-06,
+      "loss": 0.8069,
+      "step": 17919
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.829855137034394e-06,
+      "loss": 0.8354,
+      "step": 17920
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.823120653755213e-06,
+      "loss": 0.7249,
+      "step": 17921
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.816388359020734e-06,
+      "loss": 0.8345,
+      "step": 17922
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.809658252994325e-06,
+      "loss": 0.816,
+      "step": 17923
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.802930335839344e-06,
+      "loss": 0.8752,
+      "step": 17924
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.796204607719084e-06,
+      "loss": 0.9323,
+      "step": 17925
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.789481068796736e-06,
+      "loss": 0.8263,
+      "step": 17926
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.782759719235513e-06,
+      "loss": 0.7874,
+      "step": 17927
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.77604055919854e-06,
+      "loss": 0.933,
+      "step": 17928
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.769323588848866e-06,
+      "loss": 0.8245,
+      "step": 17929
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.762608808349494e-06,
+      "loss": 0.7669,
+      "step": 17930
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.755896217863437e-06,
+      "loss": 0.7637,
+      "step": 17931
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.749185817553563e-06,
+      "loss": 0.7642,
+      "step": 17932
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.742477607582756e-06,
+      "loss": 0.8725,
+      "step": 17933
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.735771588113818e-06,
+      "loss": 0.6645,
+      "step": 17934
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.729067759309485e-06,
+      "loss": 1.3379,
+      "step": 17935
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.722366121332449e-06,
+      "loss": 1.5995,
+      "step": 17936
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.71566667434538e-06,
+      "loss": 0.9925,
+      "step": 17937
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.708969418510861e-06,
+      "loss": 0.7853,
+      "step": 17938
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.702274353991414e-06,
+      "loss": 0.8118,
+      "step": 17939
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.695581480949567e-06,
+      "loss": 0.7311,
+      "step": 17940
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.688890799547712e-06,
+      "loss": 0.8133,
+      "step": 17941
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.68220230994823e-06,
+      "loss": 0.916,
+      "step": 17942
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.675516012313478e-06,
+      "loss": 0.9186,
+      "step": 17943
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.668831906805709e-06,
+      "loss": 0.7688,
+      "step": 17944
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.662149993587133e-06,
+      "loss": 0.937,
+      "step": 17945
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.655470272819945e-06,
+      "loss": 0.8717,
+      "step": 17946
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.648792744666223e-06,
+      "loss": 0.8484,
+      "step": 17947
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.642117409288031e-06,
+      "loss": 1.2268,
+      "step": 17948
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.635444266847416e-06,
+      "loss": 0.9771,
+      "step": 17949
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.628773317506302e-06,
+      "loss": 0.8084,
+      "step": 17950
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.62210456142656e-06,
+      "loss": 1.2926,
+      "step": 17951
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.615437998770105e-06,
+      "loss": 1.1283,
+      "step": 17952
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.608773629698687e-06,
+      "loss": 0.7727,
+      "step": 17953
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.602111454374052e-06,
+      "loss": 0.8353,
+      "step": 17954
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.595451472957872e-06,
+      "loss": 0.7899,
+      "step": 17955
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.588793685611818e-06,
+      "loss": 0.86,
+      "step": 17956
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.582138092497461e-06,
+      "loss": 1.2542,
+      "step": 17957
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.575484693776292e-06,
+      "loss": 1.1794,
+      "step": 17958
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.56883348960983e-06,
+      "loss": 0.7551,
+      "step": 17959
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.562184480159485e-06,
+      "loss": 0.8215,
+      "step": 17960
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.555537665586623e-06,
+      "loss": 0.672,
+      "step": 17961
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.548893046052553e-06,
+      "loss": 0.8044,
+      "step": 17962
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.54225062171853e-06,
+      "loss": 1.0239,
+      "step": 17963
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.535610392745764e-06,
+      "loss": 0.7601,
+      "step": 17964
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.52897235929543e-06,
+      "loss": 0.7957,
+      "step": 17965
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.522336521528608e-06,
+      "loss": 0.8103,
+      "step": 17966
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.515702879606348e-06,
+      "loss": 0.9263,
+      "step": 17967
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.509071433689654e-06,
+      "loss": 0.7798,
+      "step": 17968
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.502442183939475e-06,
+      "loss": 0.8291,
+      "step": 17969
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.495815130516694e-06,
+      "loss": 0.7793,
+      "step": 17970
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.489190273582116e-06,
+      "loss": 0.6959,
+      "step": 17971
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.482567613296566e-06,
+      "loss": 1.3423,
+      "step": 17972
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.475947149820752e-06,
+      "loss": 1.2141,
+      "step": 17973
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.469328883315342e-06,
+      "loss": 1.2548,
+      "step": 17974
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.462712813940977e-06,
+      "loss": 0.9827,
+      "step": 17975
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.456098941858239e-06,
+      "loss": 0.8516,
+      "step": 17976
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.449487267227586e-06,
+      "loss": 0.7552,
+      "step": 17977
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.442877790209526e-06,
+      "loss": 0.8134,
+      "step": 17978
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.436270510964451e-06,
+      "loss": 0.79,
+      "step": 17979
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.429665429652712e-06,
+      "loss": 0.8318,
+      "step": 17980
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.423062546434624e-06,
+      "loss": 0.9012,
+      "step": 17981
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.41646186147045e-06,
+      "loss": 0.8291,
+      "step": 17982
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.409863374920325e-06,
+      "loss": 0.8568,
+      "step": 17983
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.403267086944468e-06,
+      "loss": 0.8018,
+      "step": 17984
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.396672997702915e-06,
+      "loss": 0.809,
+      "step": 17985
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.39008110735573e-06,
+      "loss": 0.9043,
+      "step": 17986
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.383491416062861e-06,
+      "loss": 0.7473,
+      "step": 17987
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.37690392398427e-06,
+      "loss": 0.8038,
+      "step": 17988
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.370318631279817e-06,
+      "loss": 0.8107,
+      "step": 17989
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.36373553810933e-06,
+      "loss": 0.824,
+      "step": 17990
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.357154644632571e-06,
+      "loss": 0.8776,
+      "step": 17991
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.350575951009255e-06,
+      "loss": 0.777,
+      "step": 17992
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.343999457399033e-06,
+      "loss": 1.2333,
+      "step": 17993
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.337425163961533e-06,
+      "loss": 1.0654,
+      "step": 17994
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.330853070856304e-06,
+      "loss": 0.7039,
+      "step": 17995
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.324283178242832e-06,
+      "loss": 1.37,
+      "step": 17996
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.317715486280587e-06,
+      "loss": 1.0597,
+      "step": 17997
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.311149995128954e-06,
+      "loss": 0.7055,
+      "step": 17998
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.30458670494726e-06,
+      "loss": 0.8614,
+      "step": 17999
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.298025615894823e-06,
+      "loss": 0.8376,
+      "step": 18000
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.291466728130849e-06,
+      "loss": 0.8619,
+      "step": 18001
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.284910041814543e-06,
+      "loss": 0.7961,
+      "step": 18002
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.278355557105e-06,
+      "loss": 0.8679,
+      "step": 18003
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.271803274161327e-06,
+      "loss": 0.862,
+      "step": 18004
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.265253193142532e-06,
+      "loss": 0.7478,
+      "step": 18005
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.258705314207572e-06,
+      "loss": 0.7922,
+      "step": 18006
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.252159637515379e-06,
+      "loss": 0.7929,
+      "step": 18007
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.245616163224801e-06,
+      "loss": 0.9349,
+      "step": 18008
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.239074891494626e-06,
+      "loss": 0.8988,
+      "step": 18009
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.232535822483646e-06,
+      "loss": 0.7086,
+      "step": 18010
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.225998956350546e-06,
+      "loss": 0.8391,
+      "step": 18011
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.219464293253943e-06,
+      "loss": 0.8804,
+      "step": 18012
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.212931833352478e-06,
+      "loss": 0.8752,
+      "step": 18013
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.206401576804668e-06,
+      "loss": 0.6846,
+      "step": 18014
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.199873523768977e-06,
+      "loss": 0.8765,
+      "step": 18015
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.193347674403884e-06,
+      "loss": 0.6938,
+      "step": 18016
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.186824028867735e-06,
+      "loss": 1.084,
+      "step": 18017
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.180302587318867e-06,
+      "loss": 1.2863,
+      "step": 18018
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.173783349915521e-06,
+      "loss": 0.7902,
+      "step": 18019
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.167266316815992e-06,
+      "loss": 0.9166,
+      "step": 18020
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.160751488178354e-06,
+      "loss": 0.6736,
+      "step": 18021
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.15423886416078e-06,
+      "loss": 0.8853,
+      "step": 18022
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.1477284449213e-06,
+      "loss": 0.7,
+      "step": 18023
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.14122023061792e-06,
+      "loss": 0.8732,
+      "step": 18024
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.134714221408603e-06,
+      "loss": 0.7463,
+      "step": 18025
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.128210417451244e-06,
+      "loss": 0.9456,
+      "step": 18026
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.121708818903684e-06,
+      "loss": 0.7838,
+      "step": 18027
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.115209425923699e-06,
+      "loss": 0.7669,
+      "step": 18028
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.108712238669049e-06,
+      "loss": 0.9295,
+      "step": 18029
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.102217257297419e-06,
+      "loss": 0.7926,
+      "step": 18030
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.095724481966394e-06,
+      "loss": 1.2686,
+      "step": 18031
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.089233912833617e-06,
+      "loss": 1.0547,
+      "step": 18032
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.082745550056559e-06,
+      "loss": 0.7416,
+      "step": 18033
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.076259393792708e-06,
+      "loss": 0.8338,
+      "step": 18034
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.069775444199503e-06,
+      "loss": 1.3185,
+      "step": 18035
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.063293701434273e-06,
+      "loss": 0.9903,
+      "step": 18036
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.056814165654304e-06,
+      "loss": 0.7928,
+      "step": 18037
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.050336837016915e-06,
+      "loss": 0.8738,
+      "step": 18038
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.043861715679259e-06,
+      "loss": 0.8239,
+      "step": 18039
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.037388801798486e-06,
+      "loss": 0.7772,
+      "step": 18040
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.030918095531726e-06,
+      "loss": 0.8532,
+      "step": 18041
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.024449597036e-06,
+      "loss": 0.7428,
+      "step": 18042
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.017983306468281e-06,
+      "loss": 0.7249,
+      "step": 18043
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.011519223985498e-06,
+      "loss": 0.7409,
+      "step": 18044
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.005057349744573e-06,
+      "loss": 0.922,
+      "step": 18045
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.998597683902298e-06,
+      "loss": 0.8745,
+      "step": 18046
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.992140226615441e-06,
+      "loss": 0.7738,
+      "step": 18047
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.985684978040742e-06,
+      "loss": 0.8651,
+      "step": 18048
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.979231938334876e-06,
+      "loss": 0.8342,
+      "step": 18049
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.972781107654404e-06,
+      "loss": 0.7663,
+      "step": 18050
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.966332486155937e-06,
+      "loss": 0.8349,
+      "step": 18051
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.95988607399596e-06,
+      "loss": 0.8567,
+      "step": 18052
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.953441871330902e-06,
+      "loss": 0.8361,
+      "step": 18053
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.946999878317198e-06,
+      "loss": 0.8135,
+      "step": 18054
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.940560095111172e-06,
+      "loss": 0.6771,
+      "step": 18055
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.934122521869105e-06,
+      "loss": 0.8921,
+      "step": 18056
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.927687158747266e-06,
+      "loss": 0.8453,
+      "step": 18057
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.921254005901813e-06,
+      "loss": 1.1776,
+      "step": 18058
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.914823063488874e-06,
+      "loss": 1.4434,
+      "step": 18059
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.908394331664515e-06,
+      "loss": 1.0424,
+      "step": 18060
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.9019678105848e-06,
+      "loss": 0.7835,
+      "step": 18061
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.895543500405668e-06,
+      "loss": 0.7461,
+      "step": 18062
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.889121401283007e-06,
+      "loss": 0.8544,
+      "step": 18063
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.88270151337276e-06,
+      "loss": 0.8564,
+      "step": 18064
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.876283836830646e-06,
+      "loss": 0.8773,
+      "step": 18065
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.869868371812451e-06,
+      "loss": 1.2811,
+      "step": 18066
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.863455118473884e-06,
+      "loss": 1.0903,
+      "step": 18067
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.857044076970589e-06,
+      "loss": 0.6752,
+      "step": 18068
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.850635247458138e-06,
+      "loss": 0.9528,
+      "step": 18069
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.8442286300921e-06,
+      "loss": 0.9147,
+      "step": 18070
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.837824225027946e-06,
+      "loss": 0.7962,
+      "step": 18071
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.8314220324211e-06,
+      "loss": 0.8786,
+      "step": 18072
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.825022052426956e-06,
+      "loss": 0.7692,
+      "step": 18073
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.818624285200828e-06,
+      "loss": 0.8845,
+      "step": 18074
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.812228730897998e-06,
+      "loss": 0.8444,
+      "step": 18075
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.805835389673645e-06,
+      "loss": 0.8683,
+      "step": 18076
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.799444261682987e-06,
+      "loss": 0.8702,
+      "step": 18077
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.793055347081102e-06,
+      "loss": 0.8608,
+      "step": 18078
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.786668646023044e-06,
+      "loss": 1.0385,
+      "step": 18079
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.780284158663832e-06,
+      "loss": 0.8997,
+      "step": 18080
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.773901885158375e-06,
+      "loss": 0.7575,
+      "step": 18081
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.767521825661618e-06,
+      "loss": 0.7019,
+      "step": 18082
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.761143980328368e-06,
+      "loss": 0.8615,
+      "step": 18083
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.754768349313435e-06,
+      "loss": 0.8447,
+      "step": 18084
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.74839493277152e-06,
+      "loss": 0.82,
+      "step": 18085
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.74202373085734e-06,
+      "loss": 0.8674,
+      "step": 18086
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.735654743725496e-06,
+      "loss": 1.1544,
+      "step": 18087
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.729287971530554e-06,
+      "loss": 1.552,
+      "step": 18088
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.722923414427065e-06,
+      "loss": 1.0134,
+      "step": 18089
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.716561072569474e-06,
+      "loss": 0.827,
+      "step": 18090
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.710200946112168e-06,
+      "loss": 0.9106,
+      "step": 18091
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.703843035209547e-06,
+      "loss": 0.8723,
+      "step": 18092
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.697487340015909e-06,
+      "loss": 0.6884,
+      "step": 18093
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.691133860685452e-06,
+      "loss": 0.8364,
+      "step": 18094
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.684782597372432e-06,
+      "loss": 0.8259,
+      "step": 18095
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.678433550230958e-06,
+      "loss": 0.8182,
+      "step": 18096
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.672086719415117e-06,
+      "loss": 0.7429,
+      "step": 18097
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.665742105078966e-06,
+      "loss": 0.9067,
+      "step": 18098
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.65939970737647e-06,
+      "loss": 0.8383,
+      "step": 18099
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.65305952646156e-06,
+      "loss": 0.7746,
+      "step": 18100
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.64672156248809e-06,
+      "loss": 0.8727,
+      "step": 18101
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.640385815609909e-06,
+      "loss": 0.6736,
+      "step": 18102
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.634052285980776e-06,
+      "loss": 0.7455,
+      "step": 18103
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.627720973754372e-06,
+      "loss": 0.9054,
+      "step": 18104
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.621391879084395e-06,
+      "loss": 0.7773,
+      "step": 18105
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.615065002124434e-06,
+      "loss": 0.7524,
+      "step": 18106
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.60874034302802e-06,
+      "loss": 0.8457,
+      "step": 18107
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.602417901948701e-06,
+      "loss": 0.792,
+      "step": 18108
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.59609767903986e-06,
+      "loss": 0.8669,
+      "step": 18109
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.58977967445489e-06,
+      "loss": 0.6689,
+      "step": 18110
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.583463888347164e-06,
+      "loss": 0.8135,
+      "step": 18111
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.577150320869953e-06,
+      "loss": 0.7455,
+      "step": 18112
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.570838972176454e-06,
+      "loss": 0.837,
+      "step": 18113
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.564529842419865e-06,
+      "loss": 0.8832,
+      "step": 18114
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.558222931753312e-06,
+      "loss": 0.7017,
+      "step": 18115
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.551918240329849e-06,
+      "loss": 0.844,
+      "step": 18116
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.545615768302462e-06,
+      "loss": 0.7563,
+      "step": 18117
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.539315515824165e-06,
+      "loss": 0.8658,
+      "step": 18118
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.533017483047822e-06,
+      "loss": 0.8648,
+      "step": 18119
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.526721670126269e-06,
+      "loss": 0.763,
+      "step": 18120
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.52042807721235e-06,
+      "loss": 0.8655,
+      "step": 18121
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.514136704458787e-06,
+      "loss": 0.8108,
+      "step": 18122
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.507847552018255e-06,
+      "loss": 0.9291,
+      "step": 18123
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.501560620043404e-06,
+      "loss": 0.8287,
+      "step": 18124
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.495275908686795e-06,
+      "loss": 0.7461,
+      "step": 18125
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.488993418100955e-06,
+      "loss": 0.7902,
+      "step": 18126
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.482713148438382e-06,
+      "loss": 0.8375,
+      "step": 18127
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.476435099851488e-06,
+      "loss": 0.8743,
+      "step": 18128
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.470159272492617e-06,
+      "loss": 0.8109,
+      "step": 18129
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.463885666514105e-06,
+      "loss": 0.8437,
+      "step": 18130
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.457614282068194e-06,
+      "loss": 0.6361,
+      "step": 18131
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.451345119307097e-06,
+      "loss": 0.9416,
+      "step": 18132
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.445078178382938e-06,
+      "loss": 0.8786,
+      "step": 18133
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.438813459447847e-06,
+      "loss": 0.9301,
+      "step": 18134
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.432550962653852e-06,
+      "loss": 0.7046,
+      "step": 18135
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.426290688152926e-06,
+      "loss": 0.6997,
+      "step": 18136
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.420032636097042e-06,
+      "loss": 0.7421,
+      "step": 18137
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.413776806638041e-06,
+      "loss": 0.8696,
+      "step": 18138
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.407523199927736e-06,
+      "loss": 0.8419,
+      "step": 18139
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.40127181611795e-06,
+      "loss": 0.8436,
+      "step": 18140
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.395022655360374e-06,
+      "loss": 0.756,
+      "step": 18141
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.388775717806641e-06,
+      "loss": 0.8422,
+      "step": 18142
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.382531003608418e-06,
+      "loss": 0.8095,
+      "step": 18143
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.37628851291723e-06,
+      "loss": 0.7152,
+      "step": 18144
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.370048245884577e-06,
+      "loss": 0.8402,
+      "step": 18145
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.363810202661915e-06,
+      "loss": 0.8828,
+      "step": 18146
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.357574383400646e-06,
+      "loss": 0.7479,
+      "step": 18147
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.351340788252093e-06,
+      "loss": 0.8941,
+      "step": 18148
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.345109417367537e-06,
+      "loss": 0.8587,
+      "step": 18149
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.338880270898241e-06,
+      "loss": 0.8503,
+      "step": 18150
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.332653348995368e-06,
+      "loss": 0.8379,
+      "step": 18151
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.326428651810026e-06,
+      "loss": 0.8429,
+      "step": 18152
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.320206179493307e-06,
+      "loss": 0.9301,
+      "step": 18153
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.3139859321962e-06,
+      "loss": 0.76,
+      "step": 18154
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.307767910069697e-06,
+      "loss": 0.682,
+      "step": 18155
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.301552113264687e-06,
+      "loss": 0.8696,
+      "step": 18156
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.295338541932041e-06,
+      "loss": 0.6125,
+      "step": 18157
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.289127196222535e-06,
+      "loss": 0.7966,
+      "step": 18158
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.282918076286939e-06,
+      "loss": 0.731,
+      "step": 18159
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.276711182275932e-06,
+      "loss": 0.7597,
+      "step": 18160
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.270506514340137e-06,
+      "loss": 1.1787,
+      "step": 18161
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.26430407263018e-06,
+      "loss": 1.1559,
+      "step": 18162
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.258103857296562e-06,
+      "loss": 1.1396,
+      "step": 18163
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.25190586848975e-06,
+      "loss": 1.0651,
+      "step": 18164
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.245710106360195e-06,
+      "loss": 0.7805,
+      "step": 18165
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.239516571058248e-06,
+      "loss": 0.8192,
+      "step": 18166
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.233325262734226e-06,
+      "loss": 0.6477,
+      "step": 18167
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.227136181538386e-06,
+      "loss": 0.9181,
+      "step": 18168
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.220949327620941e-06,
+      "loss": 0.8309,
+      "step": 18169
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.214764701132016e-06,
+      "loss": 0.8299,
+      "step": 18170
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.208582302221734e-06,
+      "loss": 0.8078,
+      "step": 18171
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.202402131040144e-06,
+      "loss": 1.305,
+      "step": 18172
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.196224187737223e-06,
+      "loss": 1.0159,
+      "step": 18173
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.190048472462897e-06,
+      "loss": 0.87,
+      "step": 18174
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.183874985367068e-06,
+      "loss": 0.8044,
+      "step": 18175
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.177703726599551e-06,
+      "loss": 0.7721,
+      "step": 18176
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.171534696310113e-06,
+      "loss": 0.7712,
+      "step": 18177
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.1653678946485e-06,
+      "loss": 0.9004,
+      "step": 18178
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.159203321764353e-06,
+      "loss": 0.7904,
+      "step": 18179
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.15304097780727e-06,
+      "loss": 0.8343,
+      "step": 18180
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.146880862926843e-06,
+      "loss": 0.8768,
+      "step": 18181
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.140722977272586e-06,
+      "loss": 0.8624,
+      "step": 18182
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.13456732099388e-06,
+      "loss": 0.7121,
+      "step": 18183
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.128413894240173e-06,
+      "loss": 0.8132,
+      "step": 18184
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.122262697160787e-06,
+      "loss": 0.8227,
+      "step": 18185
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.116113729905007e-06,
+      "loss": 0.856,
+      "step": 18186
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.109966992622087e-06,
+      "loss": 0.8459,
+      "step": 18187
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.103822485461188e-06,
+      "loss": 0.8422,
+      "step": 18188
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.097680208571423e-06,
+      "loss": 0.9508,
+      "step": 18189
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.091540162101874e-06,
+      "loss": 0.6391,
+      "step": 18190
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.085402346201565e-06,
+      "loss": 0.7121,
+      "step": 18191
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.079266761019455e-06,
+      "loss": 0.6685,
+      "step": 18192
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.073133406704426e-06,
+      "loss": 0.771,
+      "step": 18193
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.067002283405357e-06,
+      "loss": 0.7356,
+      "step": 18194
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.060873391271051e-06,
+      "loss": 0.8158,
+      "step": 18195
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.054746730450246e-06,
+      "loss": 0.8738,
+      "step": 18196
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.048622301091623e-06,
+      "loss": 0.8857,
+      "step": 18197
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.042500103343832e-06,
+      "loss": 0.8139,
+      "step": 18198
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.036380137355415e-06,
+      "loss": 0.8287,
+      "step": 18199
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.030262403274969e-06,
+      "loss": 0.8223,
+      "step": 18200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.024146901250929e-06,
+      "loss": 0.7265,
+      "step": 18201
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.018033631431699e-06,
+      "loss": 0.7038,
+      "step": 18202
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.011922593965682e-06,
+      "loss": 0.7148,
+      "step": 18203
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.005813789001171e-06,
+      "loss": 0.8282,
+      "step": 18204
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.999707216686425e-06,
+      "loss": 0.8754,
+      "step": 18205
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.993602877169625e-06,
+      "loss": 0.804,
+      "step": 18206
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.987500770598965e-06,
+      "loss": 0.9063,
+      "step": 18207
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.981400897122526e-06,
+      "loss": 0.8973,
+      "step": 18208
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.975303256888311e-06,
+      "loss": 0.7837,
+      "step": 18209
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.969207850044358e-06,
+      "loss": 0.7198,
+      "step": 18210
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.963114676738593e-06,
+      "loss": 0.8591,
+      "step": 18211
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.95702373711884e-06,
+      "loss": 0.7897,
+      "step": 18212
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.950935031332973e-06,
+      "loss": 0.9054,
+      "step": 18213
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.944848559528762e-06,
+      "loss": 0.8507,
+      "step": 18214
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.938764321853876e-06,
+      "loss": 0.7326,
+      "step": 18215
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.932682318456042e-06,
+      "loss": 0.8287,
+      "step": 18216
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.926602549482831e-06,
+      "loss": 0.8221,
+      "step": 18217
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.92052501508177e-06,
+      "loss": 0.8924,
+      "step": 18218
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.914449715400418e-06,
+      "loss": 0.6899,
+      "step": 18219
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.908376650586181e-06,
+      "loss": 0.9249,
+      "step": 18220
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.902305820786448e-06,
+      "loss": 0.7041,
+      "step": 18221
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.896237226148573e-06,
+      "loss": 0.8805,
+      "step": 18222
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.890170866819824e-06,
+      "loss": 0.8362,
+      "step": 18223
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.884106742947439e-06,
+      "loss": 0.7871,
+      "step": 18224
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.878044854678556e-06,
+      "loss": 0.7067,
+      "step": 18225
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.87198520216037e-06,
+      "loss": 0.8666,
+      "step": 18226
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.86592778553985e-06,
+      "loss": 0.7327,
+      "step": 18227
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.859872604964081e-06,
+      "loss": 0.8113,
+      "step": 18228
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.853819660579987e-06,
+      "loss": 0.7849,
+      "step": 18229
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.847768952534474e-06,
+      "loss": 0.7674,
+      "step": 18230
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.841720480974369e-06,
+      "loss": 0.8712,
+      "step": 18231
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.835674246046509e-06,
+      "loss": 0.8917,
+      "step": 18232
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.829630247897613e-06,
+      "loss": 0.9873,
+      "step": 18233
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.823588486674338e-06,
+      "loss": 0.8482,
+      "step": 18234
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.817548962523358e-06,
+      "loss": 0.8916,
+      "step": 18235
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.81151167559122e-06,
+      "loss": 0.7784,
+      "step": 18236
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.805476626024454e-06,
+      "loss": 1.0134,
+      "step": 18237
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.79944381396953e-06,
+      "loss": 0.7217,
+      "step": 18238
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.793413239572877e-06,
+      "loss": 0.6962,
+      "step": 18239
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.78738490298082e-06,
+      "loss": 0.8209,
+      "step": 18240
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.7813588043397e-06,
+      "loss": 0.8998,
+      "step": 18241
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.77533494379573e-06,
+      "loss": 0.862,
+      "step": 18242
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.769313321495108e-06,
+      "loss": 0.8055,
+      "step": 18243
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.763293937584015e-06,
+      "loss": 0.9207,
+      "step": 18244
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.757276792208511e-06,
+      "loss": 0.882,
+      "step": 18245
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.751261885514638e-06,
+      "loss": 0.8659,
+      "step": 18246
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.745249217648342e-06,
+      "loss": 0.8874,
+      "step": 18247
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.7392387887556e-06,
+      "loss": 0.7739,
+      "step": 18248
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.733230598982243e-06,
+      "loss": 0.936,
+      "step": 18249
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.727224648474096e-06,
+      "loss": 0.9167,
+      "step": 18250
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.721220937376927e-06,
+      "loss": 0.7885,
+      "step": 18251
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.715219465836454e-06,
+      "loss": 0.8045,
+      "step": 18252
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.709220233998282e-06,
+      "loss": 0.7668,
+      "step": 18253
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.703223242008073e-06,
+      "loss": 0.8708,
+      "step": 18254
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.697228490011355e-06,
+      "loss": 0.8278,
+      "step": 18255
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.691235978153555e-06,
+      "loss": 0.8433,
+      "step": 18256
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.68524570658019e-06,
+      "loss": 0.8976,
+      "step": 18257
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.679257675436591e-06,
+      "loss": 0.7037,
+      "step": 18258
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.673271884868094e-06,
+      "loss": 0.8067,
+      "step": 18259
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.667288335019985e-06,
+      "loss": 0.7786,
+      "step": 18260
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.66130702603748e-06,
+      "loss": 0.8738,
+      "step": 18261
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.655327958065727e-06,
+      "loss": 0.8483,
+      "step": 18262
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.649351131249826e-06,
+      "loss": 0.8688,
+      "step": 18263
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.643376545734871e-06,
+      "loss": 0.973,
+      "step": 18264
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.637404201665843e-06,
+      "loss": 0.8713,
+      "step": 18265
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.631434099187662e-06,
+      "loss": 0.7905,
+      "step": 18266
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.625466238445255e-06,
+      "loss": 0.7558,
+      "step": 18267
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.619500619583442e-06,
+      "loss": 0.6715,
+      "step": 18268
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.6135372427469935e-06,
+      "loss": 0.8382,
+      "step": 18269
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.607576108080694e-06,
+      "loss": 0.8538,
+      "step": 18270
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.601617215729151e-06,
+      "loss": 1.1324,
+      "step": 18271
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.595660565836993e-06,
+      "loss": 1.0948,
+      "step": 18272
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.589706158548815e-06,
+      "loss": 0.8128,
+      "step": 18273
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.583753994009113e-06,
+      "loss": 0.8334,
+      "step": 18274
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.577804072362316e-06,
+      "loss": 0.8034,
+      "step": 18275
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.571856393752885e-06,
+      "loss": 1.1809,
+      "step": 18276
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.5659109583251175e-06,
+      "loss": 1.1077,
+      "step": 18277
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.55996776622333e-06,
+      "loss": 0.8071,
+      "step": 18278
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.554026817591731e-06,
+      "loss": 0.8849,
+      "step": 18279
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.548088112574547e-06,
+      "loss": 0.8893,
+      "step": 18280
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.542151651315887e-06,
+      "loss": 0.8062,
+      "step": 18281
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.536217433959803e-06,
+      "loss": 0.7003,
+      "step": 18282
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.530285460650355e-06,
+      "loss": 0.9584,
+      "step": 18283
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.524355731531485e-06,
+      "loss": 0.7951,
+      "step": 18284
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.518428246747111e-06,
+      "loss": 0.7369,
+      "step": 18285
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.512503006441096e-06,
+      "loss": 0.7405,
+      "step": 18286
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.506580010757225e-06,
+      "loss": 0.8256,
+      "step": 18287
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.500659259839249e-06,
+      "loss": 0.7189,
+      "step": 18288
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.494740753830875e-06,
+      "loss": 0.6689,
+      "step": 18289
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.488824492875745e-06,
+      "loss": 0.9222,
+      "step": 18290
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.48291047711741e-06,
+      "loss": 1.1413,
+      "step": 18291
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.476998706699445e-06,
+      "loss": 1.1228,
+      "step": 18292
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.471089181765312e-06,
+      "loss": 1.1986,
+      "step": 18293
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.4651819024583955e-06,
+      "loss": 1.0311,
+      "step": 18294
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.459276868922116e-06,
+      "loss": 0.7829,
+      "step": 18295
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.453374081299746e-06,
+      "loss": 0.8673,
+      "step": 18296
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.447473539734573e-06,
+      "loss": 0.6429,
+      "step": 18297
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.441575244369758e-06,
+      "loss": 0.8723,
+      "step": 18298
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.435679195348522e-06,
+      "loss": 0.8787,
+      "step": 18299
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.429785392813871e-06,
+      "loss": 0.8748,
+      "step": 18300
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.423893836908901e-06,
+      "loss": 0.6585,
+      "step": 18301
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.418004527776578e-06,
+      "loss": 0.8585,
+      "step": 18302
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.412117465559842e-06,
+      "loss": 0.8198,
+      "step": 18303
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.406232650401556e-06,
+      "loss": 0.8322,
+      "step": 18304
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.400350082444551e-06,
+      "loss": 0.9575,
+      "step": 18305
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.394469761831602e-06,
+      "loss": 0.8008,
+      "step": 18306
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.388591688705393e-06,
+      "loss": 0.8922,
+      "step": 18307
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.382715863208612e-06,
+      "loss": 0.8517,
+      "step": 18308
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.376842285483853e-06,
+      "loss": 0.8422,
+      "step": 18309
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.370970955673639e-06,
+      "loss": 0.8174,
+      "step": 18310
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.36510187392051e-06,
+      "loss": 0.7828,
+      "step": 18311
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.359235040366874e-06,
+      "loss": 0.875,
+      "step": 18312
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.353370455155129e-06,
+      "loss": 0.7628,
+      "step": 18313
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.347508118427604e-06,
+      "loss": 0.7812,
+      "step": 18314
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.341648030326564e-06,
+      "loss": 0.7425,
+      "step": 18315
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.335790190994218e-06,
+      "loss": 0.7494,
+      "step": 18316
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.329934600572774e-06,
+      "loss": 0.8433,
+      "step": 18317
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3240812592043185e-06,
+      "loss": 0.8424,
+      "step": 18318
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.318230167030915e-06,
+      "loss": 0.7789,
+      "step": 18319
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.312381324194551e-06,
+      "loss": 0.7281,
+      "step": 18320
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3065347308372e-06,
+      "loss": 0.8243,
+      "step": 18321
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3006903871007395e-06,
+      "loss": 0.8827,
+      "step": 18322
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.294848293126999e-06,
+      "loss": 0.8571,
+      "step": 18323
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2890084490577885e-06,
+      "loss": 0.8799,
+      "step": 18324
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2831708550348264e-06,
+      "loss": 0.7702,
+      "step": 18325
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.277335511199779e-06,
+      "loss": 0.8663,
+      "step": 18326
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.271502417694287e-06,
+      "loss": 0.9126,
+      "step": 18327
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.265671574659905e-06,
+      "loss": 0.8465,
+      "step": 18328
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2598429822381406e-06,
+      "loss": 0.8664,
+      "step": 18329
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.25401664057046e-06,
+      "loss": 0.7662,
+      "step": 18330
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.248192549798249e-06,
+      "loss": 0.8243,
+      "step": 18331
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.242370710062863e-06,
+      "loss": 0.7962,
+      "step": 18332
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.236551121505597e-06,
+      "loss": 0.6866,
+      "step": 18333
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.230733784267707e-06,
+      "loss": 0.9298,
+      "step": 18334
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2249186984903464e-06,
+      "loss": 0.8969,
+      "step": 18335
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.219105864314646e-06,
+      "loss": 0.7803,
+      "step": 18336
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.213295281881716e-06,
+      "loss": 0.8982,
+      "step": 18337
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.207486951332542e-06,
+      "loss": 0.6791,
+      "step": 18338
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2016808728080805e-06,
+      "loss": 0.8671,
+      "step": 18339
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.195877046449284e-06,
+      "loss": 0.9323,
+      "step": 18340
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.190075472396984e-06,
+      "loss": 0.836,
+      "step": 18341
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.184276150791969e-06,
+      "loss": 0.7823,
+      "step": 18342
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.178479081775025e-06,
+      "loss": 0.9552,
+      "step": 18343
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.172684265486796e-06,
+      "loss": 0.8369,
+      "step": 18344
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.166891702067935e-06,
+      "loss": 0.8386,
+      "step": 18345
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.161101391659042e-06,
+      "loss": 1.0494,
+      "step": 18346
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.1553133344006265e-06,
+      "loss": 0.8313,
+      "step": 18347
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.149527530433154e-06,
+      "loss": 0.8343,
+      "step": 18348
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.143743979897078e-06,
+      "loss": 0.9626,
+      "step": 18349
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.137962682932742e-06,
+      "loss": 0.8213,
+      "step": 18350
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.132183639680445e-06,
+      "loss": 0.8008,
+      "step": 18351
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.126406850280432e-06,
+      "loss": 0.6791,
+      "step": 18352
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.120632314872944e-06,
+      "loss": 0.7931,
+      "step": 18353
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.114860033598103e-06,
+      "loss": 0.7645,
+      "step": 18354
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1090900065959866e-06,
+      "loss": 0.6401,
+      "step": 18355
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.103322234006649e-06,
+      "loss": 0.7675,
+      "step": 18356
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.097556715970066e-06,
+      "loss": 0.8944,
+      "step": 18357
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.091793452626161e-06,
+      "loss": 0.6688,
+      "step": 18358
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.086032444114809e-06,
+      "loss": 0.8201,
+      "step": 18359
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.080273690575823e-06,
+      "loss": 0.8533,
+      "step": 18360
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.074517192148955e-06,
+      "loss": 0.8844,
+      "step": 18361
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.0687629489739285e-06,
+      "loss": 0.7489,
+      "step": 18362
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.063010961190397e-06,
+      "loss": 0.7266,
+      "step": 18363
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.057261228937939e-06,
+      "loss": 0.9938,
+      "step": 18364
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.051513752356121e-06,
+      "loss": 1.1469,
+      "step": 18365
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.045768531584418e-06,
+      "loss": 1.179,
+      "step": 18366
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.0400255667622536e-06,
+      "loss": 0.8138,
+      "step": 18367
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.0342848580290275e-06,
+      "loss": 0.7211,
+      "step": 18368
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.028546405524061e-06,
+      "loss": 0.8229,
+      "step": 18369
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.0228102093866095e-06,
+      "loss": 0.8568,
+      "step": 18370
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.017076269755885e-06,
+      "loss": 0.8168,
+      "step": 18371
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.011344586771073e-06,
+      "loss": 0.7111,
+      "step": 18372
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.005615160571266e-06,
+      "loss": 0.832,
+      "step": 18373
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.999887991295506e-06,
+      "loss": 0.7659,
+      "step": 18374
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.9941630790827935e-06,
+      "loss": 1.2035,
+      "step": 18375
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.988440424072062e-06,
+      "loss": 1.0618,
+      "step": 18376
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.9827200264021895e-06,
+      "loss": 0.8748,
+      "step": 18377
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.9770018862120426e-06,
+      "loss": 0.8167,
+      "step": 18378
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.971286003640376e-06,
+      "loss": 0.8297,
+      "step": 18379
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.965572378825891e-06,
+      "loss": 0.9677,
+      "step": 18380
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.959861011907287e-06,
+      "loss": 0.7735,
+      "step": 18381
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.9541519030231765e-06,
+      "loss": 0.8789,
+      "step": 18382
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.94844505231207e-06,
+      "loss": 0.7585,
+      "step": 18383
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.942740459912533e-06,
+      "loss": 0.9286,
+      "step": 18384
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.93703812596298e-06,
+      "loss": 0.6484,
+      "step": 18385
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.931338050601799e-06,
+      "loss": 0.8843,
+      "step": 18386
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.9256402339673345e-06,
+      "loss": 0.8614,
+      "step": 18387
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.919944676197876e-06,
+      "loss": 0.9309,
+      "step": 18388
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.9142513774316354e-06,
+      "loss": 0.8121,
+      "step": 18389
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.908560337806802e-06,
+      "loss": 0.8752,
+      "step": 18390
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.902871557461488e-06,
+      "loss": 0.8072,
+      "step": 18391
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.89718503653376e-06,
+      "loss": 0.8117,
+      "step": 18392
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.891500775161619e-06,
+      "loss": 0.8302,
+      "step": 18393
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.885818773483033e-06,
+      "loss": 0.7268,
+      "step": 18394
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.880139031635891e-06,
+      "loss": 0.7386,
+      "step": 18395
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.874461549758016e-06,
+      "loss": 0.7916,
+      "step": 18396
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.8687863279872426e-06,
+      "loss": 0.8503,
+      "step": 18397
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.863113366461293e-06,
+      "loss": 0.8792,
+      "step": 18398
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.857442665317804e-06,
+      "loss": 0.9188,
+      "step": 18399
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.851774224694463e-06,
+      "loss": 0.6412,
+      "step": 18400
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.846108044728805e-06,
+      "loss": 0.826,
+      "step": 18401
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.8404441255583544e-06,
+      "loss": 0.758,
+      "step": 18402
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.834782467320567e-06,
+      "loss": 0.6884,
+      "step": 18403
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.829123070152843e-06,
+      "loss": 0.7641,
+      "step": 18404
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.82346593419253e-06,
+      "loss": 0.6629,
+      "step": 18405
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.817811059576939e-06,
+      "loss": 0.8577,
+      "step": 18406
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.8121584464433176e-06,
+      "loss": 0.6328,
+      "step": 18407
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.806508094928832e-06,
+      "loss": 0.8153,
+      "step": 18408
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.800860005170606e-06,
+      "loss": 0.6386,
+      "step": 18409
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.795214177305742e-06,
+      "loss": 0.821,
+      "step": 18410
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.789570611471252e-06,
+      "loss": 0.773,
+      "step": 18411
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.7839293078040825e-06,
+      "loss": 0.7983,
+      "step": 18412
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.778290266441179e-06,
+      "loss": 0.7242,
+      "step": 18413
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.772653487519376e-06,
+      "loss": 0.8152,
+      "step": 18414
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.767018971175465e-06,
+      "loss": 0.8511,
+      "step": 18415
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.7613867175462255e-06,
+      "loss": 0.7904,
+      "step": 18416
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.7557567267683475e-06,
+      "loss": 1.2477,
+      "step": 18417
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.7501289989784225e-06,
+      "loss": 0.9837,
+      "step": 18418
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.744503534313073e-06,
+      "loss": 0.8521,
+      "step": 18419
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.738880332908815e-06,
+      "loss": 0.8735,
+      "step": 18420
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.733259394902103e-06,
+      "loss": 0.7732,
+      "step": 18421
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.727640720429396e-06,
+      "loss": 0.7101,
+      "step": 18422
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.722024309627017e-06,
+      "loss": 0.914,
+      "step": 18423
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.716410162631282e-06,
+      "loss": 1.3267,
+      "step": 18424
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.710798279578467e-06,
+      "loss": 0.9962,
+      "step": 18425
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.705188660604744e-06,
+      "loss": 0.7451,
+      "step": 18426
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.699581305846269e-06,
+      "loss": 0.7467,
+      "step": 18427
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.6939762154391015e-06,
+      "loss": 1.215,
+      "step": 18428
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.68837338951932e-06,
+      "loss": 1.1263,
+      "step": 18429
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.682772828222872e-06,
+      "loss": 1.2856,
+      "step": 18430
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.677174531685682e-06,
+      "loss": 1.1229,
+      "step": 18431
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.671578500043629e-06,
+      "loss": 0.8442,
+      "step": 18432
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.665984733432518e-06,
+      "loss": 0.7513,
+      "step": 18433
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.660393231988094e-06,
+      "loss": 0.7854,
+      "step": 18434
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.654803995846081e-06,
+      "loss": 0.8211,
+      "step": 18435
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.64921702514213e-06,
+      "loss": 0.7344,
+      "step": 18436
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.643632320011795e-06,
+      "loss": 0.7445,
+      "step": 18437
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.6380498805906576e-06,
+      "loss": 0.8304,
+      "step": 18438
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.6324697070142e-06,
+      "loss": 0.7292,
+      "step": 18439
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.6268917994178e-06,
+      "loss": 0.7815,
+      "step": 18440
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.621316157936897e-06,
+      "loss": 0.8333,
+      "step": 18441
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.615742782706769e-06,
+      "loss": 0.926,
+      "step": 18442
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.610171673862686e-06,
+      "loss": 0.8781,
+      "step": 18443
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.604602831539841e-06,
+      "loss": 0.7741,
+      "step": 18444
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.599036255873414e-06,
+      "loss": 0.7189,
+      "step": 18445
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.593471946998497e-06,
+      "loss": 0.7552,
+      "step": 18446
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.587909905050127e-06,
+      "loss": 0.8439,
+      "step": 18447
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.582350130163295e-06,
+      "loss": 1.2672,
+      "step": 18448
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.576792622472927e-06,
+      "loss": 1.0009,
+      "step": 18449
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.571237382113882e-06,
+      "loss": 0.8558,
+      "step": 18450
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 6.565684409221029e-06,
+      "loss": 0.7779,
+      "step": 18451
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.560133703929105e-06,
+      "loss": 0.7422,
+      "step": 18452
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.554585266372825e-06,
+      "loss": 0.8983,
+      "step": 18453
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.54903909668686e-06,
+      "loss": 1.1873,
+      "step": 18454
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.5434951950058e-06,
+      "loss": 1.4475,
+      "step": 18455
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.537953561464183e-06,
+      "loss": 1.0964,
+      "step": 18456
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.532414196196534e-06,
+      "loss": 0.7537,
+      "step": 18457
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.52687709933727e-06,
+      "loss": 0.7924,
+      "step": 18458
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.5213422710207695e-06,
+      "loss": 0.7811,
+      "step": 18459
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.515809711381349e-06,
+      "loss": 0.6512,
+      "step": 18460
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.510279420553333e-06,
+      "loss": 0.8025,
+      "step": 18461
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.504751398670872e-06,
+      "loss": 0.725,
+      "step": 18462
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.499225645868167e-06,
+      "loss": 0.773,
+      "step": 18463
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.4937021622793246e-06,
+      "loss": 0.7633,
+      "step": 18464
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.488180948038381e-06,
+      "loss": 0.7687,
+      "step": 18465
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.482662003279327e-06,
+      "loss": 0.7254,
+      "step": 18466
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.477145328136136e-06,
+      "loss": 0.7142,
+      "step": 18467
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.471630922742678e-06,
+      "loss": 0.8672,
+      "step": 18468
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.466118787232756e-06,
+      "loss": 0.7151,
+      "step": 18469
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.460608921740208e-06,
+      "loss": 0.8061,
+      "step": 18470
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.4551013263987045e-06,
+      "loss": 0.9437,
+      "step": 18471
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.449596001341929e-06,
+      "loss": 0.6392,
+      "step": 18472
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.444092946703495e-06,
+      "loss": 1.2119,
+      "step": 18473
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.438592162616963e-06,
+      "loss": 0.9164,
+      "step": 18474
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.433093649215827e-06,
+      "loss": 0.7964,
+      "step": 18475
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.427597406633534e-06,
+      "loss": 0.7228,
+      "step": 18476
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.422103435003479e-06,
+      "loss": 0.7811,
+      "step": 18477
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.416611734458977e-06,
+      "loss": 0.7321,
+      "step": 18478
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.4111223051333325e-06,
+      "loss": 0.8499,
+      "step": 18479
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.405635147159772e-06,
+      "loss": 0.7805,
+      "step": 18480
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.400150260671444e-06,
+      "loss": 0.8385,
+      "step": 18481
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.394667645801477e-06,
+      "loss": 0.7713,
+      "step": 18482
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.389187302682942e-06,
+      "loss": 0.9276,
+      "step": 18483
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.3837092314488314e-06,
+      "loss": 0.8204,
+      "step": 18484
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.378233432232084e-06,
+      "loss": 0.8802,
+      "step": 18485
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.372759905165615e-06,
+      "loss": 0.7793,
+      "step": 18486
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.3672886503822636e-06,
+      "loss": 0.7776,
+      "step": 18487
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.3618196680147905e-06,
+      "loss": 0.768,
+      "step": 18488
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.356352958195955e-06,
+      "loss": 0.7652,
+      "step": 18489
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.350888521058429e-06,
+      "loss": 0.7022,
+      "step": 18490
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.345426356734796e-06,
+      "loss": 0.83,
+      "step": 18491
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.339966465357661e-06,
+      "loss": 0.8021,
+      "step": 18492
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.334508847059517e-06,
+      "loss": 0.8238,
+      "step": 18493
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.329053501972793e-06,
+      "loss": 0.8278,
+      "step": 18494
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.323600430229925e-06,
+      "loss": 0.9293,
+      "step": 18495
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.318149631963244e-06,
+      "loss": 0.9143,
+      "step": 18496
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.3127011073050306e-06,
+      "loss": 0.835,
+      "step": 18497
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.307254856387523e-06,
+      "loss": 0.7433,
+      "step": 18498
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.301810879342907e-06,
+      "loss": 0.9159,
+      "step": 18499
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.296369176303296e-06,
+      "loss": 0.7834,
+      "step": 18500
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.290929747400742e-06,
+      "loss": 0.8162,
+      "step": 18501
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.2854925927672835e-06,
+      "loss": 0.6941,
+      "step": 18502
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.28005771253487e-06,
+      "loss": 0.7934,
+      "step": 18503
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.274625106835375e-06,
+      "loss": 0.7292,
+      "step": 18504
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.269194775800713e-06,
+      "loss": 0.7334,
+      "step": 18505
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.263766719562614e-06,
+      "loss": 0.798,
+      "step": 18506
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.2583409382528155e-06,
+      "loss": 0.6572,
+      "step": 18507
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.252917432003025e-06,
+      "loss": 0.7924,
+      "step": 18508
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.247496200944858e-06,
+      "loss": 0.7421,
+      "step": 18509
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.242077245209876e-06,
+      "loss": 0.8198,
+      "step": 18510
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.236660564929608e-06,
+      "loss": 0.8504,
+      "step": 18511
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.231246160235515e-06,
+      "loss": 0.7082,
+      "step": 18512
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.225834031258981e-06,
+      "loss": 0.8891,
+      "step": 18513
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.22042417813139e-06,
+      "loss": 0.8387,
+      "step": 18514
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.215016600984014e-06,
+      "loss": 0.8394,
+      "step": 18515
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.209611299948082e-06,
+      "loss": 0.8432,
+      "step": 18516
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.204208275154788e-06,
+      "loss": 0.8668,
+      "step": 18517
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.198807526735273e-06,
+      "loss": 0.9031,
+      "step": 18518
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.193409054820598e-06,
+      "loss": 0.5833,
+      "step": 18519
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.188012859541781e-06,
+      "loss": 0.796,
+      "step": 18520
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.182618941029794e-06,
+      "loss": 0.874,
+      "step": 18521
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.177227299415533e-06,
+      "loss": 0.8324,
+      "step": 18522
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.171837934829839e-06,
+      "loss": 0.7886,
+      "step": 18523
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.166450847403538e-06,
+      "loss": 0.8168,
+      "step": 18524
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.16106603726736e-06,
+      "loss": 0.8373,
+      "step": 18525
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.155683504551968e-06,
+      "loss": 0.8898,
+      "step": 18526
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.150303249388045e-06,
+      "loss": 1.2561,
+      "step": 18527
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.144925271906121e-06,
+      "loss": 1.1638,
+      "step": 18528
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.139549572236725e-06,
+      "loss": 0.8618,
+      "step": 18529
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.1341761505103405e-06,
+      "loss": 0.8319,
+      "step": 18530
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.128805006857375e-06,
+      "loss": 0.7433,
+      "step": 18531
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.123436141408168e-06,
+      "loss": 1.2496,
+      "step": 18532
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.118069554293004e-06,
+      "loss": 1.0207,
+      "step": 18533
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.112705245642192e-06,
+      "loss": 1.0013,
+      "step": 18534
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.1073432155858475e-06,
+      "loss": 0.8132,
+      "step": 18535
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.101983464254135e-06,
+      "loss": 0.7044,
+      "step": 18536
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.0966259917771495e-06,
+      "loss": 0.7784,
+      "step": 18537
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.0912707982848874e-06,
+      "loss": 0.8256,
+      "step": 18538
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.0859178839073106e-06,
+      "loss": 0.7906,
+      "step": 18539
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.0805672487743716e-06,
+      "loss": 0.7556,
+      "step": 18540
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.075218893015888e-06,
+      "loss": 0.6784,
+      "step": 18541
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.069872816761679e-06,
+      "loss": 0.8094,
+      "step": 18542
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.064529020141485e-06,
+      "loss": 0.7225,
+      "step": 18543
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.0591875032850135e-06,
+      "loss": 0.7793,
+      "step": 18544
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.053848266321871e-06,
+      "loss": 0.7799,
+      "step": 18545
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.048511309381666e-06,
+      "loss": 0.8908,
+      "step": 18546
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.0431766325939056e-06,
+      "loss": 0.8051,
+      "step": 18547
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.037844236088075e-06,
+      "loss": 0.9431,
+      "step": 18548
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.032514119993571e-06,
+      "loss": 0.8865,
+      "step": 18549
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.027186284439767e-06,
+      "loss": 0.822,
+      "step": 18550
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.021860729555939e-06,
+      "loss": 0.7946,
+      "step": 18551
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.016537455471383e-06,
+      "loss": 0.7372,
+      "step": 18552
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.01121646231525e-06,
+      "loss": 0.7387,
+      "step": 18553
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.005897750216705e-06,
+      "loss": 1.198,
+      "step": 18554
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.00058131930481e-06,
+      "loss": 0.9454,
+      "step": 18555
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.995267169708607e-06,
+      "loss": 0.7961,
+      "step": 18556
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.9899553015570595e-06,
+      "loss": 0.9234,
+      "step": 18557
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.984645714979076e-06,
+      "loss": 0.9024,
+      "step": 18558
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.979338410103541e-06,
+      "loss": 0.7479,
+      "step": 18559
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.9740333870592415e-06,
+      "loss": 0.89,
+      "step": 18560
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.968730645974918e-06,
+      "loss": 0.8941,
+      "step": 18561
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.963430186979291e-06,
+      "loss": 1.2072,
+      "step": 18562
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.9581320102010005e-06,
+      "loss": 1.1211,
+      "step": 18563
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.952836115768612e-06,
+      "loss": 0.775,
+      "step": 18564
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.9475425038106545e-06,
+      "loss": 0.7425,
+      "step": 18565
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.942251174455605e-06,
+      "loss": 0.8226,
+      "step": 18566
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.93696212783188e-06,
+      "loss": 0.778,
+      "step": 18567
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.931675364067846e-06,
+      "loss": 0.7183,
+      "step": 18568
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.926390883291821e-06,
+      "loss": 0.8012,
+      "step": 18569
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.921108685632026e-06,
+      "loss": 0.7217,
+      "step": 18570
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.915828771216681e-06,
+      "loss": 0.8098,
+      "step": 18571
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.9105511401739255e-06,
+      "loss": 0.8046,
+      "step": 18572
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.905275792631848e-06,
+      "loss": 0.7281,
+      "step": 18573
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.900002728718446e-06,
+      "loss": 0.7909,
+      "step": 18574
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.894731948561738e-06,
+      "loss": 0.8511,
+      "step": 18575
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.889463452289623e-06,
+      "loss": 0.8947,
+      "step": 18576
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.884197240029954e-06,
+      "loss": 0.7774,
+      "step": 18577
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.878933311910573e-06,
+      "loss": 0.7882,
+      "step": 18578
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.8736716680591995e-06,
+      "loss": 0.7311,
+      "step": 18579
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.868412308603533e-06,
+      "loss": 0.877,
+      "step": 18580
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.863155233671225e-06,
+      "loss": 0.7387,
+      "step": 18581
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.857900443389863e-06,
+      "loss": 0.8528,
+      "step": 18582
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.852647937886968e-06,
+      "loss": 0.8473,
+      "step": 18583
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.847397717290037e-06,
+      "loss": 0.8027,
+      "step": 18584
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.842149781726469e-06,
+      "loss": 0.7173,
+      "step": 18585
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.836904131323628e-06,
+      "loss": 0.8089,
+      "step": 18586
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.831660766208846e-06,
+      "loss": 0.7938,
+      "step": 18587
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.826419686509354e-06,
+      "loss": 0.8219,
+      "step": 18588
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.821180892352362e-06,
+      "loss": 0.8599,
+      "step": 18589
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.815944383864991e-06,
+      "loss": 1.2297,
+      "step": 18590
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.810710161174371e-06,
+      "loss": 1.0899,
+      "step": 18591
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.805478224407501e-06,
+      "loss": 1.32,
+      "step": 18592
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.800248573691358e-06,
+      "loss": 1.2142,
+      "step": 18593
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.795021209152873e-06,
+      "loss": 0.7975,
+      "step": 18594
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.789796130918912e-06,
+      "loss": 0.8192,
+      "step": 18595
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.784573339116273e-06,
+      "loss": 0.8096,
+      "step": 18596
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.779352833871721e-06,
+      "loss": 0.7785,
+      "step": 18597
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.774134615311955e-06,
+      "loss": 0.769,
+      "step": 18598
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7689186835635955e-06,
+      "loss": 0.6835,
+      "step": 18599
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.763705038753275e-06,
+      "loss": 0.7863,
+      "step": 18600
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.758493681007493e-06,
+      "loss": 0.6865,
+      "step": 18601
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.753284610452725e-06,
+      "loss": 0.8955,
+      "step": 18602
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.748077827215414e-06,
+      "loss": 0.8936,
+      "step": 18603
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.742873331421916e-06,
+      "loss": 1.0023,
+      "step": 18604
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.73767112319854e-06,
+      "loss": 0.8549,
+      "step": 18605
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7324712026715295e-06,
+      "loss": 0.5689,
+      "step": 18606
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.727273569967095e-06,
+      "loss": 0.9161,
+      "step": 18607
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.722078225211391e-06,
+      "loss": 0.8182,
+      "step": 18608
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.716885168530495e-06,
+      "loss": 0.801,
+      "step": 18609
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.711694400050427e-06,
+      "loss": 0.8452,
+      "step": 18610
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.706505919897187e-06,
+      "loss": 0.9101,
+      "step": 18611
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.701319728196675e-06,
+      "loss": 0.8725,
+      "step": 18612
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.696135825074766e-06,
+      "loss": 0.7871,
+      "step": 18613
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.690954210657285e-06,
+      "loss": 0.8745,
+      "step": 18614
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.685774885069961e-06,
+      "loss": 0.8957,
+      "step": 18615
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.680597848438518e-06,
+      "loss": 0.7323,
+      "step": 18616
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.675423100888577e-06,
+      "loss": 0.8887,
+      "step": 18617
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.670250642545738e-06,
+      "loss": 0.7452,
+      "step": 18618
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.665080473535544e-06,
+      "loss": 0.886,
+      "step": 18619
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.659912593983463e-06,
+      "loss": 1.3298,
+      "step": 18620
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6547470040149155e-06,
+      "loss": 1.0347,
+      "step": 18621
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.649583703755257e-06,
+      "loss": 0.8075,
+      "step": 18622
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.644422693329832e-06,
+      "loss": 0.7085,
+      "step": 18623
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6392639728638505e-06,
+      "loss": 0.8388,
+      "step": 18624
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6341075424825475e-06,
+      "loss": 0.9305,
+      "step": 18625
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.628953402311043e-06,
+      "loss": 0.8315,
+      "step": 18626
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.623801552474428e-06,
+      "loss": 0.8562,
+      "step": 18627
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.618651993097757e-06,
+      "loss": 0.8695,
+      "step": 18628
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.613504724305996e-06,
+      "loss": 0.9944,
+      "step": 18629
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.608359746224057e-06,
+      "loss": 0.8208,
+      "step": 18630
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.603217058976807e-06,
+      "loss": 0.7346,
+      "step": 18631
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.5980766626890665e-06,
+      "loss": 0.827,
+      "step": 18632
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.592938557485583e-06,
+      "loss": 0.7325,
+      "step": 18633
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.587802743491044e-06,
+      "loss": 0.6487,
+      "step": 18634
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.5826692208301275e-06,
+      "loss": 0.8383,
+      "step": 18635
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.5775379896274016e-06,
+      "loss": 0.9949,
+      "step": 18636
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.572409050007388e-06,
+      "loss": 0.7722,
+      "step": 18637
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.567282402094576e-06,
+      "loss": 0.5563,
+      "step": 18638
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.5621580460133886e-06,
+      "loss": 0.8005,
+      "step": 18639
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.55703598188817e-06,
+      "loss": 0.7379,
+      "step": 18640
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.551916209843255e-06,
+      "loss": 0.8242,
+      "step": 18641
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.546798730002889e-06,
+      "loss": 0.7997,
+      "step": 18642
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.541683542491261e-06,
+      "loss": 0.7911,
+      "step": 18643
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.536570647432526e-06,
+      "loss": 0.8851,
+      "step": 18644
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.531460044950776e-06,
+      "loss": 0.7757,
+      "step": 18645
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.526351735170033e-06,
+      "loss": 0.854,
+      "step": 18646
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.521245718214263e-06,
+      "loss": 0.843,
+      "step": 18647
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.516141994207413e-06,
+      "loss": 0.7443,
+      "step": 18648
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.511040563273328e-06,
+      "loss": 0.7659,
+      "step": 18649
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.50594142553581e-06,
+      "loss": 0.7579,
+      "step": 18650
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.500844581118636e-06,
+      "loss": 0.7399,
+      "step": 18651
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.495750030145508e-06,
+      "loss": 0.7959,
+      "step": 18652
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.490657772740027e-06,
+      "loss": 0.9449,
+      "step": 18653
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.485567809025804e-06,
+      "loss": 0.8013,
+      "step": 18654
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.480480139126387e-06,
+      "loss": 0.6644,
+      "step": 18655
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.475394763165209e-06,
+      "loss": 0.7201,
+      "step": 18656
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.470311681265727e-06,
+      "loss": 0.8575,
+      "step": 18657
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.465230893551299e-06,
+      "loss": 0.8296,
+      "step": 18658
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.460152400145213e-06,
+      "loss": 0.7136,
+      "step": 18659
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.455076201170739e-06,
+      "loss": 0.8033,
+      "step": 18660
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.450002296751078e-06,
+      "loss": 0.7879,
+      "step": 18661
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.444930687009353e-06,
+      "loss": 0.695,
+      "step": 18662
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.439861372068644e-06,
+      "loss": 0.882,
+      "step": 18663
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.4347943520520196e-06,
+      "loss": 0.7498,
+      "step": 18664
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.429729627082414e-06,
+      "loss": 0.7786,
+      "step": 18665
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.424667197282751e-06,
+      "loss": 0.8733,
+      "step": 18666
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.419607062775933e-06,
+      "loss": 0.8537,
+      "step": 18667
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.414549223684729e-06,
+      "loss": 0.8607,
+      "step": 18668
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.409493680131872e-06,
+      "loss": 0.8214,
+      "step": 18669
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.404440432240099e-06,
+      "loss": 0.7086,
+      "step": 18670
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.399389480132033e-06,
+      "loss": 0.7253,
+      "step": 18671
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.394340823930244e-06,
+      "loss": 0.83,
+      "step": 18672
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.389294463757288e-06,
+      "loss": 1.1238,
+      "step": 18673
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.384250399735624e-06,
+      "loss": 1.001,
+      "step": 18674
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.379208631987654e-06,
+      "loss": 0.8478,
+      "step": 18675
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.374169160635767e-06,
+      "loss": 0.7483,
+      "step": 18676
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.369131985802245e-06,
+      "loss": 0.8633,
+      "step": 18677
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.364097107609356e-06,
+      "loss": 0.8164,
+      "step": 18678
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.359064526179269e-06,
+      "loss": 0.7011,
+      "step": 18679
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.354034241634154e-06,
+      "loss": 0.8845,
+      "step": 18680
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.349006254096078e-06,
+      "loss": 0.7553,
+      "step": 18681
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.343980563687067e-06,
+      "loss": 0.7502,
+      "step": 18682
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.338957170529091e-06,
+      "loss": 0.76,
+      "step": 18683
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.333936074744062e-06,
+      "loss": 0.8472,
+      "step": 18684
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.328917276453826e-06,
+      "loss": 0.92,
+      "step": 18685
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.3239007757802214e-06,
+      "loss": 0.8103,
+      "step": 18686
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.318886572844983e-06,
+      "loss": 0.7998,
+      "step": 18687
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.313874667769781e-06,
+      "loss": 0.7993,
+      "step": 18688
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.308865060676282e-06,
+      "loss": 0.8904,
+      "step": 18689
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.3038577516860475e-06,
+      "loss": 0.7592,
+      "step": 18690
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.2988527409206014e-06,
+      "loss": 0.8236,
+      "step": 18691
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.293850028501424e-06,
+      "loss": 0.7583,
+      "step": 18692
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.288849614549929e-06,
+      "loss": 0.861,
+      "step": 18693
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.283851499187475e-06,
+      "loss": 0.9166,
+      "step": 18694
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.278855682535333e-06,
+      "loss": 0.8646,
+      "step": 18695
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.273862164714804e-06,
+      "loss": 0.8846,
+      "step": 18696
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.268870945847016e-06,
+      "loss": 0.7022,
+      "step": 18697
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.263882026053158e-06,
+      "loss": 0.7441,
+      "step": 18698
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.258895405454278e-06,
+      "loss": 0.8532,
+      "step": 18699
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.253911084171382e-06,
+      "loss": 0.9479,
+      "step": 18700
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.2489290623254805e-06,
+      "loss": 0.8166,
+      "step": 18701
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.243949340037468e-06,
+      "loss": 0.7935,
+      "step": 18702
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.238971917428193e-06,
+      "loss": 0.8456,
+      "step": 18703
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.233996794618445e-06,
+      "loss": 0.9187,
+      "step": 18704
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.229023971728985e-06,
+      "loss": 1.1896,
+      "step": 18705
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.224053448880506e-06,
+      "loss": 0.9888,
+      "step": 18706
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.21908522619361e-06,
+      "loss": 0.8383,
+      "step": 18707
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.2141193037889e-06,
+      "loss": 0.8313,
+      "step": 18708
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.2091556817869034e-06,
+      "loss": 0.8655,
+      "step": 18709
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.2041943603080455e-06,
+      "loss": 0.8245,
+      "step": 18710
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.1992353394727965e-06,
+      "loss": 0.8781,
+      "step": 18711
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.19427861940145e-06,
+      "loss": 0.6509,
+      "step": 18712
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.189324200214318e-06,
+      "loss": 0.9433,
+      "step": 18713
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.184372082031652e-06,
+      "loss": 0.9275,
+      "step": 18714
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.179422264973643e-06,
+      "loss": 0.745,
+      "step": 18715
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.174474749160396e-06,
+      "loss": 0.899,
+      "step": 18716
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.169529534712014e-06,
+      "loss": 0.8116,
+      "step": 18717
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.164586621748502e-06,
+      "loss": 0.7398,
+      "step": 18718
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.1596460103898185e-06,
+      "loss": 0.9001,
+      "step": 18719
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.154707700755856e-06,
+      "loss": 0.7543,
+      "step": 18720
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.149771692966499e-06,
+      "loss": 0.9679,
+      "step": 18721
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.144837987141527e-06,
+      "loss": 0.8478,
+      "step": 18722
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.1399065834006444e-06,
+      "loss": 0.7275,
+      "step": 18723
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.13497748186359e-06,
+      "loss": 0.8697,
+      "step": 18724
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.130050682649978e-06,
+      "loss": 0.9328,
+      "step": 18725
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.125126185879336e-06,
+      "loss": 1.2832,
+      "step": 18726
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.120203991671224e-06,
+      "loss": 1.018,
+      "step": 18727
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.115284100145079e-06,
+      "loss": 0.8347,
+      "step": 18728
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.110366511420306e-06,
+      "loss": 0.7277,
+      "step": 18729
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.105451225616265e-06,
+      "loss": 0.7377,
+      "step": 18730
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.100538242852237e-06,
+      "loss": 0.7203,
+      "step": 18731
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0956275632474515e-06,
+      "loss": 0.9253,
+      "step": 18732
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0907191869211e-06,
+      "loss": 0.8223,
+      "step": 18733
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.08581311399231e-06,
+      "loss": 0.8529,
+      "step": 18734
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.080909344580143e-06,
+      "loss": 0.8115,
+      "step": 18735
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0760078788036015e-06,
+      "loss": 0.7531,
+      "step": 18736
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.071108716781647e-06,
+      "loss": 0.6491,
+      "step": 18737
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.066211858633196e-06,
+      "loss": 0.8632,
+      "step": 18738
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0613173044770534e-06,
+      "loss": 0.8117,
+      "step": 18739
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.056425054432068e-06,
+      "loss": 0.8583,
+      "step": 18740
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.051535108616923e-06,
+      "loss": 0.7554,
+      "step": 18741
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0466474671502895e-06,
+      "loss": 0.7796,
+      "step": 18742
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.041762130150829e-06,
+      "loss": 1.1613,
+      "step": 18743
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0368790977370795e-06,
+      "loss": 1.02,
+      "step": 18744
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.0319983700275354e-06,
+      "loss": 0.7888,
+      "step": 18745
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.027119947140691e-06,
+      "loss": 0.772,
+      "step": 18746
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.022243829194917e-06,
+      "loss": 0.7219,
+      "step": 18747
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.017370016308542e-06,
+      "loss": 1.2036,
+      "step": 18748
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.012498508599872e-06,
+      "loss": 0.9443,
+      "step": 18749
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.007629306187145e-06,
+      "loss": 0.8111,
+      "step": 18750
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.002762409188511e-06,
+      "loss": 0.8682,
+      "step": 18751
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.9978978177220745e-06,
+      "loss": 0.7636,
+      "step": 18752
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.993035531905943e-06,
+      "loss": 0.7631,
+      "step": 18753
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.9881755518580875e-06,
+      "loss": 0.7459,
+      "step": 18754
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.983317877696469e-06,
+      "loss": 0.8029,
+      "step": 18755
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.978462509538972e-06,
+      "loss": 0.6661,
+      "step": 18756
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.973609447503425e-06,
+      "loss": 0.8348,
+      "step": 18757
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.968758691707643e-06,
+      "loss": 0.7971,
+      "step": 18758
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.963910242269332e-06,
+      "loss": 0.7399,
+      "step": 18759
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.959064099306154e-06,
+      "loss": 0.829,
+      "step": 18760
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.9542202629357156e-06,
+      "loss": 0.8589,
+      "step": 18761
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.9493787332756e-06,
+      "loss": 0.6777,
+      "step": 18762
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.9445395104433025e-06,
+      "loss": 0.8491,
+      "step": 18763
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.939702594556239e-06,
+      "loss": 0.9014,
+      "step": 18764
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.934867985731839e-06,
+      "loss": 0.6844,
+      "step": 18765
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.930035684087419e-06,
+      "loss": 0.7246,
+      "step": 18766
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.925205689740242e-06,
+      "loss": 0.6331,
+      "step": 18767
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.920378002807558e-06,
+      "loss": 0.8118,
+      "step": 18768
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.915552623406527e-06,
+      "loss": 0.7153,
+      "step": 18769
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.910729551654214e-06,
+      "loss": 0.6988,
+      "step": 18770
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.905908787667734e-06,
+      "loss": 1.3557,
+      "step": 18771
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.901090331564051e-06,
+      "loss": 1.0702,
+      "step": 18772
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.896274183460103e-06,
+      "loss": 0.7489,
+      "step": 18773
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.891460343472798e-06,
+      "loss": 0.8418,
+      "step": 18774
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.886648811718952e-06,
+      "loss": 0.9242,
+      "step": 18775
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.88183958831534e-06,
+      "loss": 0.8732,
+      "step": 18776
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.877032673378667e-06,
+      "loss": 0.9227,
+      "step": 18777
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.872228067025608e-06,
+      "loss": 0.7828,
+      "step": 18778
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.86742576937278e-06,
+      "loss": 0.8935,
+      "step": 18779
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.8626257805367e-06,
+      "loss": 0.8941,
+      "step": 18780
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.8578281006338874e-06,
+      "loss": 1.2986,
+      "step": 18781
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.853032729780771e-06,
+      "loss": 1.1617,
+      "step": 18782
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.848239668093723e-06,
+      "loss": 0.8819,
+      "step": 18783
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.8434489156890975e-06,
+      "loss": 0.7933,
+      "step": 18784
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.838660472683143e-06,
+      "loss": 0.8621,
+      "step": 18785
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.833874339192046e-06,
+      "loss": 1.2895,
+      "step": 18786
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.829090515332013e-06,
+      "loss": 0.9469,
+      "step": 18787
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.824309001219108e-06,
+      "loss": 0.9286,
+      "step": 18788
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.8195297969693905e-06,
+      "loss": 0.8607,
+      "step": 18789
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.81475290269886e-06,
+      "loss": 0.7944,
+      "step": 18790
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.809978318523433e-06,
+      "loss": 1.1359,
+      "step": 18791
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.8052060445589855e-06,
+      "loss": 1.0789,
+      "step": 18792
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.800436080921344e-06,
+      "loss": 0.8967,
+      "step": 18793
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.7956684277262744e-06,
+      "loss": 0.6675,
+      "step": 18794
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.7909030850894935e-06,
+      "loss": 0.7196,
+      "step": 18795
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.78614005312662e-06,
+      "loss": 0.8096,
+      "step": 18796
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.781379331953284e-06,
+      "loss": 0.7954,
+      "step": 18797
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.776620921685027e-06,
+      "loss": 0.882,
+      "step": 18798
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.771864822437311e-06,
+      "loss": 0.6906,
+      "step": 18799
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.767111034325578e-06,
+      "loss": 0.7651,
+      "step": 18800
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.7623595574652016e-06,
+      "loss": 0.8647,
+      "step": 18801
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.757610391971467e-06,
+      "loss": 0.8597,
+      "step": 18802
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.752863537959684e-06,
+      "loss": 0.8513,
+      "step": 18803
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.748118995545026e-06,
+      "loss": 0.7349,
+      "step": 18804
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.743376764842622e-06,
+      "loss": 0.8265,
+      "step": 18805
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.738636845967614e-06,
+      "loss": 0.8566,
+      "step": 18806
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.733899239035e-06,
+      "loss": 0.8834,
+      "step": 18807
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.729163944159776e-06,
+      "loss": 0.6242,
+      "step": 18808
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.72443096145685e-06,
+      "loss": 0.7865,
+      "step": 18809
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.719700291041107e-06,
+      "loss": 0.852,
+      "step": 18810
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.714971933027346e-06,
+      "loss": 0.7351,
+      "step": 18811
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.7102458875303065e-06,
+      "loss": 0.7211,
+      "step": 18812
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.705522154664721e-06,
+      "loss": 0.8865,
+      "step": 18813
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.7008007345452295e-06,
+      "loss": 1.2039,
+      "step": 18814
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.696081627286386e-06,
+      "loss": 0.9516,
+      "step": 18815
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.691364833002743e-06,
+      "loss": 0.9033,
+      "step": 18816
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.6866503518087745e-06,
+      "loss": 0.7986,
+      "step": 18817
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.681938183818879e-06,
+      "loss": 1.3402,
+      "step": 18818
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.677228329147443e-06,
+      "loss": 1.0097,
+      "step": 18819
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.672520787908774e-06,
+      "loss": 0.8094,
+      "step": 18820
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.667815560217081e-06,
+      "loss": 0.8444,
+      "step": 18821
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.663112646186618e-06,
+      "loss": 1.0129,
+      "step": 18822
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.658412045931482e-06,
+      "loss": 1.4465,
+      "step": 18823
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.65371375956577e-06,
+      "loss": 1.1005,
+      "step": 18824
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.64901778720348e-06,
+      "loss": 0.9017,
+      "step": 18825
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.64432412895861e-06,
+      "loss": 0.6941,
+      "step": 18826
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.639632784945069e-06,
+      "loss": 0.8416,
+      "step": 18827
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.63494375527671e-06,
+      "loss": 0.8384,
+      "step": 18828
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.630257040067332e-06,
+      "loss": 0.942,
+      "step": 18829
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.625572639430664e-06,
+      "loss": 0.7743,
+      "step": 18830
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.620890553480417e-06,
+      "loss": 0.8445,
+      "step": 18831
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.616210782330221e-06,
+      "loss": 0.8716,
+      "step": 18832
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.611533326093642e-06,
+      "loss": 0.6844,
+      "step": 18833
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.606858184884178e-06,
+      "loss": 0.775,
+      "step": 18834
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.602185358815347e-06,
+      "loss": 1.2704,
+      "step": 18835
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.597514848000517e-06,
+      "loss": 1.0012,
+      "step": 18836
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.592846652553029e-06,
+      "loss": 0.8981,
+      "step": 18837
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.588180772586215e-06,
+      "loss": 0.9113,
+      "step": 18838
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.583517208213284e-06,
+      "loss": 0.8182,
+      "step": 18839
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.578855959547423e-06,
+      "loss": 0.8158,
+      "step": 18840
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.574197026701776e-06,
+      "loss": 0.7849,
+      "step": 18841
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.569540409789397e-06,
+      "loss": 0.8239,
+      "step": 18842
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.564886108923305e-06,
+      "loss": 0.8095,
+      "step": 18843
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5602341242164556e-06,
+      "loss": 0.799,
+      "step": 18844
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.555584455781747e-06,
+      "loss": 0.7061,
+      "step": 18845
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.550937103732011e-06,
+      "loss": 1.2841,
+      "step": 18846
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5462920681800695e-06,
+      "loss": 1.0779,
+      "step": 18847
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.54164934923863e-06,
+      "loss": 0.8269,
+      "step": 18848
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5370089470203825e-06,
+      "loss": 1.2187,
+      "step": 18849
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.532370861637924e-06,
+      "loss": 1.2162,
+      "step": 18850
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5277350932038554e-06,
+      "loss": 1.3583,
+      "step": 18851
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.523101641830663e-06,
+      "loss": 1.0125,
+      "step": 18852
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5184705076307785e-06,
+      "loss": 0.749,
+      "step": 18853
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.513841690716636e-06,
+      "loss": 0.7255,
+      "step": 18854
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.509215191200555e-06,
+      "loss": 1.2636,
+      "step": 18855
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.504591009194803e-06,
+      "loss": 1.0537,
+      "step": 18856
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.499969144811633e-06,
+      "loss": 0.7616,
+      "step": 18857
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.495349598163223e-06,
+      "loss": 0.8676,
+      "step": 18858
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.49073236936165e-06,
+      "loss": 0.8052,
+      "step": 18859
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.48611745851899e-06,
+      "loss": 0.9458,
+      "step": 18860
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.4815048657472545e-06,
+      "loss": 0.842,
+      "step": 18861
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.4768945911583645e-06,
+      "loss": 0.6522,
+      "step": 18862
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.472286634864231e-06,
+      "loss": 0.7201,
+      "step": 18863
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.467680996976686e-06,
+      "loss": 1.3281,
+      "step": 18864
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.463077677607497e-06,
+      "loss": 1.123,
+      "step": 18865
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.4584766768683725e-06,
+      "loss": 0.8563,
+      "step": 18866
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.453877994871003e-06,
+      "loss": 0.7873,
+      "step": 18867
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.449281631726976e-06,
+      "loss": 0.753,
+      "step": 18868
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.4446875875478465e-06,
+      "loss": 0.8303,
+      "step": 18869
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.440095862445115e-06,
+      "loss": 0.6538,
+      "step": 18870
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.4355064565302255e-06,
+      "loss": 0.7502,
+      "step": 18871
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.430919369914555e-06,
+      "loss": 0.8423,
+      "step": 18872
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.426334602709425e-06,
+      "loss": 0.774,
+      "step": 18873
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.421752155026105e-06,
+      "loss": 0.909,
+      "step": 18874
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.417172026975802e-06,
+      "loss": 0.9035,
+      "step": 18875
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.412594218669696e-06,
+      "loss": 0.845,
+      "step": 18876
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.408018730218877e-06,
+      "loss": 0.7124,
+      "step": 18877
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.403445561734365e-06,
+      "loss": 1.2175,
+      "step": 18878
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.398874713327195e-06,
+      "loss": 0.9038,
+      "step": 18879
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3943061851082655e-06,
+      "loss": 0.8071,
+      "step": 18880
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3897399771884675e-06,
+      "loss": 0.8387,
+      "step": 18881
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3851760896785995e-06,
+      "loss": 0.7923,
+      "step": 18882
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.380614522689453e-06,
+      "loss": 0.7497,
+      "step": 18883
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3760552763317145e-06,
+      "loss": 0.8159,
+      "step": 18884
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3714983507160416e-06,
+      "loss": 0.7236,
+      "step": 18885
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3669437459530224e-06,
+      "loss": 0.7381,
+      "step": 18886
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.362391462153225e-06,
+      "loss": 0.6693,
+      "step": 18887
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.357841499427084e-06,
+      "loss": 0.8596,
+      "step": 18888
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.353293857885055e-06,
+      "loss": 0.7561,
+      "step": 18889
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.348748537637504e-06,
+      "loss": 0.7258,
+      "step": 18890
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.344205538794721e-06,
+      "loss": 0.8899,
+      "step": 18891
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.339664861466986e-06,
+      "loss": 0.9673,
+      "step": 18892
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.335126505764497e-06,
+      "loss": 0.8804,
+      "step": 18893
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.33059047179738e-06,
+      "loss": 1.2786,
+      "step": 18894
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3260567596757564e-06,
+      "loss": 1.146,
+      "step": 18895
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.321525369509627e-06,
+      "loss": 0.8463,
+      "step": 18896
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.316996301408971e-06,
+      "loss": 0.879,
+      "step": 18897
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.3124695554837e-06,
+      "loss": 0.7693,
+      "step": 18898
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.307945131843704e-06,
+      "loss": 0.8534,
+      "step": 18899
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.303423030598763e-06,
+      "loss": 0.7952,
+      "step": 18900
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.298903251858622e-06,
+      "loss": 0.8367,
+      "step": 18901
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.294385795733014e-06,
+      "loss": 0.8484,
+      "step": 18902
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.289870662331518e-06,
+      "loss": 0.686,
+      "step": 18903
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.2853578517637605e-06,
+      "loss": 0.8603,
+      "step": 18904
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.2808473641392395e-06,
+      "loss": 0.719,
+      "step": 18905
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.276339199567425e-06,
+      "loss": 0.9205,
+      "step": 18906
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.271833358157729e-06,
+      "loss": 0.8109,
+      "step": 18907
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.267329840019507e-06,
+      "loss": 0.6912,
+      "step": 18908
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.262828645262073e-06,
+      "loss": 0.8263,
+      "step": 18909
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.25832977399464e-06,
+      "loss": 0.8091,
+      "step": 18910
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.253833226326409e-06,
+      "loss": 0.7647,
+      "step": 18911
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.2493390023665145e-06,
+      "loss": 0.7795,
+      "step": 18912
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.244847102224014e-06,
+      "loss": 1.1469,
+      "step": 18913
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.240357526007932e-06,
+      "loss": 1.0688,
+      "step": 18914
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.235870273827236e-06,
+      "loss": 0.8532,
+      "step": 18915
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.2313853457908174e-06,
+      "loss": 0.8229,
+      "step": 18916
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.226902742007521e-06,
+      "loss": 0.8024,
+      "step": 18917
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.222422462586151e-06,
+      "loss": 0.8061,
+      "step": 18918
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.217944507635418e-06,
+      "loss": 0.8555,
+      "step": 18919
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.213468877264026e-06,
+      "loss": 0.7216,
+      "step": 18920
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.208995571580588e-06,
+      "loss": 0.7546,
+      "step": 18921
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.20452459069366e-06,
+      "loss": 0.8576,
+      "step": 18922
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.200055934711744e-06,
+      "loss": 0.7351,
+      "step": 18923
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.195589603743311e-06,
+      "loss": 0.806,
+      "step": 18924
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.19112559789675e-06,
+      "loss": 0.8004,
+      "step": 18925
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.186663917280398e-06,
+      "loss": 0.8125,
+      "step": 18926
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.182204562002534e-06,
+      "loss": 0.669,
+      "step": 18927
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.1777475321713945e-06,
+      "loss": 0.7599,
+      "step": 18928
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.173292827895137e-06,
+      "loss": 0.8425,
+      "step": 18929
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.168840449281896e-06,
+      "loss": 1.1784,
+      "step": 18930
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.16439039643971e-06,
+      "loss": 0.9906,
+      "step": 18931
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.159942669476569e-06,
+      "loss": 0.8945,
+      "step": 18932
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.1554972685004416e-06,
+      "loss": 0.7427,
+      "step": 18933
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.151054193619208e-06,
+      "loss": 0.8117,
+      "step": 18934
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.146613444940684e-06,
+      "loss": 0.7598,
+      "step": 18935
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.142175022572659e-06,
+      "loss": 0.6963,
+      "step": 18936
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.137738926622858e-06,
+      "loss": 0.8425,
+      "step": 18937
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.13330515719893e-06,
+      "loss": 0.8715,
+      "step": 18938
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.128873714408465e-06,
+      "loss": 0.7182,
+      "step": 18939
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.1244445983590455e-06,
+      "loss": 0.8618,
+      "step": 18940
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.12001780915815e-06,
+      "loss": 0.8891,
+      "step": 18941
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.115593346913193e-06,
+      "loss": 0.7708,
+      "step": 18942
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.11117121173159e-06,
+      "loss": 0.7853,
+      "step": 18943
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.106751403720654e-06,
+      "loss": 0.7312,
+      "step": 18944
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.102333922987622e-06,
+      "loss": 0.9235,
+      "step": 18945
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.097918769639764e-06,
+      "loss": 0.6299,
+      "step": 18946
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.09350594378417e-06,
+      "loss": 0.788,
+      "step": 18947
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.089095445527957e-06,
+      "loss": 0.7363,
+      "step": 18948
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.084687274978183e-06,
+      "loss": 0.7568,
+      "step": 18949
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.080281432241817e-06,
+      "loss": 0.9152,
+      "step": 18950
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.075877917425774e-06,
+      "loss": 0.8313,
+      "step": 18951
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.071476730636959e-06,
+      "loss": 0.7684,
+      "step": 18952
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.067077871982161e-06,
+      "loss": 0.8149,
+      "step": 18953
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.062681341568153e-06,
+      "loss": 1.1832,
+      "step": 18954
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.058287139501615e-06,
+      "loss": 1.1989,
+      "step": 18955
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.053895265889218e-06,
+      "loss": 0.7689,
+      "step": 18956
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.049505720837532e-06,
+      "loss": 0.7995,
+      "step": 18957
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.045118504453083e-06,
+      "loss": 0.7958,
+      "step": 18958
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.0407336168423735e-06,
+      "loss": 0.9408,
+      "step": 18959
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.036351058111809e-06,
+      "loss": 0.7938,
+      "step": 18960
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.031970828367748e-06,
+      "loss": 0.8315,
+      "step": 18961
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.027592927716495e-06,
+      "loss": 0.9325,
+      "step": 18962
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.023217356264297e-06,
+      "loss": 0.846,
+      "step": 18963
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.018844114117348e-06,
+      "loss": 0.7749,
+      "step": 18964
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.014473201381808e-06,
+      "loss": 0.8615,
+      "step": 18965
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.010104618163723e-06,
+      "loss": 0.9178,
+      "step": 18966
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.0057383645691226e-06,
+      "loss": 0.6447,
+      "step": 18967
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.001374440703998e-06,
+      "loss": 0.8054,
+      "step": 18968
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.997012846674242e-06,
+      "loss": 0.7446,
+      "step": 18969
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.992653582585692e-06,
+      "loss": 0.8997,
+      "step": 18970
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.9882966485441764e-06,
+      "loss": 0.8347,
+      "step": 18971
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.98394204465542e-06,
+      "loss": 0.806,
+      "step": 18972
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.9795897710251055e-06,
+      "loss": 0.7887,
+      "step": 18973
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.975239827758848e-06,
+      "loss": 0.7692,
+      "step": 18974
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.970892214962263e-06,
+      "loss": 0.7922,
+      "step": 18975
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.9665469327408e-06,
+      "loss": 0.837,
+      "step": 18976
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.962203981199974e-06,
+      "loss": 0.8803,
+      "step": 18977
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.957863360445157e-06,
+      "loss": 0.8608,
+      "step": 18978
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.953525070581709e-06,
+      "loss": 0.7736,
+      "step": 18979
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.94918911171489e-06,
+      "loss": 0.9141,
+      "step": 18980
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.944855483949961e-06,
+      "loss": 0.7238,
+      "step": 18981
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.940524187392092e-06,
+      "loss": 0.7995,
+      "step": 18982
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.936195222146377e-06,
+      "loss": 0.8664,
+      "step": 18983
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.9318685883179105e-06,
+      "loss": 0.917,
+      "step": 18984
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.9275442860116755e-06,
+      "loss": 0.8995,
+      "step": 18985
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.923222315332619e-06,
+      "loss": 0.8065,
+      "step": 18986
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.918902676385661e-06,
+      "loss": 0.8585,
+      "step": 18987
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.914585369275614e-06,
+      "loss": 0.6689,
+      "step": 18988
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.910270394107263e-06,
+      "loss": 0.8472,
+      "step": 18989
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.905957750985323e-06,
+      "loss": 0.8798,
+      "step": 18990
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.901647440014467e-06,
+      "loss": 0.893,
+      "step": 18991
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.897339461299288e-06,
+      "loss": 0.7454,
+      "step": 18992
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.893033814944358e-06,
+      "loss": 0.7698,
+      "step": 18993
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.888730501054172e-06,
+      "loss": 0.7784,
+      "step": 18994
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.8844295197331685e-06,
+      "loss": 0.8799,
+      "step": 18995
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.880130871085707e-06,
+      "loss": 0.9632,
+      "step": 18996
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.87583455521614e-06,
+      "loss": 0.7324,
+      "step": 18997
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.871540572228738e-06,
+      "loss": 0.9015,
+      "step": 18998
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.867248922227684e-06,
+      "loss": 0.7106,
+      "step": 18999
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.862959605317162e-06,
+      "loss": 0.8013,
+      "step": 19000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.858672621601267e-06,
+      "loss": 0.7861,
+      "step": 19001
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.854387971184026e-06,
+      "loss": 0.8398,
+      "step": 19002
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.850105654169456e-06,
+      "loss": 0.9434,
+      "step": 19003
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.845825670661462e-06,
+      "loss": 0.7583,
+      "step": 19004
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.841548020763918e-06,
+      "loss": 0.754,
+      "step": 19005
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.837272704580652e-06,
+      "loss": 0.8327,
+      "step": 19006
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.832999722215413e-06,
+      "loss": 0.7735,
+      "step": 19007
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.828729073771897e-06,
+      "loss": 1.0395,
+      "step": 19008
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.824460759353765e-06,
+      "loss": 0.818,
+      "step": 19009
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.820194779064612e-06,
+      "loss": 0.798,
+      "step": 19010
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.815931133007955e-06,
+      "loss": 0.8136,
+      "step": 19011
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.811669821287278e-06,
+      "loss": 0.8742,
+      "step": 19012
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.8074108440059984e-06,
+      "loss": 0.8777,
+      "step": 19013
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.8031542012674893e-06,
+      "loss": 0.7162,
+      "step": 19014
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.798899893175034e-06,
+      "loss": 0.8377,
+      "step": 19015
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.794647919831906e-06,
+      "loss": 0.7721,
+      "step": 19016
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7903982813412897e-06,
+      "loss": 0.7441,
+      "step": 19017
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.786150977806302e-06,
+      "loss": 0.7927,
+      "step": 19018
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.781906009330083e-06,
+      "loss": 0.7346,
+      "step": 19019
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7776633760155943e-06,
+      "loss": 0.8016,
+      "step": 19020
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7734230779658097e-06,
+      "loss": 0.7878,
+      "step": 19021
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7691851152836575e-06,
+      "loss": 0.8851,
+      "step": 19022
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7649494880719894e-06,
+      "loss": 0.8576,
+      "step": 19023
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.760716196433589e-06,
+      "loss": 0.7893,
+      "step": 19024
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.756485240471219e-06,
+      "loss": 1.2756,
+      "step": 19025
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.752256620287542e-06,
+      "loss": 1.0459,
+      "step": 19026
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7480303359851975e-06,
+      "loss": 0.8757,
+      "step": 19027
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7438063876667374e-06,
+      "loss": 0.8105,
+      "step": 19028
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.73958477543469e-06,
+      "loss": 0.8641,
+      "step": 19029
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7353654993915077e-06,
+      "loss": 0.787,
+      "step": 19030
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7311485596395745e-06,
+      "loss": 0.8827,
+      "step": 19031
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7269339562812643e-06,
+      "loss": 0.7013,
+      "step": 19032
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.722721689418851e-06,
+      "loss": 0.955,
+      "step": 19033
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7185117591545526e-06,
+      "loss": 0.8094,
+      "step": 19034
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7143041655905543e-06,
+      "loss": 0.7718,
+      "step": 19035
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.710098908828963e-06,
+      "loss": 0.8994,
+      "step": 19036
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7058959889718414e-06,
+      "loss": 0.7151,
+      "step": 19037
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7016954061211974e-06,
+      "loss": 0.9119,
+      "step": 19038
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.697497160378971e-06,
+      "loss": 0.7531,
+      "step": 19039
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6933012518470366e-06,
+      "loss": 0.8616,
+      "step": 19040
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6891076806272574e-06,
+      "loss": 0.8279,
+      "step": 19041
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.684916446821407e-06,
+      "loss": 0.8087,
+      "step": 19042
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.680727550531171e-06,
+      "loss": 0.8087,
+      "step": 19043
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.676540991858235e-06,
+      "loss": 0.7747,
+      "step": 19044
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6723567709042173e-06,
+      "loss": 0.8626,
+      "step": 19045
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.668174887770648e-06,
+      "loss": 0.7973,
+      "step": 19046
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6639953425590123e-06,
+      "loss": 0.76,
+      "step": 19047
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6598181353707628e-06,
+      "loss": 0.8741,
+      "step": 19048
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6556432663072737e-06,
+      "loss": 0.7156,
+      "step": 19049
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.651470735469864e-06,
+      "loss": 0.8411,
+      "step": 19050
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6473005429597974e-06,
+      "loss": 0.8422,
+      "step": 19051
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6431326888782926e-06,
+      "loss": 0.8318,
+      "step": 19052
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6389671733264686e-06,
+      "loss": 0.768,
+      "step": 19053
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.634803996405467e-06,
+      "loss": 1.3547,
+      "step": 19054
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.630643158216296e-06,
+      "loss": 1.0028,
+      "step": 19055
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.62648465885993e-06,
+      "loss": 0.7592,
+      "step": 19056
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6223284984373327e-06,
+      "loss": 0.8849,
+      "step": 19057
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6181746770493353e-06,
+      "loss": 0.7887,
+      "step": 19058
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.614023194796756e-06,
+      "loss": 0.8178,
+      "step": 19059
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.60987405178036e-06,
+      "loss": 0.8147,
+      "step": 19060
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.605727248100843e-06,
+      "loss": 0.9039,
+      "step": 19061
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.6015827838588366e-06,
+      "loss": 1.1371,
+      "step": 19062
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.597440659154927e-06,
+      "loss": 1.0445,
+      "step": 19063
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.593300874089656e-06,
+      "loss": 0.8265,
+      "step": 19064
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5891634287634534e-06,
+      "loss": 0.7706,
+      "step": 19065
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5850283232767844e-06,
+      "loss": 0.8465,
+      "step": 19066
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5808955577299795e-06,
+      "loss": 0.7104,
+      "step": 19067
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5767651322233363e-06,
+      "loss": 1.1648,
+      "step": 19068
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5726370468570968e-06,
+      "loss": 1.0011,
+      "step": 19069
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5685113017314697e-06,
+      "loss": 0.8044,
+      "step": 19070
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.564387896946564e-06,
+      "loss": 0.8212,
+      "step": 19071
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5602668326024437e-06,
+      "loss": 1.1893,
+      "step": 19072
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.556148108799151e-06,
+      "loss": 0.9466,
+      "step": 19073
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.55203172563664e-06,
+      "loss": 0.8798,
+      "step": 19074
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.5479176832147854e-06,
+      "loss": 0.7495,
+      "step": 19075
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.5438059816334636e-06,
+      "loss": 0.7578,
+      "step": 19076
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.539696620992461e-06,
+      "loss": 0.7106,
+      "step": 19077
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.535589601391498e-06,
+      "loss": 0.9357,
+      "step": 19078
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.5314849229302503e-06,
+      "loss": 0.678,
+      "step": 19079
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.5273825857083386e-06,
+      "loss": 0.9918,
+      "step": 19080
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.5232825898253163e-06,
+      "loss": 0.7422,
+      "step": 19081
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.519184935380704e-06,
+      "loss": 0.8295,
+      "step": 19082
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.515089622473955e-06,
+      "loss": 0.7137,
+      "step": 19083
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.5109966512044345e-06,
+      "loss": 0.7118,
+      "step": 19084
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.5069060216714853e-06,
+      "loss": 0.8004,
+      "step": 19085
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.502817733974395e-06,
+      "loss": 0.7452,
+      "step": 19086
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.498731788212384e-06,
+      "loss": 0.8533,
+      "step": 19087
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4946481844846056e-06,
+      "loss": 0.8025,
+      "step": 19088
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4905669228901704e-06,
+      "loss": 0.7877,
+      "step": 19089
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.486488003528132e-06,
+      "loss": 0.7857,
+      "step": 19090
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.482411426497478e-06,
+      "loss": 0.7061,
+      "step": 19091
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.478337191897152e-06,
+      "loss": 0.7578,
+      "step": 19092
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4742652998260515e-06,
+      "loss": 0.7684,
+      "step": 19093
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.470195750382943e-06,
+      "loss": 0.8739,
+      "step": 19094
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4661285436666357e-06,
+      "loss": 1.0788,
+      "step": 19095
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4620636797758397e-06,
+      "loss": 0.9821,
+      "step": 19096
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4580011588091874e-06,
+      "loss": 0.8512,
+      "step": 19097
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.453940980865289e-06,
+      "loss": 0.8426,
+      "step": 19098
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4498831460426763e-06,
+      "loss": 0.8353,
+      "step": 19099
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.445827654439815e-06,
+      "loss": 0.9038,
+      "step": 19100
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.441774506155171e-06,
+      "loss": 0.9442,
+      "step": 19101
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.437723701287077e-06,
+      "loss": 0.6872,
+      "step": 19102
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4336752399338643e-06,
+      "loss": 0.8822,
+      "step": 19103
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.429629122193756e-06,
+      "loss": 1.2834,
+      "step": 19104
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4255853481649835e-06,
+      "loss": 1.0012,
+      "step": 19105
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.42154391794568e-06,
+      "loss": 0.7891,
+      "step": 19106
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4175048316338997e-06,
+      "loss": 0.8286,
+      "step": 19107
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4134680893277316e-06,
+      "loss": 0.7887,
+      "step": 19108
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4094336911250744e-06,
+      "loss": 0.9454,
+      "step": 19109
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.405401637123873e-06,
+      "loss": 0.9856,
+      "step": 19110
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.4013719274219923e-06,
+      "loss": 1.1674,
+      "step": 19111
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.397344562117222e-06,
+      "loss": 1.0166,
+      "step": 19112
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3933195413072936e-06,
+      "loss": 0.7079,
+      "step": 19113
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.389296865089908e-06,
+      "loss": 1.2203,
+      "step": 19114
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.385276533562687e-06,
+      "loss": 1.0461,
+      "step": 19115
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.381258546823196e-06,
+      "loss": 0.7492,
+      "step": 19116
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.377242904968958e-06,
+      "loss": 0.7128,
+      "step": 19117
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.373229608097439e-06,
+      "loss": 0.8497,
+      "step": 19118
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3692186563060167e-06,
+      "loss": 0.8039,
+      "step": 19119
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.365210049692047e-06,
+      "loss": 0.8774,
+      "step": 19120
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3612037883528293e-06,
+      "loss": 0.8634,
+      "step": 19121
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3571998723855637e-06,
+      "loss": 0.8425,
+      "step": 19122
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3531983018874504e-06,
+      "loss": 0.9165,
+      "step": 19123
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3491990769556002e-06,
+      "loss": 0.7403,
+      "step": 19124
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3452021976870585e-06,
+      "loss": 0.7807,
+      "step": 19125
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.341207664178825e-06,
+      "loss": 0.855,
+      "step": 19126
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3372154765278775e-06,
+      "loss": 0.8892,
+      "step": 19127
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.333225634831072e-06,
+      "loss": 1.2139,
+      "step": 19128
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3292381391852425e-06,
+      "loss": 1.5272,
+      "step": 19129
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3252529896871777e-06,
+      "loss": 1.0341,
+      "step": 19130
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3212701864335893e-06,
+      "loss": 0.8765,
+      "step": 19131
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3172897295211225e-06,
+      "loss": 0.7482,
+      "step": 19132
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.313311619046411e-06,
+      "loss": 0.7597,
+      "step": 19133
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3093358551059883e-06,
+      "loss": 0.7517,
+      "step": 19134
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3053624377963445e-06,
+      "loss": 0.7214,
+      "step": 19135
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.3013913672138906e-06,
+      "loss": 0.8701,
+      "step": 19136
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2974226434550503e-06,
+      "loss": 0.7756,
+      "step": 19137
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2934562666160795e-06,
+      "loss": 0.6548,
+      "step": 19138
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.28949223679329e-06,
+      "loss": 0.7472,
+      "step": 19139
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.285530554082872e-06,
+      "loss": 0.6948,
+      "step": 19140
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2815712185809812e-06,
+      "loss": 0.7809,
+      "step": 19141
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2776142303836745e-06,
+      "loss": 0.8472,
+      "step": 19142
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.27365958958703e-06,
+      "loss": 0.936,
+      "step": 19143
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2697072962869947e-06,
+      "loss": 0.5097,
+      "step": 19144
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.265757350579501e-06,
+      "loss": 0.6961,
+      "step": 19145
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2618097525604076e-06,
+      "loss": 0.818,
+      "step": 19146
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.257864502325536e-06,
+      "loss": 0.8262,
+      "step": 19147
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2539215999706e-06,
+      "loss": 0.7423,
+      "step": 19148
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2499810455913327e-06,
+      "loss": 0.7919,
+      "step": 19149
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2460428392833473e-06,
+      "loss": 0.7612,
+      "step": 19150
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2421069811422343e-06,
+      "loss": 0.8125,
+      "step": 19151
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2381734712634948e-06,
+      "loss": 0.8021,
+      "step": 19152
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2342423097426077e-06,
+      "loss": 0.8305,
+      "step": 19153
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2303134966749748e-06,
+      "loss": 1.1462,
+      "step": 19154
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.226387032155942e-06,
+      "loss": 1.11,
+      "step": 19155
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2224629162808217e-06,
+      "loss": 0.7874,
+      "step": 19156
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2185411491448382e-06,
+      "loss": 0.8401,
+      "step": 19157
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.214621730843148e-06,
+      "loss": 0.7984,
+      "step": 19158
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.210704661470909e-06,
+      "loss": 0.8253,
+      "step": 19159
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2067899411231783e-06,
+      "loss": 0.7943,
+      "step": 19160
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.2028775698949463e-06,
+      "loss": 0.7696,
+      "step": 19161
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.198967547881182e-06,
+      "loss": 0.7997,
+      "step": 19162
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1950598751767757e-06,
+      "loss": 0.7811,
+      "step": 19163
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1911545518765517e-06,
+      "loss": 0.8817,
+      "step": 19164
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.187251578075312e-06,
+      "loss": 0.8407,
+      "step": 19165
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1833509538677808e-06,
+      "loss": 0.7482,
+      "step": 19166
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.179452679348582e-06,
+      "loss": 0.8375,
+      "step": 19167
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1755567546123745e-06,
+      "loss": 1.215,
+      "step": 19168
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.171663179753681e-06,
+      "loss": 1.1685,
+      "step": 19169
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.167771954867005e-06,
+      "loss": 0.8041,
+      "step": 19170
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.163883080046792e-06,
+      "loss": 0.7972,
+      "step": 19171
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1599965553874122e-06,
+      "loss": 0.8142,
+      "step": 19172
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1561123809831894e-06,
+      "loss": 0.8087,
+      "step": 19173
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.152230556928393e-06,
+      "loss": 0.7799,
+      "step": 19174
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1483510833172468e-06,
+      "loss": 0.8763,
+      "step": 19175
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1444739602438876e-06,
+      "loss": 0.8996,
+      "step": 19176
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1405991878024068e-06,
+      "loss": 0.6302,
+      "step": 19177
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1367267660868504e-06,
+      "loss": 0.7165,
+      "step": 19178
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.1328566951912107e-06,
+      "loss": 0.8865,
+      "step": 19179
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1289889752093903e-06,
+      "loss": 0.8347,
+      "step": 19180
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1251236062352806e-06,
+      "loss": 0.8203,
+      "step": 19181
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1212605883626735e-06,
+      "loss": 0.7634,
+      "step": 19182
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1173999216853268e-06,
+      "loss": 0.7349,
+      "step": 19183
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1135416062969438e-06,
+      "loss": 0.8555,
+      "step": 19184
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1096856422911492e-06,
+      "loss": 0.7957,
+      "step": 19185
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.105832029761535e-06,
+      "loss": 0.8531,
+      "step": 19186
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.1019807688016267e-06,
+      "loss": 0.7822,
+      "step": 19187
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0981318595049046e-06,
+      "loss": 0.851,
+      "step": 19188
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.09428530196475e-06,
+      "loss": 0.7538,
+      "step": 19189
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0904410962745434e-06,
+      "loss": 0.7896,
+      "step": 19190
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.086599242527566e-06,
+      "loss": 0.8848,
+      "step": 19191
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.082759740817065e-06,
+      "loss": 0.7993,
+      "step": 19192
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.078922591236222e-06,
+      "loss": 0.8135,
+      "step": 19193
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.075087793878162e-06,
+      "loss": 0.8259,
+      "step": 19194
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0712553488359554e-06,
+      "loss": 0.8103,
+      "step": 19195
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.067425256202605e-06,
+      "loss": 0.7768,
+      "step": 19196
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.063597516071082e-06,
+      "loss": 0.7273,
+      "step": 19197
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.059772128534266e-06,
+      "loss": 0.7713,
+      "step": 19198
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0559490936849955e-06,
+      "loss": 0.8366,
+      "step": 19199
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.052128411616073e-06,
+      "loss": 0.8703,
+      "step": 19200
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.048310082420214e-06,
+      "loss": 0.8174,
+      "step": 19201
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0444941061900766e-06,
+      "loss": 1.1459,
+      "step": 19202
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.040680483018299e-06,
+      "loss": 1.0433,
+      "step": 19203
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0368692129974175e-06,
+      "loss": 0.8651,
+      "step": 19204
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.033060296219914e-06,
+      "loss": 0.816,
+      "step": 19205
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0292537327782588e-06,
+      "loss": 0.9254,
+      "step": 19206
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.025449522764834e-06,
+      "loss": 0.8495,
+      "step": 19207
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.021647666271943e-06,
+      "loss": 0.9551,
+      "step": 19208
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0178481633918565e-06,
+      "loss": 0.9427,
+      "step": 19209
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.014051014216823e-06,
+      "loss": 0.7021,
+      "step": 19210
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.010256218838947e-06,
+      "loss": 0.8703,
+      "step": 19211
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.0064637773503546e-06,
+      "loss": 1.1862,
+      "step": 19212
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.002673689843094e-06,
+      "loss": 1.1028,
+      "step": 19213
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9988859564091255e-06,
+      "loss": 0.9205,
+      "step": 19214
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.995100577140375e-06,
+      "loss": 1.1843,
+      "step": 19215
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9913175521287475e-06,
+      "loss": 0.9395,
+      "step": 19216
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.987536881466013e-06,
+      "loss": 0.845,
+      "step": 19217
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.983758565243944e-06,
+      "loss": 0.8272,
+      "step": 19218
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9799826035542432e-06,
+      "loss": 1.1476,
+      "step": 19219
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.97620899648855e-06,
+      "loss": 1.0208,
+      "step": 19220
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9724377441384342e-06,
+      "loss": 0.8057,
+      "step": 19221
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.968668846595435e-06,
+      "loss": 0.7926,
+      "step": 19222
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9649023039510225e-06,
+      "loss": 0.7077,
+      "step": 19223
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9611381162966022e-06,
+      "loss": 0.8215,
+      "step": 19224
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.957376283723534e-06,
+      "loss": 0.6771,
+      "step": 19225
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.953616806323112e-06,
+      "loss": 0.8098,
+      "step": 19226
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.949859684186562e-06,
+      "loss": 0.8024,
+      "step": 19227
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.94610491740509e-06,
+      "loss": 0.8847,
+      "step": 19228
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9423525060698233e-06,
+      "loss": 0.7602,
+      "step": 19229
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9386024502718212e-06,
+      "loss": 0.8119,
+      "step": 19230
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9348547501020895e-06,
+      "loss": 0.8272,
+      "step": 19231
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9311094056515885e-06,
+      "loss": 0.8869,
+      "step": 19232
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.927366417011235e-06,
+      "loss": 1.2565,
+      "step": 19233
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.923625784271822e-06,
+      "loss": 0.9479,
+      "step": 19234
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9198875075241883e-06,
+      "loss": 0.7837,
+      "step": 19235
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9161515868590283e-06,
+      "loss": 0.744,
+      "step": 19236
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9124180223670026e-06,
+      "loss": 0.7799,
+      "step": 19237
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9086868141387504e-06,
+      "loss": 0.7959,
+      "step": 19238
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9049579622648095e-06,
+      "loss": 0.6878,
+      "step": 19239
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9012314668356855e-06,
+      "loss": 0.9589,
+      "step": 19240
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8975073279418064e-06,
+      "loss": 0.8695,
+      "step": 19241
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8937855456735662e-06,
+      "loss": 0.735,
+      "step": 19242
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.890066120121271e-06,
+      "loss": 0.8507,
+      "step": 19243
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8863490513752144e-06,
+      "loss": 0.8322,
+      "step": 19244
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.882634339525603e-06,
+      "loss": 0.9245,
+      "step": 19245
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.878921984662575e-06,
+      "loss": 0.7769,
+      "step": 19246
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.875211986876236e-06,
+      "loss": 0.812,
+      "step": 19247
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8715043462566373e-06,
+      "loss": 1.2511,
+      "step": 19248
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8677990628937503e-06,
+      "loss": 0.9852,
+      "step": 19249
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.864096136877492e-06,
+      "loss": 0.7798,
+      "step": 19250
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.860395568297758e-06,
+      "loss": 0.8785,
+      "step": 19251
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.856697357244331e-06,
+      "loss": 0.8966,
+      "step": 19252
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8530015038069625e-06,
+      "loss": 0.7704,
+      "step": 19253
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.849308008075369e-06,
+      "loss": 0.7997,
+      "step": 19254
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8456168701392005e-06,
+      "loss": 0.7751,
+      "step": 19255
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8419280900879973e-06,
+      "loss": 1.268,
+      "step": 19256
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.83824166801131e-06,
+      "loss": 1.1616,
+      "step": 19257
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.834557603998611e-06,
+      "loss": 0.6493,
+      "step": 19258
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8308758981392847e-06,
+      "loss": 0.8619,
+      "step": 19259
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.827196550522726e-06,
+      "loss": 0.8258,
+      "step": 19260
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.823519561238197e-06,
+      "loss": 0.8476,
+      "step": 19261
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8198449303749377e-06,
+      "loss": 0.6577,
+      "step": 19262
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8161726580221427e-06,
+      "loss": 1.1448,
+      "step": 19263
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.81250274426893e-06,
+      "loss": 1.026,
+      "step": 19264
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.808835189204373e-06,
+      "loss": 0.8013,
+      "step": 19265
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.805169992917467e-06,
+      "loss": 0.8117,
+      "step": 19266
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.801507155497185e-06,
+      "loss": 0.9031,
+      "step": 19267
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7978466770324006e-06,
+      "loss": 0.8358,
+      "step": 19268
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7941885576119654e-06,
+      "loss": 0.8051,
+      "step": 19269
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.790532797324663e-06,
+      "loss": 0.6973,
+      "step": 19270
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.786879396259201e-06,
+      "loss": 0.8711,
+      "step": 19271
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7832283545042413e-06,
+      "loss": 0.8246,
+      "step": 19272
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7795796721484247e-06,
+      "loss": 0.9064,
+      "step": 19273
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7759333492802795e-06,
+      "loss": 0.7538,
+      "step": 19274
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7722893859882914e-06,
+      "loss": 0.7932,
+      "step": 19275
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7686477823609224e-06,
+      "loss": 0.7549,
+      "step": 19276
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7650085384865464e-06,
+      "loss": 0.6692,
+      "step": 19277
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7613716544534598e-06,
+      "loss": 0.7331,
+      "step": 19278
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7577371303499577e-06,
+      "loss": 0.7869,
+      "step": 19279
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.754104966264248e-06,
+      "loss": 0.8775,
+      "step": 19280
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.7504751622844604e-06,
+      "loss": 0.8007,
+      "step": 19281
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.746847718498691e-06,
+      "loss": 0.855,
+      "step": 19282
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.743222634995002e-06,
+      "loss": 0.7825,
+      "step": 19283
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.739599911861346e-06,
+      "loss": 0.755,
+      "step": 19284
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.7359795491856632e-06,
+      "loss": 0.8839,
+      "step": 19285
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.732361547055795e-06,
+      "loss": 0.7045,
+      "step": 19286
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.728745905559571e-06,
+      "loss": 0.7654,
+      "step": 19287
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.7251326247847097e-06,
+      "loss": 0.7954,
+      "step": 19288
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.721521704818941e-06,
+      "loss": 0.8473,
+      "step": 19289
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.717913145749884e-06,
+      "loss": 0.726,
+      "step": 19290
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.714306947665113e-06,
+      "loss": 0.8621,
+      "step": 19291
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.7107031106521576e-06,
+      "loss": 0.8527,
+      "step": 19292
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.70710163479847e-06,
+      "loss": 0.9057,
+      "step": 19293
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.7035025201914475e-06,
+      "loss": 0.7131,
+      "step": 19294
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6999057669184757e-06,
+      "loss": 0.7568,
+      "step": 19295
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.696311375066807e-06,
+      "loss": 0.818,
+      "step": 19296
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.692719344723704e-06,
+      "loss": 0.8133,
+      "step": 19297
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6891296759763096e-06,
+      "loss": 0.8718,
+      "step": 19298
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6855423689117974e-06,
+      "loss": 0.7555,
+      "step": 19299
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6819574236171653e-06,
+      "loss": 0.8385,
+      "step": 19300
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.678374840179454e-06,
+      "loss": 0.8044,
+      "step": 19301
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6747946186856164e-06,
+      "loss": 0.762,
+      "step": 19302
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6712167592225167e-06,
+      "loss": 0.7751,
+      "step": 19303
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6676412618770185e-06,
+      "loss": 0.823,
+      "step": 19304
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6640681267358745e-06,
+      "loss": 0.8142,
+      "step": 19305
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.660497353885805e-06,
+      "loss": 0.8858,
+      "step": 19306
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6569289434134723e-06,
+      "loss": 0.7655,
+      "step": 19307
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.653362895405498e-06,
+      "loss": 0.9299,
+      "step": 19308
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6497992099484004e-06,
+      "loss": 0.9108,
+      "step": 19309
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6462378871286775e-06,
+      "loss": 0.9213,
+      "step": 19310
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6426789270327713e-06,
+      "loss": 0.8021,
+      "step": 19311
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6391223297470567e-06,
+      "loss": 0.901,
+      "step": 19312
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6355680953578323e-06,
+      "loss": 1.2944,
+      "step": 19313
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.632016223951372e-06,
+      "loss": 1.0934,
+      "step": 19314
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6284667156138752e-06,
+      "loss": 0.9256,
+      "step": 19315
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6249195704314723e-06,
+      "loss": 0.8215,
+      "step": 19316
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6213747884902716e-06,
+      "loss": 0.8126,
+      "step": 19317
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6178323698762943e-06,
+      "loss": 0.7504,
+      "step": 19318
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6142923146755037e-06,
+      "loss": 0.8451,
+      "step": 19319
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6107546229738435e-06,
+      "loss": 0.7964,
+      "step": 19320
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.607219294857144e-06,
+      "loss": 0.9197,
+      "step": 19321
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.6036863304112145e-06,
+      "loss": 0.7206,
+      "step": 19322
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.600155729721787e-06,
+      "loss": 0.722,
+      "step": 19323
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5966274928745705e-06,
+      "loss": 0.802,
+      "step": 19324
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.593101619955174e-06,
+      "loss": 0.8098,
+      "step": 19325
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5895781110491733e-06,
+      "loss": 0.7034,
+      "step": 19326
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5860569662420897e-06,
+      "loss": 0.7561,
+      "step": 19327
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5825381856193875e-06,
+      "loss": 0.7813,
+      "step": 19328
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.579021769266432e-06,
+      "loss": 0.751,
+      "step": 19329
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5755077172685993e-06,
+      "loss": 0.7341,
+      "step": 19330
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5719960297111545e-06,
+      "loss": 0.7578,
+      "step": 19331
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.568486706679318e-06,
+      "loss": 0.7675,
+      "step": 19332
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5649797482582894e-06,
+      "loss": 0.7884,
+      "step": 19333
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.561475154533155e-06,
+      "loss": 0.8129,
+      "step": 19334
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5579729255889694e-06,
+      "loss": 0.6889,
+      "step": 19335
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5544730615107425e-06,
+      "loss": 0.8903,
+      "step": 19336
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5509755623834064e-06,
+      "loss": 1.4036,
+      "step": 19337
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5474804282918485e-06,
+      "loss": 1.0559,
+      "step": 19338
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.543987659320879e-06,
+      "loss": 0.6124,
+      "step": 19339
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5404972555552965e-06,
+      "loss": 0.8786,
+      "step": 19340
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5370092170797777e-06,
+      "loss": 0.9126,
+      "step": 19341
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5335235439789883e-06,
+      "loss": 0.877,
+      "step": 19342
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5300402363375496e-06,
+      "loss": 0.8094,
+      "step": 19343
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5265592942399606e-06,
+      "loss": 0.756,
+      "step": 19344
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.523080717770698e-06,
+      "loss": 0.8789,
+      "step": 19345
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5196045070142284e-06,
+      "loss": 0.8281,
+      "step": 19346
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.516130662054883e-06,
+      "loss": 0.8067,
+      "step": 19347
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.512659182976973e-06,
+      "loss": 0.7969,
+      "step": 19348
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5091900698647642e-06,
+      "loss": 0.7533,
+      "step": 19349
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.5057233228024444e-06,
+      "loss": 0.7159,
+      "step": 19350
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.502258941874136e-06,
+      "loss": 1.2689,
+      "step": 19351
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.498796927163938e-06,
+      "loss": 0.9524,
+      "step": 19352
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4953372787558604e-06,
+      "loss": 0.8359,
+      "step": 19353
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4918799967338703e-06,
+      "loss": 0.8332,
+      "step": 19354
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4884250811818664e-06,
+      "loss": 0.6793,
+      "step": 19355
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4849725321837158e-06,
+      "loss": 0.7296,
+      "step": 19356
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4815223498231953e-06,
+      "loss": 0.8919,
+      "step": 19357
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4780745341840495e-06,
+      "loss": 0.8973,
+      "step": 19358
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4746290853499444e-06,
+      "loss": 0.9206,
+      "step": 19359
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4711860034044907e-06,
+      "loss": 0.728,
+      "step": 19360
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.467745288431267e-06,
+      "loss": 0.7589,
+      "step": 19361
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.464306940513772e-06,
+      "loss": 0.7999,
+      "step": 19362
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4608709597354616e-06,
+      "loss": 0.8152,
+      "step": 19363
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.457437346179703e-06,
+      "loss": 1.0275,
+      "step": 19364
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.454006099929862e-06,
+      "loss": 0.9212,
+      "step": 19365
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4505772210691837e-06,
+      "loss": 0.6744,
+      "step": 19366
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.447150709680879e-06,
+      "loss": 0.8831,
+      "step": 19367
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4437265658481367e-06,
+      "loss": 0.7203,
+      "step": 19368
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4403047896540577e-06,
+      "loss": 0.8082,
+      "step": 19369
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4368853811816637e-06,
+      "loss": 0.8619,
+      "step": 19370
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.433468340513945e-06,
+      "loss": 0.8014,
+      "step": 19371
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4300536677338558e-06,
+      "loss": 0.7955,
+      "step": 19372
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.426641362924231e-06,
+      "loss": 0.9352,
+      "step": 19373
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4232314261679156e-06,
+      "loss": 0.8449,
+      "step": 19374
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.419823857547665e-06,
+      "loss": 0.731,
+      "step": 19375
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4164186571461577e-06,
+      "loss": 0.8399,
+      "step": 19376
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.41301582504605e-06,
+      "loss": 1.353,
+      "step": 19377
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.409615361329942e-06,
+      "loss": 1.0973,
+      "step": 19378
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4062172660803352e-06,
+      "loss": 1.2147,
+      "step": 19379
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.402821539379707e-06,
+      "loss": 0.9592,
+      "step": 19380
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.399428181310481e-06,
+      "loss": 0.8677,
+      "step": 19381
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.396037191955003e-06,
+      "loss": 0.7101,
+      "step": 19382
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.3926485713955617e-06,
+      "loss": 0.6935,
+      "step": 19383
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.3892623197144138e-06,
+      "loss": 1.2606,
+      "step": 19384
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.3858784369937382e-06,
+      "loss": 1.1531,
+      "step": 19385
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.382496923315647e-06,
+      "loss": 0.9152,
+      "step": 19386
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.37911777876223e-06,
+      "loss": 0.7795,
+      "step": 19387
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.375741003415488e-06,
+      "loss": 0.8419,
+      "step": 19388
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.372366597357356e-06,
+      "loss": 0.7653,
+      "step": 19389
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3689945606697463e-06,
+      "loss": 0.8241,
+      "step": 19390
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3656248934345036e-06,
+      "loss": 0.8155,
+      "step": 19391
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.362257595733375e-06,
+      "loss": 0.7961,
+      "step": 19392
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.358892667648127e-06,
+      "loss": 0.7476,
+      "step": 19393
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3555301092604067e-06,
+      "loss": 0.8636,
+      "step": 19394
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.352169920651803e-06,
+      "loss": 0.8363,
+      "step": 19395
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3488121019038854e-06,
+      "loss": 0.7985,
+      "step": 19396
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3454566530981436e-06,
+      "loss": 0.862,
+      "step": 19397
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3421035743160124e-06,
+      "loss": 0.7414,
+      "step": 19398
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.33875286563886e-06,
+      "loss": 1.2611,
+      "step": 19399
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3354045271480217e-06,
+      "loss": 0.9673,
+      "step": 19400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3320585589247545e-06,
+      "loss": 0.7542,
+      "step": 19401
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.32871496105026e-06,
+      "loss": 0.8772,
+      "step": 19402
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3253737336056956e-06,
+      "loss": 0.8021,
+      "step": 19403
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3220348766721302e-06,
+      "loss": 0.8409,
+      "step": 19404
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.318698390330598e-06,
+      "loss": 0.8789,
+      "step": 19405
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3153642746621016e-06,
+      "loss": 0.9394,
+      "step": 19406
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.312032529747543e-06,
+      "loss": 0.8815,
+      "step": 19407
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3087031556677573e-06,
+      "loss": 0.8476,
+      "step": 19408
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.305376152503591e-06,
+      "loss": 0.8075,
+      "step": 19409
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3020515203357574e-06,
+      "loss": 0.7971,
+      "step": 19410
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.298729259244958e-06,
+      "loss": 0.8949,
+      "step": 19411
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2954093693117963e-06,
+      "loss": 0.919,
+      "step": 19412
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2920918506168843e-06,
+      "loss": 0.6101,
+      "step": 19413
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2887767032407135e-06,
+      "loss": 0.7844,
+      "step": 19414
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2854639272637424e-06,
+      "loss": 0.6922,
+      "step": 19415
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2821535227663837e-06,
+      "loss": 0.8953,
+      "step": 19416
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.278845489828951e-06,
+      "loss": 0.7446,
+      "step": 19417
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2755398285317475e-06,
+      "loss": 0.895,
+      "step": 19418
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2722365389549973e-06,
+      "loss": 0.6626,
+      "step": 19419
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.26893562117888e-06,
+      "loss": 0.767,
+      "step": 19420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2656370752834777e-06,
+      "loss": 1.1388,
+      "step": 19421
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2623409013488805e-06,
+      "loss": 1.0547,
+      "step": 19422
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2590470994550585e-06,
+      "loss": 0.7359,
+      "step": 19423
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.255755669681958e-06,
+      "loss": 0.7726,
+      "step": 19424
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.252466612109472e-06,
+      "loss": 0.7562,
+      "step": 19425
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.249179926817413e-06,
+      "loss": 0.8457,
+      "step": 19426
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2458956138855404e-06,
+      "loss": 1.1667,
+      "step": 19427
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.242613673393568e-06,
+      "loss": 1.0403,
+      "step": 19428
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2393341054211535e-06,
+      "loss": 0.7917,
+      "step": 19429
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2360569100478902e-06,
+      "loss": 0.7861,
+      "step": 19430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2327820873533135e-06,
+      "loss": 0.8335,
+      "step": 19431
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.229509637416882e-06,
+      "loss": 1.144,
+      "step": 19432
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2262395603180442e-06,
+      "loss": 1.0282,
+      "step": 19433
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2229718561361467e-06,
+      "loss": 0.7135,
+      "step": 19434
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2197065249504934e-06,
+      "loss": 0.8375,
+      "step": 19435
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.216443566840354e-06,
+      "loss": 0.7897,
+      "step": 19436
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.213182981884876e-06,
+      "loss": 0.765,
+      "step": 19437
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2099247701632407e-06,
+      "loss": 0.8308,
+      "step": 19438
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2066689317544963e-06,
+      "loss": 0.8411,
+      "step": 19439
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.203415466737657e-06,
+      "loss": 0.7757,
+      "step": 19440
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2001643751917045e-06,
+      "loss": 0.8557,
+      "step": 19441
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.19691565719552e-06,
+      "loss": 0.69,
+      "step": 19442
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1936693128279505e-06,
+      "loss": 0.8236,
+      "step": 19443
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.19042534216779e-06,
+      "loss": 0.7464,
+      "step": 19444
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.187183745293775e-06,
+      "loss": 0.8173,
+      "step": 19445
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1839445222845644e-06,
+      "loss": 0.7201,
+      "step": 19446
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1807076732187737e-06,
+      "loss": 0.9315,
+      "step": 19447
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1774731981749617e-06,
+      "loss": 0.8431,
+      "step": 19448
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1742410972316218e-06,
+      "loss": 1.2058,
+      "step": 19449
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1710113704672132e-06,
+      "loss": 1.0434,
+      "step": 19450
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.167784017960106e-06,
+      "loss": 0.9074,
+      "step": 19451
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1645590397886272e-06,
+      "loss": 0.9443,
+      "step": 19452
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.161336436031036e-06,
+      "loss": 0.7288,
+      "step": 19453
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1581162067655592e-06,
+      "loss": 0.8292,
+      "step": 19454
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1548983520703557e-06,
+      "loss": 0.812,
+      "step": 19455
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.151682872023486e-06,
+      "loss": 0.8356,
+      "step": 19456
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.148469766703032e-06,
+      "loss": 0.7292,
+      "step": 19457
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.145259036186953e-06,
+      "loss": 0.7369,
+      "step": 19458
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1420506805531536e-06,
+      "loss": 0.7372,
+      "step": 19459
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.13884469987955e-06,
+      "loss": 1.0994,
+      "step": 19460
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1356410942439007e-06,
+      "loss": 1.4364,
+      "step": 19461
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.132439863723956e-06,
+      "loss": 1.0956,
+      "step": 19462
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1292410083974422e-06,
+      "loss": 0.7943,
+      "step": 19463
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.126044528341975e-06,
+      "loss": 0.7358,
+      "step": 19464
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1228504236351255e-06,
+      "loss": 0.8143,
+      "step": 19465
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1196586943544207e-06,
+      "loss": 0.7305,
+      "step": 19466
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.116469340577332e-06,
+      "loss": 0.9218,
+      "step": 19467
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1132823623812414e-06,
+      "loss": 1.3783,
+      "step": 19468
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.110097759843499e-06,
+      "loss": 1.4279,
+      "step": 19469
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1069155330414204e-06,
+      "loss": 0.992,
+      "step": 19470
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.10373568205221e-06,
+      "loss": 0.7899,
+      "step": 19471
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.1005582069530294e-06,
+      "loss": 0.7721,
+      "step": 19472
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.097383107821038e-06,
+      "loss": 0.8186,
+      "step": 19473
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.094210384733253e-06,
+      "loss": 0.9342,
+      "step": 19474
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0910400377667006e-06,
+      "loss": 0.9263,
+      "step": 19475
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.087872066998309e-06,
+      "loss": 0.8111,
+      "step": 19476
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0847064725049713e-06,
+      "loss": 0.8412,
+      "step": 19477
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.081543254363494e-06,
+      "loss": 0.7422,
+      "step": 19478
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0783824126506814e-06,
+      "loss": 0.7804,
+      "step": 19479
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0752239474432165e-06,
+      "loss": 0.78,
+      "step": 19480
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0720678588177613e-06,
+      "loss": 0.8571,
+      "step": 19481
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.06891414685092e-06,
+      "loss": 1.3151,
+      "step": 19482
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.065762811619232e-06,
+      "loss": 1.0877,
+      "step": 19483
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.06261385319918e-06,
+      "loss": 0.8341,
+      "step": 19484
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.059467271667159e-06,
+      "loss": 1.2469,
+      "step": 19485
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0563230670995747e-06,
+      "loss": 0.9157,
+      "step": 19486
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0531812395727102e-06,
+      "loss": 0.8747,
+      "step": 19487
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0500417891628155e-06,
+      "loss": 0.7974,
+      "step": 19488
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0469047159461075e-06,
+      "loss": 0.7682,
+      "step": 19489
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.043770019998703e-06,
+      "loss": 0.7605,
+      "step": 19490
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0406377013966747e-06,
+      "loss": 0.8932,
+      "step": 19491
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0375077602160507e-06,
+      "loss": 0.7843,
+      "step": 19492
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0343801965327923e-06,
+      "loss": 0.812,
+      "step": 19493
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0312550104227944e-06,
+      "loss": 0.8262,
+      "step": 19494
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0281322019619185e-06,
+      "loss": 0.798,
+      "step": 19495
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0250117712259487e-06,
+      "loss": 0.8424,
+      "step": 19496
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0218937182906017e-06,
+      "loss": 0.8055,
+      "step": 19497
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.018778043231584e-06,
+      "loss": 0.8358,
+      "step": 19498
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0156647461244794e-06,
+      "loss": 0.8423,
+      "step": 19499
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.012553827044861e-06,
+      "loss": 0.9479,
+      "step": 19500
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0094452860682123e-06,
+      "loss": 0.8513,
+      "step": 19501
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.006339123270007e-06,
+      "loss": 0.8347,
+      "step": 19502
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.0032353387256063e-06,
+      "loss": 0.7946,
+      "step": 19503
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.000133932510351e-06,
+      "loss": 0.8675,
+      "step": 19504
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.997034904699491e-06,
+      "loss": 0.8296,
+      "step": 19505
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.993938255368244e-06,
+      "loss": 0.7986,
+      "step": 19506
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9908439845917838e-06,
+      "loss": 0.7744,
+      "step": 19507
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9877520924451943e-06,
+      "loss": 0.7924,
+      "step": 19508
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9846625790035043e-06,
+      "loss": 0.8577,
+      "step": 19509
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.98157544434171e-06,
+      "loss": 0.7558,
+      "step": 19510
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.978490688534718e-06,
+      "loss": 0.7753,
+      "step": 19511
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.975408311657412e-06,
+      "loss": 0.8468,
+      "step": 19512
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9723283137845883e-06,
+      "loss": 0.7289,
+      "step": 19513
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9692506949909985e-06,
+      "loss": 0.912,
+      "step": 19514
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9661754553513378e-06,
+      "loss": 0.7914,
+      "step": 19515
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.963102594940236e-06,
+      "loss": 0.5991,
+      "step": 19516
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9600321138322664e-06,
+      "loss": 0.8807,
+      "step": 19517
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9569640121019696e-06,
+      "loss": 0.8232,
+      "step": 19518
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.953898289823786e-06,
+      "loss": 0.8001,
+      "step": 19519
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.950834947072122e-06,
+      "loss": 0.783,
+      "step": 19520
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9477739839213194e-06,
+      "loss": 0.6808,
+      "step": 19521
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9447154004456737e-06,
+      "loss": 0.8583,
+      "step": 19522
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.941659196719414e-06,
+      "loss": 0.9178,
+      "step": 19523
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9386053728167155e-06,
+      "loss": 0.7337,
+      "step": 19524
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9355539288116955e-06,
+      "loss": 0.6689,
+      "step": 19525
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.932504864778395e-06,
+      "loss": 0.7649,
+      "step": 19526
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.929458180790822e-06,
+      "loss": 1.2576,
+      "step": 19527
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9264138769229278e-06,
+      "loss": 1.4855,
+      "step": 19528
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.923371953248576e-06,
+      "loss": 0.9645,
+      "step": 19529
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9203324098416075e-06,
+      "loss": 0.9064,
+      "step": 19530
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9172952467757964e-06,
+      "loss": 0.6782,
+      "step": 19531
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9142604641248287e-06,
+      "loss": 0.8487,
+      "step": 19532
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9112280619623777e-06,
+      "loss": 0.8097,
+      "step": 19533
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9081980403620412e-06,
+      "loss": 0.8953,
+      "step": 19534
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.905170399397338e-06,
+      "loss": 0.7814,
+      "step": 19535
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9021451391417532e-06,
+      "loss": 0.8638,
+      "step": 19536
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8991222596687174e-06,
+      "loss": 0.7393,
+      "step": 19537
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.896101761051583e-06,
+      "loss": 0.7041,
+      "step": 19538
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.893083643363669e-06,
+      "loss": 0.8822,
+      "step": 19539
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8900679066782167e-06,
+      "loss": 1.2585,
+      "step": 19540
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8870545510684124e-06,
+      "loss": 1.0763,
+      "step": 19541
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.884043576607375e-06,
+      "loss": 0.8566,
+      "step": 19542
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8810349833682129e-06,
+      "loss": 0.8333,
+      "step": 19543
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8780287714239232e-06,
+      "loss": 0.7571,
+      "step": 19544
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8750249408474585e-06,
+      "loss": 0.8429,
+      "step": 19545
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8720234917117495e-06,
+      "loss": 0.7592,
+      "step": 19546
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.869024424089605e-06,
+      "loss": 0.8293,
+      "step": 19547
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8660277380538326e-06,
+      "loss": 0.9681,
+      "step": 19548
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8630334336771527e-06,
+      "loss": 0.8527,
+      "step": 19549
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.86004151103224e-06,
+      "loss": 0.9056,
+      "step": 19550
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.857051970191681e-06,
+      "loss": 0.8163,
+      "step": 19551
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8540648112280735e-06,
+      "loss": 0.8893,
+      "step": 19552
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8510800342138813e-06,
+      "loss": 0.8839,
+      "step": 19553
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8480976392215577e-06,
+      "loss": 0.8118,
+      "step": 19554
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8451176263234892e-06,
+      "loss": 0.8771,
+      "step": 19555
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.842139995591985e-06,
+      "loss": 0.8971,
+      "step": 19556
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8391647470993201e-06,
+      "loss": 0.7547,
+      "step": 19557
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8361918809176815e-06,
+      "loss": 0.9227,
+      "step": 19558
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8332213971192558e-06,
+      "loss": 0.7619,
+      "step": 19559
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8302532957761077e-06,
+      "loss": 0.7904,
+      "step": 19560
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8272875769602792e-06,
+      "loss": 0.7799,
+      "step": 19561
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8243242407437467e-06,
+      "loss": 0.9368,
+      "step": 19562
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.821363287198441e-06,
+      "loss": 0.6826,
+      "step": 19563
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8184047163961937e-06,
+      "loss": 0.7088,
+      "step": 19564
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8154485284088362e-06,
+      "loss": 0.7492,
+      "step": 19565
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8124947233081002e-06,
+      "loss": 0.8463,
+      "step": 19566
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8095433011656615e-06,
+      "loss": 0.8495,
+      "step": 19567
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.806594262053174e-06,
+      "loss": 0.8504,
+      "step": 19568
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8036476060422024e-06,
+      "loss": 0.7525,
+      "step": 19569
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8007033332042456e-06,
+      "loss": 0.8856,
+      "step": 19570
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7977614436107793e-06,
+      "loss": 0.8333,
+      "step": 19571
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.794821937333191e-06,
+      "loss": 0.6977,
+      "step": 19572
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7918848144428234e-06,
+      "loss": 0.9915,
+      "step": 19573
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7889500750109422e-06,
+      "loss": 0.9283,
+      "step": 19574
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7860177191088013e-06,
+      "loss": 1.3503,
+      "step": 19575
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7830877468075658e-06,
+      "loss": 0.9383,
+      "step": 19576
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7801601581783123e-06,
+      "loss": 0.8289,
+      "step": 19577
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7772349532921285e-06,
+      "loss": 0.885,
+      "step": 19578
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7743121322199686e-06,
+      "loss": 1.1319,
+      "step": 19579
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7713916950328091e-06,
+      "loss": 1.004,
+      "step": 19580
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7684736418015046e-06,
+      "loss": 0.7914,
+      "step": 19581
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7655579725968873e-06,
+      "loss": 0.8925,
+      "step": 19582
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7626446874896895e-06,
+      "loss": 0.7934,
+      "step": 19583
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7597337865506547e-06,
+      "loss": 0.868,
+      "step": 19584
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.756825269850404e-06,
+      "loss": 0.7837,
+      "step": 19585
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7539191374595255e-06,
+      "loss": 0.7601,
+      "step": 19586
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7510153894485627e-06,
+      "loss": 1.2942,
+      "step": 19587
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7481140258879813e-06,
+      "loss": 1.4841,
+      "step": 19588
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.745215046848192e-06,
+      "loss": 1.0404,
+      "step": 19589
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7423184523995605e-06,
+      "loss": 0.9013,
+      "step": 19590
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7394242426123753e-06,
+      "loss": 0.7571,
+      "step": 19591
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.736532417556891e-06,
+      "loss": 0.9574,
+      "step": 19592
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7336429773032735e-06,
+      "loss": 1.1602,
+      "step": 19593
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7307559219216675e-06,
+      "loss": 1.0898,
+      "step": 19594
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.727871251482105e-06,
+      "loss": 0.7973,
+      "step": 19595
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7249889660546415e-06,
+      "loss": 0.7282,
+      "step": 19596
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7221090657091988e-06,
+      "loss": 0.8764,
+      "step": 19597
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7192315505156875e-06,
+      "loss": 0.7569,
+      "step": 19598
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7163564205439187e-06,
+      "loss": 0.8524,
+      "step": 19599
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.713483675863692e-06,
+      "loss": 0.8268,
+      "step": 19600
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7106133165447292e-06,
+      "loss": 0.7521,
+      "step": 19601
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7077453426566747e-06,
+      "loss": 0.8404,
+      "step": 19602
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7048797542691396e-06,
+      "loss": 1.1932,
+      "step": 19603
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7020165514516794e-06,
+      "loss": 1.0725,
+      "step": 19604
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6991557342737718e-06,
+      "loss": 0.8957,
+      "step": 19605
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6962973028048612e-06,
+      "loss": 0.7705,
+      "step": 19606
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6934412571143143e-06,
+      "loss": 0.8158,
+      "step": 19607
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6905875972714312e-06,
+      "loss": 1.1952,
+      "step": 19608
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.687736323345479e-06,
+      "loss": 1.4706,
+      "step": 19609
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6848874354056576e-06,
+      "loss": 1.5338,
+      "step": 19610
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6820409335211117e-06,
+      "loss": 0.9901,
+      "step": 19611
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6791968177609197e-06,
+      "loss": 0.8355,
+      "step": 19612
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6763550881941036e-06,
+      "loss": 0.9315,
+      "step": 19613
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6735157448896422e-06,
+      "loss": 0.7983,
+      "step": 19614
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6706787879164244e-06,
+      "loss": 0.8738,
+      "step": 19615
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6678442173433172e-06,
+      "loss": 0.9126,
+      "step": 19616
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6650120332391217e-06,
+      "loss": 0.775,
+      "step": 19617
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6621822356725492e-06,
+      "loss": 0.8754,
+      "step": 19618
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6593548247123003e-06,
+      "loss": 0.8922,
+      "step": 19619
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6565298004269868e-06,
+      "loss": 0.8329,
+      "step": 19620
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6537071628851542e-06,
+      "loss": 0.8308,
+      "step": 19621
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6508869121553472e-06,
+      "loss": 0.7755,
+      "step": 19622
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.648069048305978e-06,
+      "loss": 0.8719,
+      "step": 19623
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.645253571405425e-06,
+      "loss": 0.7379,
+      "step": 19624
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6424404815220562e-06,
+      "loss": 1.0358,
+      "step": 19625
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6396297787241166e-06,
+      "loss": 1.0911,
+      "step": 19626
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6368214630798184e-06,
+      "loss": 0.8529,
+      "step": 19627
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6340155346573405e-06,
+      "loss": 0.8232,
+      "step": 19628
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6312119935247616e-06,
+      "loss": 0.8367,
+      "step": 19629
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6284108397501385e-06,
+      "loss": 0.8778,
+      "step": 19630
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.625612073401428e-06,
+      "loss": 0.6631,
+      "step": 19631
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6228156945465757e-06,
+      "loss": 0.9139,
+      "step": 19632
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6200217032534492e-06,
+      "loss": 0.7278,
+      "step": 19633
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6172300995898392e-06,
+      "loss": 0.8782,
+      "step": 19634
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6144408836235136e-06,
+      "loss": 0.7804,
+      "step": 19635
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6116540554221625e-06,
+      "loss": 0.9502,
+      "step": 19636
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.608869615053421e-06,
+      "loss": 0.9025,
+      "step": 19637
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6060875625848571e-06,
+      "loss": 0.7983,
+      "step": 19638
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6033078980839944e-06,
+      "loss": 0.8557,
+      "step": 19639
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6005306216182902e-06,
+      "loss": 0.762,
+      "step": 19640
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5977557332551685e-06,
+      "loss": 0.8412,
+      "step": 19641
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5949832330619419e-06,
+      "loss": 0.8354,
+      "step": 19642
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5922131211059232e-06,
+      "loss": 0.7928,
+      "step": 19643
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5894453974543256e-06,
+      "loss": 0.8442,
+      "step": 19644
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5866800621743284e-06,
+      "loss": 0.7521,
+      "step": 19645
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5839171153330446e-06,
+      "loss": 0.8831,
+      "step": 19646
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5811565569975318e-06,
+      "loss": 0.7827,
+      "step": 19647
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5783983872347808e-06,
+      "loss": 0.9371,
+      "step": 19648
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5756426061117268e-06,
+      "loss": 0.7534,
+      "step": 19649
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5728892136952611e-06,
+      "loss": 0.7834,
+      "step": 19650
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5701382100522189e-06,
+      "loss": 0.7267,
+      "step": 19651
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5673895952493244e-06,
+      "loss": 0.8813,
+      "step": 19652
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5646433693533247e-06,
+      "loss": 0.8584,
+      "step": 19653
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.561899532430866e-06,
+      "loss": 0.7057,
+      "step": 19654
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5591580845485176e-06,
+      "loss": 0.7885,
+      "step": 19655
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.556419025772815e-06,
+      "loss": 0.8856,
+      "step": 19656
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5536823561702606e-06,
+      "loss": 0.8278,
+      "step": 19657
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.550948075807246e-06,
+      "loss": 0.7645,
+      "step": 19658
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5482161847501397e-06,
+      "loss": 0.7796,
+      "step": 19659
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5454866830652448e-06,
+      "loss": 0.8692,
+      "step": 19660
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.542759570818808e-06,
+      "loss": 0.8048,
+      "step": 19661
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5400348480769988e-06,
+      "loss": 0.7944,
+      "step": 19662
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5373125149059642e-06,
+      "loss": 0.6655,
+      "step": 19663
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5345925713717623e-06,
+      "loss": 0.762,
+      "step": 19664
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5318750175404185e-06,
+      "loss": 0.7661,
+      "step": 19665
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5291598534778683e-06,
+      "loss": 0.762,
+      "step": 19666
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5264470792500152e-06,
+      "loss": 0.787,
+      "step": 19667
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5237366949226838e-06,
+      "loss": 0.8219,
+      "step": 19668
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5210287005616776e-06,
+      "loss": 0.7696,
+      "step": 19669
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5183230962327099e-06,
+      "loss": 0.7812,
+      "step": 19670
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.51561988200144e-06,
+      "loss": 0.9197,
+      "step": 19671
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5129190579334707e-06,
+      "loss": 0.7859,
+      "step": 19672
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5102206240943495e-06,
+      "loss": 0.927,
+      "step": 19673
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5075245805495797e-06,
+      "loss": 0.8216,
+      "step": 19674
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.5048309273645755e-06,
+      "loss": 0.8142,
+      "step": 19675
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.502139664604718e-06,
+      "loss": 0.8371,
+      "step": 19676
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.499450792335333e-06,
+      "loss": 0.7912,
+      "step": 19677
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4967643106216457e-06,
+      "loss": 0.8277,
+      "step": 19678
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4940802195288928e-06,
+      "loss": 0.7641,
+      "step": 19679
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4913985191222112e-06,
+      "loss": 0.7625,
+      "step": 19680
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4887192094666603e-06,
+      "loss": 0.8105,
+      "step": 19681
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4860422906272763e-06,
+      "loss": 0.8119,
+      "step": 19682
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4833677626690413e-06,
+      "loss": 0.8391,
+      "step": 19683
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4806956256568361e-06,
+      "loss": 0.8246,
+      "step": 19684
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4780258796555423e-06,
+      "loss": 0.7874,
+      "step": 19685
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4753585247299306e-06,
+      "loss": 0.731,
+      "step": 19686
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.472693560944749e-06,
+      "loss": 0.7382,
+      "step": 19687
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4700309883646679e-06,
+      "loss": 0.8042,
+      "step": 19688
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4673708070543135e-06,
+      "loss": 0.7829,
+      "step": 19689
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4647130170782453e-06,
+      "loss": 1.2246,
+      "step": 19690
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.462057618500956e-06,
+      "loss": 0.9953,
+      "step": 19691
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4594046113869052e-06,
+      "loss": 0.7957,
+      "step": 19692
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.456753995800475e-06,
+      "loss": 0.8039,
+      "step": 19693
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4541057718059803e-06,
+      "loss": 0.8371,
+      "step": 19694
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4514599394677253e-06,
+      "loss": 0.7668,
+      "step": 19695
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.448816498849892e-06,
+      "loss": 0.7862,
+      "step": 19696
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4461754500166403e-06,
+      "loss": 0.8069,
+      "step": 19697
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4435367930320742e-06,
+      "loss": 0.7481,
+      "step": 19698
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.4409005279602316e-06,
+      "loss": 0.867,
+      "step": 19699
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4382666548650947e-06,
+      "loss": 0.8241,
+      "step": 19700
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4356351738105788e-06,
+      "loss": 0.7775,
+      "step": 19701
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4330060848605553e-06,
+      "loss": 0.7967,
+      "step": 19702
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4303793880788396e-06,
+      "loss": 0.7823,
+      "step": 19703
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4277550835291475e-06,
+      "loss": 0.9481,
+      "step": 19704
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.425133171275217e-06,
+      "loss": 0.7966,
+      "step": 19705
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4225136513806414e-06,
+      "loss": 0.8055,
+      "step": 19706
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4198965239090035e-06,
+      "loss": 0.7861,
+      "step": 19707
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4172817889238298e-06,
+      "loss": 0.9419,
+      "step": 19708
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4146694464885812e-06,
+      "loss": 0.8419,
+      "step": 19709
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4120594966666512e-06,
+      "loss": 0.8278,
+      "step": 19710
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4094519395213778e-06,
+      "loss": 0.8875,
+      "step": 19711
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4068467751160441e-06,
+      "loss": 0.885,
+      "step": 19712
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.404244003513877e-06,
+      "loss": 0.7394,
+      "step": 19713
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4016436247780484e-06,
+      "loss": 0.7461,
+      "step": 19714
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3990456389716743e-06,
+      "loss": 0.868,
+      "step": 19715
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3964500461577935e-06,
+      "loss": 0.8354,
+      "step": 19716
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3938568463994106e-06,
+      "loss": 0.9167,
+      "step": 19717
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3912660397594534e-06,
+      "loss": 0.7944,
+      "step": 19718
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.388677626300794e-06,
+      "loss": 0.7553,
+      "step": 19719
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3860916060862705e-06,
+      "loss": 0.6707,
+      "step": 19720
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3835079791786331e-06,
+      "loss": 0.8804,
+      "step": 19721
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3809267456405762e-06,
+      "loss": 0.7188,
+      "step": 19722
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3783479055347603e-06,
+      "loss": 0.9862,
+      "step": 19723
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.375771458923769e-06,
+      "loss": 0.6694,
+      "step": 19724
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3731974058701303e-06,
+      "loss": 0.8849,
+      "step": 19725
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3706257464363048e-06,
+      "loss": 0.8129,
+      "step": 19726
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3680564806847207e-06,
+      "loss": 0.9586,
+      "step": 19727
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.365489608677728e-06,
+      "loss": 1.2123,
+      "step": 19728
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3629251304776213e-06,
+      "loss": 1.0408,
+      "step": 19729
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3603630461466287e-06,
+      "loss": 0.7134,
+      "step": 19730
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3578033557469562e-06,
+      "loss": 0.6135,
+      "step": 19731
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3552460593406979e-06,
+      "loss": 0.8827,
+      "step": 19732
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.352691156989938e-06,
+      "loss": 0.8358,
+      "step": 19733
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3501386487566715e-06,
+      "loss": 0.8213,
+      "step": 19734
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3475885347028484e-06,
+      "loss": 0.7802,
+      "step": 19735
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3450408148903636e-06,
+      "loss": 0.9029,
+      "step": 19736
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3424954893810572e-06,
+      "loss": 0.7228,
+      "step": 19737
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.339952558236679e-06,
+      "loss": 0.7817,
+      "step": 19738
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3374120215189578e-06,
+      "loss": 0.8538,
+      "step": 19739
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3348738792895666e-06,
+      "loss": 0.6884,
+      "step": 19740
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3323381316100669e-06,
+      "loss": 0.7586,
+      "step": 19741
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.329804778542032e-06,
+      "loss": 1.2221,
+      "step": 19742
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.327273820146935e-06,
+      "loss": 1.0064,
+      "step": 19743
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.324745256486204e-06,
+      "loss": 0.857,
+      "step": 19744
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3222190876212016e-06,
+      "loss": 0.78,
+      "step": 19745
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.319695313613234e-06,
+      "loss": 0.7134,
+      "step": 19746
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3171739345235523e-06,
+      "loss": 0.6875,
+      "step": 19747
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.314654950413352e-06,
+      "loss": 0.8402,
+      "step": 19748
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3121383613437732e-06,
+      "loss": 0.7743,
+      "step": 19749
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3096241673758891e-06,
+      "loss": 0.8317,
+      "step": 19750
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3071123685707066e-06,
+      "loss": 1.231,
+      "step": 19751
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.3046029649891989e-06,
+      "loss": 1.0464,
+      "step": 19752
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.302095956692273e-06,
+      "loss": 0.7757,
+      "step": 19753
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2995913437407581e-06,
+      "loss": 0.8118,
+      "step": 19754
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2970891261954389e-06,
+      "loss": 0.7884,
+      "step": 19755
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2945893041170553e-06,
+      "loss": 0.8327,
+      "step": 19756
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2920918775662595e-06,
+      "loss": 0.8898,
+      "step": 19757
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.289596846603691e-06,
+      "loss": 0.8129,
+      "step": 19758
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.287104211289869e-06,
+      "loss": 0.7245,
+      "step": 19759
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2846139716853112e-06,
+      "loss": 0.9618,
+      "step": 19760
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2821261278504471e-06,
+      "loss": 0.8372,
+      "step": 19761
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2796406798456617e-06,
+      "loss": 0.8559,
+      "step": 19762
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2771576277312626e-06,
+      "loss": 0.7261,
+      "step": 19763
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2746769715675234e-06,
+      "loss": 0.812,
+      "step": 19764
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2721987114146406e-06,
+      "loss": 0.8159,
+      "step": 19765
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.269722847332766e-06,
+      "loss": 0.787,
+      "step": 19766
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2672493793819852e-06,
+      "loss": 0.8581,
+      "step": 19767
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2647783076223273e-06,
+      "loss": 0.8686,
+      "step": 19768
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2623096321137783e-06,
+      "loss": 0.8091,
+      "step": 19769
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2598433529162234e-06,
+      "loss": 0.8853,
+      "step": 19770
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.257379470089537e-06,
+      "loss": 0.9048,
+      "step": 19771
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2549179836935154e-06,
+      "loss": 0.8617,
+      "step": 19772
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2524588937878778e-06,
+      "loss": 0.9457,
+      "step": 19773
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.250002200432343e-06,
+      "loss": 0.6794,
+      "step": 19774
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2475479036864968e-06,
+      "loss": 0.9072,
+      "step": 19775
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2450960036099247e-06,
+      "loss": 0.804,
+      "step": 19776
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2426465002621346e-06,
+      "loss": 0.7921,
+      "step": 19777
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2401993937025569e-06,
+      "loss": 0.772,
+      "step": 19778
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2377546839905996e-06,
+      "loss": 0.7602,
+      "step": 19779
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2353123711855707e-06,
+      "loss": 0.784,
+      "step": 19780
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2328724553467785e-06,
+      "loss": 0.7236,
+      "step": 19781
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2304349365334088e-06,
+      "loss": 0.8705,
+      "step": 19782
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2279998148046257e-06,
+      "loss": 0.7473,
+      "step": 19783
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2255670902195592e-06,
+      "loss": 0.8741,
+      "step": 19784
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2231367628371959e-06,
+      "loss": 0.8737,
+      "step": 19785
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.220708832716555e-06,
+      "loss": 0.8548,
+      "step": 19786
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.218283299916556e-06,
+      "loss": 0.8039,
+      "step": 19787
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2158601644960521e-06,
+      "loss": 0.7265,
+      "step": 19788
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2134394265138627e-06,
+      "loss": 0.8545,
+      "step": 19789
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2110210860287409e-06,
+      "loss": 0.854,
+      "step": 19790
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.208605143099384e-06,
+      "loss": 0.8513,
+      "step": 19791
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2061915977843897e-06,
+      "loss": 0.8314,
+      "step": 19792
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.2037804501423778e-06,
+      "loss": 0.8319,
+      "step": 19793
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.201371700231846e-06,
+      "loss": 0.816,
+      "step": 19794
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1989653481112584e-06,
+      "loss": 0.7977,
+      "step": 19795
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1965613938389907e-06,
+      "loss": 0.7308,
+      "step": 19796
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1941598374734298e-06,
+      "loss": 0.8664,
+      "step": 19797
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1917606790728287e-06,
+      "loss": 0.7553,
+      "step": 19798
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1893639186954187e-06,
+      "loss": 0.7821,
+      "step": 19799
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1869695563993755e-06,
+      "loss": 0.9175,
+      "step": 19800
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1845775922428082e-06,
+      "loss": 0.8597,
+      "step": 19801
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.182188026283748e-06,
+      "loss": 0.8353,
+      "step": 19802
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.1798008585802155e-06,
+      "loss": 0.6978,
+      "step": 19803
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.177416089190142e-06,
+      "loss": 0.9495,
+      "step": 19804
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1750337181713921e-06,
+      "loss": 0.7658,
+      "step": 19805
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1726537455817865e-06,
+      "loss": 0.7885,
+      "step": 19806
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.170276171479101e-06,
+      "loss": 0.8799,
+      "step": 19807
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.167900995921012e-06,
+      "loss": 0.7548,
+      "step": 19808
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.165528218965195e-06,
+      "loss": 0.8512,
+      "step": 19809
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1631578406692156e-06,
+      "loss": 0.796,
+      "step": 19810
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1607898610906053e-06,
+      "loss": 0.8364,
+      "step": 19811
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1584242802868295e-06,
+      "loss": 0.7812,
+      "step": 19812
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1560610983153198e-06,
+      "loss": 0.7909,
+      "step": 19813
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.153700315233397e-06,
+      "loss": 0.8616,
+      "step": 19814
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1513419310983819e-06,
+      "loss": 0.7931,
+      "step": 19815
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1489859459675067e-06,
+      "loss": 0.8504,
+      "step": 19816
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1466323598979478e-06,
+      "loss": 0.8353,
+      "step": 19817
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.144281172946804e-06,
+      "loss": 0.7356,
+      "step": 19818
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1419323851711738e-06,
+      "loss": 0.8367,
+      "step": 19819
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1395859966280454e-06,
+      "loss": 0.7082,
+      "step": 19820
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1372420073743506e-06,
+      "loss": 0.8555,
+      "step": 19821
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1349004174669886e-06,
+      "loss": 0.7508,
+      "step": 19822
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1325612269628028e-06,
+      "loss": 0.9633,
+      "step": 19823
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1302244359185365e-06,
+      "loss": 0.694,
+      "step": 19824
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1278900443909224e-06,
+      "loss": 0.729,
+      "step": 19825
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.125558052436615e-06,
+      "loss": 0.7183,
+      "step": 19826
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1232284601122022e-06,
+      "loss": 0.7431,
+      "step": 19827
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.120901267474228e-06,
+      "loss": 0.7733,
+      "step": 19828
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1185764745791583e-06,
+      "loss": 0.7481,
+      "step": 19829
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1162540814834254e-06,
+      "loss": 0.8151,
+      "step": 19830
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1139340882433958e-06,
+      "loss": 0.7335,
+      "step": 19831
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1116164949153684e-06,
+      "loss": 0.6653,
+      "step": 19832
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1093013015555876e-06,
+      "loss": 0.7934,
+      "step": 19833
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1069885082202525e-06,
+      "loss": 0.7485,
+      "step": 19834
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.104678114965485e-06,
+      "loss": 0.8389,
+      "step": 19835
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1023701218473626e-06,
+      "loss": 0.8559,
+      "step": 19836
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.1000645289218848e-06,
+      "loss": 0.8277,
+      "step": 19837
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0977613362450289e-06,
+      "loss": 0.8087,
+      "step": 19838
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.095460543872684e-06,
+      "loss": 0.6676,
+      "step": 19839
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0931621518606717e-06,
+      "loss": 0.9617,
+      "step": 19840
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.090866160264803e-06,
+      "loss": 0.7668,
+      "step": 19841
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0885725691407777e-06,
+      "loss": 0.7805,
+      "step": 19842
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0862813785442628e-06,
+      "loss": 1.2399,
+      "step": 19843
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.08399258853088e-06,
+      "loss": 0.9107,
+      "step": 19844
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0817061991561627e-06,
+      "loss": 0.8266,
+      "step": 19845
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0794222104755892e-06,
+      "loss": 0.7257,
+      "step": 19846
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0771406225446147e-06,
+      "loss": 0.8181,
+      "step": 19847
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.074861435418606e-06,
+      "loss": 0.8322,
+      "step": 19848
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0725846491528634e-06,
+      "loss": 0.8997,
+      "step": 19849
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0703102638026652e-06,
+      "loss": 0.8467,
+      "step": 19850
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0680382794232001e-06,
+      "loss": 0.8109,
+      "step": 19851
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0657686960696024e-06,
+      "loss": 1.1567,
+      "step": 19852
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0635015137969495e-06,
+      "loss": 1.1018,
+      "step": 19853
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0612367326602757e-06,
+      "loss": 0.7912,
+      "step": 19854
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.058974352714548e-06,
+      "loss": 0.9178,
+      "step": 19855
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0567143740146556e-06,
+      "loss": 0.6983,
+      "step": 19856
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0544567966154772e-06,
+      "loss": 0.8845,
+      "step": 19857
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0522016205717799e-06,
+      "loss": 0.74,
+      "step": 19858
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0499488459382867e-06,
+      "loss": 1.1517,
+      "step": 19859
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.047698472769698e-06,
+      "loss": 1.034,
+      "step": 19860
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.045450501120615e-06,
+      "loss": 0.8609,
+      "step": 19861
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.043204931045594e-06,
+      "loss": 0.7736,
+      "step": 19862
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0409617625991353e-06,
+      "loss": 0.7636,
+      "step": 19863
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0387209958356848e-06,
+      "loss": 0.9518,
+      "step": 19864
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0364826308096098e-06,
+      "loss": 0.7821,
+      "step": 19865
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.034246667575256e-06,
+      "loss": 0.7451,
+      "step": 19866
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0320131061868687e-06,
+      "loss": 1.1235,
+      "step": 19867
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0297819466986713e-06,
+      "loss": 0.9579,
+      "step": 19868
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0275531891647983e-06,
+      "loss": 0.8118,
+      "step": 19869
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0253268336393506e-06,
+      "loss": 0.6293,
+      "step": 19870
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0231028801763631e-06,
+      "loss": 0.806,
+      "step": 19871
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0208813288298035e-06,
+      "loss": 0.898,
+      "step": 19872
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0186621796535956e-06,
+      "loss": 0.7717,
+      "step": 19873
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0164454327015739e-06,
+      "loss": 0.718,
+      "step": 19874
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.014231088027562e-06,
+      "loss": 0.675,
+      "step": 19875
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0120191456852944e-06,
+      "loss": 0.816,
+      "step": 19876
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0098096057284512e-06,
+      "loss": 0.8452,
+      "step": 19877
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0076024682106443e-06,
+      "loss": 0.8185,
+      "step": 19878
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0053977331854757e-06,
+      "loss": 0.841,
+      "step": 19879
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0031954007064136e-06,
+      "loss": 1.1512,
+      "step": 19880
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0009954708269264e-06,
+      "loss": 0.9448,
+      "step": 19881
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.987979436004158e-07,
+      "loss": 0.8551,
+      "step": 19882
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.966028190801946e-07,
+      "loss": 0.8447,
+      "step": 19883
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.944100973195424e-07,
+      "loss": 0.9515,
+      "step": 19884
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.922197783716835e-07,
+      "loss": 0.8879,
+      "step": 19885
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.900318622897643e-07,
+      "loss": 0.8555,
+      "step": 19886
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.878463491268974e-07,
+      "loss": 1.2261,
+      "step": 19887
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.856632389361076e-07,
+      "loss": 0.9871,
+      "step": 19888
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.834825317703966e-07,
+      "loss": 0.7427,
+      "step": 19889
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.813042276826779e-07,
+      "loss": 0.7995,
+      "step": 19890
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.79128326725809e-07,
+      "loss": 0.813,
+      "step": 19891
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.769548289526143e-07,
+      "loss": 0.8544,
+      "step": 19892
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.747837344158405e-07,
+      "loss": 0.861,
+      "step": 19893
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.72615043168168e-07,
+      "loss": 0.7665,
+      "step": 19894
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.704487552622433e-07,
+      "loss": 0.8273,
+      "step": 19895
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.682848707506354e-07,
+      "loss": 0.7621,
+      "step": 19896
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.66123389685858e-07,
+      "loss": 0.7759,
+      "step": 19897
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.63964312120369e-07,
+      "loss": 0.7787,
+      "step": 19898
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.61807638106571e-07,
+      "loss": 0.7909,
+      "step": 19899
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.59653367696811e-07,
+      "loss": 0.8042,
+      "step": 19900
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.575015009433697e-07,
+      "loss": 0.7754,
+      "step": 19901
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.553520378984603e-07,
+      "loss": 0.8215,
+      "step": 19902
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.532049786142527e-07,
+      "loss": 0.7477,
+      "step": 19903
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.510603231428605e-07,
+      "loss": 0.8068,
+      "step": 19904
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.489180715363422e-07,
+      "loss": 0.7437,
+      "step": 19905
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.467782238466671e-07,
+      "loss": 0.6948,
+      "step": 19906
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.446407801257828e-07,
+      "loss": 0.9228,
+      "step": 19907
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.425057404255588e-07,
+      "loss": 0.8527,
+      "step": 19908
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.403731047978204e-07,
+      "loss": 0.8234,
+      "step": 19909
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.382428732943149e-07,
+      "loss": 0.7611,
+      "step": 19910
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.361150459667456e-07,
+      "loss": 0.7846,
+      "step": 19911
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.339896228667488e-07,
+      "loss": 0.8366,
+      "step": 19912
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.318666040459167e-07,
+      "loss": 0.8534,
+      "step": 19913
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.297459895557748e-07,
+      "loss": 0.7593,
+      "step": 19914
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.276277794477816e-07,
+      "loss": 0.834,
+      "step": 19915
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.255119737733409e-07,
+      "loss": 0.9352,
+      "step": 19916
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.233985725838112e-07,
+      "loss": 0.8761,
+      "step": 19917
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.212875759304851e-07,
+      "loss": 0.9053,
+      "step": 19918
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.19178983864577e-07,
+      "loss": 0.8377,
+      "step": 19919
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.170727964372905e-07,
+      "loss": 0.7889,
+      "step": 19920
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.149690136997291e-07,
+      "loss": 0.7037,
+      "step": 19921
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.128676357029298e-07,
+      "loss": 0.7825,
+      "step": 19922
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.107686624979294e-07,
+      "loss": 0.7246,
+      "step": 19923
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.08672094135643e-07,
+      "loss": 0.8375,
+      "step": 19924
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.06577930666952e-07,
+      "loss": 0.7635,
+      "step": 19925
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.044861721426934e-07,
+      "loss": 0.8015,
+      "step": 19926
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.023968186136267e-07,
+      "loss": 1.1308,
+      "step": 19927
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.003098701304669e-07,
+      "loss": 0.9569,
+      "step": 19928
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.982253267438512e-07,
+      "loss": 0.5946,
+      "step": 19929
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.961431885043836e-07,
+      "loss": 0.7319,
+      "step": 19930
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.940634554625904e-07,
+      "loss": 0.7971,
+      "step": 19931
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.919861276689312e-07,
+      "loss": 0.816,
+      "step": 19932
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.899112051738434e-07,
+      "loss": 0.7968,
+      "step": 19933
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.878386880276757e-07,
+      "loss": 0.7575,
+      "step": 19934
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.8576857628071e-07,
+      "loss": 0.7893,
+      "step": 19935
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.837008699832283e-07,
+      "loss": 0.8409,
+      "step": 19936
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.816355691853684e-07,
+      "loss": 0.8036,
+      "step": 19937
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.795726739372789e-07,
+      "loss": 0.7549,
+      "step": 19938
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.77512184289031e-07,
+      "loss": 0.8608,
+      "step": 19939
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.75454100290618e-07,
+      "loss": 0.8923,
+      "step": 19940
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.733984219919777e-07,
+      "loss": 0.8575,
+      "step": 19941
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.713451494430258e-07,
+      "loss": 0.779,
+      "step": 19942
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.692942826935779e-07,
+      "loss": 0.847,
+      "step": 19943
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.672458217934165e-07,
+      "loss": 0.8633,
+      "step": 19944
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.651997667922462e-07,
+      "loss": 0.772,
+      "step": 19945
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.631561177397274e-07,
+      "loss": 0.8654,
+      "step": 19946
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.611148746854647e-07,
+      "loss": 0.731,
+      "step": 19947
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.590760376789964e-07,
+      "loss": 0.9143,
+      "step": 19948
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.570396067697939e-07,
+      "loss": 0.8456,
+      "step": 19949
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.550055820072955e-07,
+      "loss": 0.6572,
+      "step": 19950
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.529739634408396e-07,
+      "loss": 0.8049,
+      "step": 19951
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.509447511197643e-07,
+      "loss": 1.0562,
+      "step": 19952
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.489179450933082e-07,
+      "loss": 1.507,
+      "step": 19953
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.46893545410643e-07,
+      "loss": 0.9933,
+      "step": 19954
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.448715521209294e-07,
+      "loss": 0.8429,
+      "step": 19955
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.42851965273217e-07,
+      "loss": 0.7544,
+      "step": 19956
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.408347849165332e-07,
+      "loss": 0.793,
+      "step": 19957
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.388200110998168e-07,
+      "loss": 0.6918,
+      "step": 19958
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.368076438719729e-07,
+      "loss": 0.705,
+      "step": 19959
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.347976832818627e-07,
+      "loss": 0.8083,
+      "step": 19960
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.327901293782359e-07,
+      "loss": 0.7287,
+      "step": 19961
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.307849822098312e-07,
+      "loss": 0.9009,
+      "step": 19962
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.287822418252989e-07,
+      "loss": 0.8425,
+      "step": 19963
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.267819082732664e-07,
+      "loss": 0.813,
+      "step": 19964
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.247839816022618e-07,
+      "loss": 0.8039,
+      "step": 19965
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.227884618607795e-07,
+      "loss": 0.7536,
+      "step": 19966
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.207953490972475e-07,
+      "loss": 0.8596,
+      "step": 19967
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.188046433600382e-07,
+      "loss": 0.7678,
+      "step": 19968
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.168163446974797e-07,
+      "loss": 0.7873,
+      "step": 19969
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.148304531578e-07,
+      "loss": 0.7845,
+      "step": 19970
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.12846968789216e-07,
+      "loss": 0.7975,
+      "step": 19971
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.108658916398671e-07,
+      "loss": 0.7945,
+      "step": 19972
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.088872217578258e-07,
+      "loss": 0.7192,
+      "step": 19973
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.069109591910984e-07,
+      "loss": 0.7523,
+      "step": 19974
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.049371039876796e-07,
+      "loss": 0.7568,
+      "step": 19975
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.029656561954424e-07,
+      "loss": 0.7893,
+      "step": 19976
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.009966158622595e-07,
+      "loss": 0.8256,
+      "step": 19977
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.990299830358927e-07,
+      "loss": 1.2011,
+      "step": 19978
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.970657577640928e-07,
+      "loss": 1.065,
+      "step": 19979
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.951039400945215e-07,
+      "loss": 0.794,
+      "step": 19980
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.931445300747964e-07,
+      "loss": 0.8695,
+      "step": 19981
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.911875277524572e-07,
+      "loss": 0.8592,
+      "step": 19982
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.892329331750104e-07,
+      "loss": 0.9866,
+      "step": 19983
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.872807463898957e-07,
+      "loss": 0.7081,
+      "step": 19984
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.853309674444864e-07,
+      "loss": 0.8042,
+      "step": 19985
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.833835963861003e-07,
+      "loss": 0.6321,
+      "step": 19986
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.814386332619994e-07,
+      "loss": 0.9204,
+      "step": 19987
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.794960781194017e-07,
+      "loss": 0.8357,
+      "step": 19988
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.775559310054359e-07,
+      "loss": 0.9085,
+      "step": 19989
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.756181919671867e-07,
+      "loss": 0.7139,
+      "step": 19990
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.736828610516944e-07,
+      "loss": 0.7729,
+      "step": 19991
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.717499383059212e-07,
+      "loss": 0.9311,
+      "step": 19992
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.698194237767742e-07,
+      "loss": 0.8401,
+      "step": 19993
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.678913175111157e-07,
+      "loss": 0.7699,
+      "step": 19994
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.659656195557196e-07,
+      "loss": 0.8328,
+      "step": 19995
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.640423299573596e-07,
+      "loss": 0.8158,
+      "step": 19996
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.62121448762676e-07,
+      "loss": 0.8373,
+      "step": 19997
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.602029760183094e-07,
+      "loss": 0.7958,
+      "step": 19998
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.582869117708003e-07,
+      "loss": 0.7606,
+      "step": 19999
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.563732560666669e-07,
+      "loss": 0.7742,
+      "step": 20000
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.544620089523502e-07,
+      "loss": 0.9922,
+      "step": 20001
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.52553170474235e-07,
+      "loss": 0.7359,
+      "step": 20002
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.506467406786399e-07,
+      "loss": 0.9214,
+      "step": 20003
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.487427196118502e-07,
+      "loss": 0.689,
+      "step": 20004
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.46841107320051e-07,
+      "loss": 0.7812,
+      "step": 20005
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.449419038494165e-07,
+      "loss": 0.8254,
+      "step": 20006
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.430451092460211e-07,
+      "loss": 0.8548,
+      "step": 20007
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.411507235559168e-07,
+      "loss": 0.8492,
+      "step": 20008
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.392587468250667e-07,
+      "loss": 0.879,
+      "step": 20009
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.373691790993897e-07,
+      "loss": 0.8621,
+      "step": 20010
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.354820204247381e-07,
+      "loss": 0.7111,
+      "step": 20011
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.335972708469308e-07,
+      "loss": 0.8536,
+      "step": 20012
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.317149304116977e-07,
+      "loss": 0.8789,
+      "step": 20013
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.298349991647135e-07,
+      "loss": 0.8742,
+      "step": 20014
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.279574771516195e-07,
+      "loss": 0.9223,
+      "step": 20015
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.260823644179793e-07,
+      "loss": 0.846,
+      "step": 20016
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.242096610093007e-07,
+      "loss": 0.753,
+      "step": 20017
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.223393669710255e-07,
+      "loss": 0.9099,
+      "step": 20018
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.204714823485504e-07,
+      "loss": 0.809,
+      "step": 20019
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.186060071872058e-07,
+      "loss": 1.3423,
+      "step": 20020
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.167429415322668e-07,
+      "loss": 1.1368,
+      "step": 20021
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.148822854289417e-07,
+      "loss": 0.9489,
+      "step": 20022
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.130240389224052e-07,
+      "loss": 0.8516,
+      "step": 20023
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.111682020577437e-07,
+      "loss": 0.777,
+      "step": 20024
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.093147748799989e-07,
+      "loss": 0.8702,
+      "step": 20025
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.074637574341458e-07,
+      "loss": 0.9039,
+      "step": 20026
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.056151497651264e-07,
+      "loss": 1.2221,
+      "step": 20027
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.037689519177826e-07,
+      "loss": 1.0508,
+      "step": 20028
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.019251639369451e-07,
+      "loss": 0.7985,
+      "step": 20029
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.000837858673337e-07,
+      "loss": 0.8246,
+      "step": 20030
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.982448177536571e-07,
+      "loss": 0.8817,
+      "step": 20031
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.964082596405463e-07,
+      "loss": 1.2153,
+      "step": 20032
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.945741115725657e-07,
+      "loss": 1.0335,
+      "step": 20033
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.927423735942351e-07,
+      "loss": 0.8189,
+      "step": 20034
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.909130457500079e-07,
+      "loss": 0.8056,
+      "step": 20035
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.890861280842709e-07,
+      "loss": 0.8869,
+      "step": 20036
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.872616206413662e-07,
+      "loss": 0.805,
+      "step": 20037
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.854395234655919e-07,
+      "loss": 0.852,
+      "step": 20038
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.83619836601157e-07,
+      "loss": 0.7463,
+      "step": 20039
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.818025600922152e-07,
+      "loss": 0.8088,
+      "step": 20040
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.799876939828753e-07,
+      "loss": 0.8266,
+      "step": 20041
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.781752383172024e-07,
+      "loss": 0.7159,
+      "step": 20042
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.763651931391502e-07,
+      "loss": 0.8711,
+      "step": 20043
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.745575584926722e-07,
+      "loss": 0.7706,
+      "step": 20044
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.727523344216447e-07,
+      "loss": 0.8787,
+      "step": 20045
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.709495209698546e-07,
+      "loss": 0.8611,
+      "step": 20046
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.691491181810672e-07,
+      "loss": 0.8801,
+      "step": 20047
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.673511260989807e-07,
+      "loss": 0.8132,
+      "step": 20048
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.655555447672268e-07,
+      "loss": 0.8621,
+      "step": 20049
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.637623742293819e-07,
+      "loss": 0.8101,
+      "step": 20050
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.619716145289778e-07,
+      "loss": 0.7427,
+      "step": 20051
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.601832657094575e-07,
+      "loss": 0.8203,
+      "step": 20052
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.583973278142308e-07,
+      "loss": 0.5854,
+      "step": 20053
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.566138008866518e-07,
+      "loss": 0.9481,
+      "step": 20054
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.54832684969986e-07,
+      "loss": 0.7663,
+      "step": 20055
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.530539801074764e-07,
+      "loss": 0.9826,
+      "step": 20056
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.512776863422776e-07,
+      "loss": 0.7706,
+      "step": 20057
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.495038037175105e-07,
+      "loss": 0.8311,
+      "step": 20058
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.477323322762185e-07,
+      "loss": 1.2576,
+      "step": 20059
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.459632720614117e-07,
+      "loss": 1.1878,
+      "step": 20060
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.441966231160001e-07,
+      "loss": 0.8089,
+      "step": 20061
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.424323854828718e-07,
+      "loss": 0.8113,
+      "step": 20062
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.406705592048478e-07,
+      "loss": 0.8329,
+      "step": 20063
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.38911144324672e-07,
+      "loss": 0.7665,
+      "step": 20064
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.371541408850546e-07,
+      "loss": 0.7482,
+      "step": 20065
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.353995489286391e-07,
+      "loss": 0.7539,
+      "step": 20066
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.336473684980138e-07,
+      "loss": 0.7744,
+      "step": 20067
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.318975996356779e-07,
+      "loss": 0.864,
+      "step": 20068
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.30150242384131e-07,
+      "loss": 0.7658,
+      "step": 20069
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.284052967857501e-07,
+      "loss": 0.7547,
+      "step": 20070
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.266627628829125e-07,
+      "loss": 0.8471,
+      "step": 20071
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.249226407178732e-07,
+      "loss": 0.8845,
+      "step": 20072
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.231849303328984e-07,
+      "loss": 0.7132,
+      "step": 20073
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.214496317701435e-07,
+      "loss": 0.7661,
+      "step": 20074
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.197167450717301e-07,
+      "loss": 0.8246,
+      "step": 20075
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.179862702797135e-07,
+      "loss": 1.0765,
+      "step": 20076
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.162582074360934e-07,
+      "loss": 1.1212,
+      "step": 20077
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.14532556582792e-07,
+      "loss": 0.8605,
+      "step": 20078
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.1280931776172e-07,
+      "loss": 0.8445,
+      "step": 20079
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.110884910146664e-07,
+      "loss": 0.7855,
+      "step": 20080
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.093700763834199e-07,
+      "loss": 0.7899,
+      "step": 20081
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.076540739096803e-07,
+      "loss": 1.1326,
+      "step": 20082
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.059404836350813e-07,
+      "loss": 1.0247,
+      "step": 20083
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.042293056012227e-07,
+      "loss": 0.9328,
+      "step": 20084
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.02520539849627e-07,
+      "loss": 0.7404,
+      "step": 20085
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.008141864217832e-07,
+      "loss": 0.9459,
+      "step": 20086
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.991102453590803e-07,
+      "loss": 0.7424,
+      "step": 20087
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.974087167028741e-07,
+      "loss": 0.7244,
+      "step": 20088
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.957096004944762e-07,
+      "loss": 0.8812,
+      "step": 20089
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.940128967751091e-07,
+      "loss": 0.8654,
+      "step": 20090
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.92318605585962e-07,
+      "loss": 0.7474,
+      "step": 20091
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.906267269681465e-07,
+      "loss": 0.8506,
+      "step": 20092
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.889372609627298e-07,
+      "loss": 0.7473,
+      "step": 20093
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.872502076107012e-07,
+      "loss": 0.8043,
+      "step": 20094
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.855655669530169e-07,
+      "loss": 0.8146,
+      "step": 20095
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.838833390305554e-07,
+      "loss": 1.2766,
+      "step": 20096
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.822035238841505e-07,
+      "loss": 1.1,
+      "step": 20097
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.805261215545588e-07,
+      "loss": 0.8481,
+      "step": 20098
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.788511320825141e-07,
+      "loss": 0.921,
+      "step": 20099
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.771785555086284e-07,
+      "loss": 1.2166,
+      "step": 20100
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.75508391873536e-07,
+      "loss": 0.9623,
+      "step": 20101
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.738406412177488e-07,
+      "loss": 0.8468,
+      "step": 20102
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.721753035817346e-07,
+      "loss": 0.8062,
+      "step": 20103
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.705123790059274e-07,
+      "loss": 0.8486,
+      "step": 20104
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.688518675306842e-07,
+      "loss": 0.8424,
+      "step": 20105
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.671937691962947e-07,
+      "loss": 0.8394,
+      "step": 20106
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.655380840430047e-07,
+      "loss": 0.7699,
+      "step": 20107
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.638848121110041e-07,
+      "loss": 0.8065,
+      "step": 20108
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.622339534404053e-07,
+      "loss": 0.8472,
+      "step": 20109
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.605855080712763e-07,
+      "loss": 0.8642,
+      "step": 20110
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.589394760436184e-07,
+      "loss": 0.9374,
+      "step": 20111
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.572958573973996e-07,
+      "loss": 0.8721,
+      "step": 20112
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.556546521724882e-07,
+      "loss": 0.8461,
+      "step": 20113
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.540158604087298e-07,
+      "loss": 0.7001,
+      "step": 20114
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.523794821458927e-07,
+      "loss": 0.7527,
+      "step": 20115
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.507455174236897e-07,
+      "loss": 0.6847,
+      "step": 20116
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.491139662817779e-07,
+      "loss": 0.7116,
+      "step": 20117
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.474848287597479e-07,
+      "loss": 0.623,
+      "step": 20118
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.458581048971456e-07,
+      "loss": 0.8277,
+      "step": 20119
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.442337947334508e-07,
+      "loss": 0.8927,
+      "step": 20120
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.426118983080874e-07,
+      "loss": 0.7246,
+      "step": 20121
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.409924156604018e-07,
+      "loss": 1.144,
+      "step": 20122
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.39375346829718e-07,
+      "loss": 1.49,
+      "step": 20123
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.377606918552603e-07,
+      "loss": 1.0764,
+      "step": 20124
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.361484507762305e-07,
+      "loss": 0.7043,
+      "step": 20125
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.34538623631764e-07,
+      "loss": 0.8518,
+      "step": 20126
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.329312104609074e-07,
+      "loss": 0.7927,
+      "step": 20127
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.313262113026962e-07,
+      "loss": 0.8454,
+      "step": 20128
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.297236261960658e-07,
+      "loss": 0.8134,
+      "step": 20129
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.281234551799075e-07,
+      "loss": 0.666,
+      "step": 20130
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.265256982930788e-07,
+      "loss": 0.7019,
+      "step": 20131
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.249303555743268e-07,
+      "loss": 0.804,
+      "step": 20132
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.233374270623981e-07,
+      "loss": 0.9144,
+      "step": 20133
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.217469127959395e-07,
+      "loss": 0.8597,
+      "step": 20134
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.201588128135426e-07,
+      "loss": 0.7395,
+      "step": 20135
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.185731271537653e-07,
+      "loss": 0.7424,
+      "step": 20136
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.169898558550767e-07,
+      "loss": 0.8574,
+      "step": 20137
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.15408998955913e-07,
+      "loss": 0.9464,
+      "step": 20138
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.138305564946433e-07,
+      "loss": 1.0861,
+      "step": 20139
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.122545285095703e-07,
+      "loss": 1.1089,
+      "step": 20140
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.106809150389302e-07,
+      "loss": 0.8703,
+      "step": 20141
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.091097161209368e-07,
+      "loss": 0.76,
+      "step": 20142
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.075409317937152e-07,
+      "loss": 1.2419,
+      "step": 20143
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.059745620953349e-07,
+      "loss": 0.9874,
+      "step": 20144
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.0441060706381e-07,
+      "loss": 0.7634,
+      "step": 20145
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.028490667370989e-07,
+      "loss": 0.7538,
+      "step": 20146
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.012899411530936e-07,
+      "loss": 0.8382,
+      "step": 20147
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.997332303496415e-07,
+      "loss": 0.9497,
+      "step": 20148
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.981789343645238e-07,
+      "loss": 0.9008,
+      "step": 20149
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.966270532354544e-07,
+      "loss": 0.9658,
+      "step": 20150
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.950775870001034e-07,
+      "loss": 0.7551,
+      "step": 20151
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.93530535696074e-07,
+      "loss": 0.915,
+      "step": 20152
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.919858993609139e-07,
+      "loss": 0.8108,
+      "step": 20153
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.904436780321042e-07,
+      "loss": 0.7358,
+      "step": 20154
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.889038717470817e-07,
+      "loss": 0.7903,
+      "step": 20155
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.873664805432055e-07,
+      "loss": 1.2713,
+      "step": 20156
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.85831504457801e-07,
+      "loss": 1.1855,
+      "step": 20157
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.842989435281054e-07,
+      "loss": 0.8692,
+      "step": 20158
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.827687977913331e-07,
+      "loss": 0.9191,
+      "step": 20159
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.812410672846102e-07,
+      "loss": 0.8885,
+      "step": 20160
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.797157520450069e-07,
+      "loss": 0.8039,
+      "step": 20161
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.781928521095491e-07,
+      "loss": 0.8557,
+      "step": 20162
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.7667236751520737e-07,
+      "loss": 0.7827,
+      "step": 20163
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.7515429829885217e-07,
+      "loss": 0.7223,
+      "step": 20164
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.73638644497354e-07,
+      "loss": 0.7675,
+      "step": 20165
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.721254061474945e-07,
+      "loss": 0.7669,
+      "step": 20166
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.706145832859887e-07,
+      "loss": 1.2739,
+      "step": 20167
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.6910617594950746e-07,
+      "loss": 1.0111,
+      "step": 20168
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.676001841746658e-07,
+      "loss": 0.7857,
+      "step": 20169
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.6609660799799005e-07,
+      "loss": 0.9053,
+      "step": 20170
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.6459544745600656e-07,
+      "loss": 0.8636,
+      "step": 20171
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.630967025851196e-07,
+      "loss": 0.8555,
+      "step": 20172
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.6160037342171116e-07,
+      "loss": 0.6713,
+      "step": 20173
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.6010646000209654e-07,
+      "loss": 0.8314,
+      "step": 20174
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.586149623625469e-07,
+      "loss": 0.8582,
+      "step": 20175
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.571258805392331e-07,
+      "loss": 0.7224,
+      "step": 20176
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5563921456831525e-07,
+      "loss": 0.8847,
+      "step": 20177
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.541549644858645e-07,
+      "loss": 0.7248,
+      "step": 20178
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5267313032789637e-07,
+      "loss": 0.7654,
+      "step": 20179
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.511937121303933e-07,
+      "loss": 0.8403,
+      "step": 20180
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.497167099292487e-07,
+      "loss": 0.7821,
+      "step": 20181
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.482421237603007e-07,
+      "loss": 0.9112,
+      "step": 20182
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.467699536593539e-07,
+      "loss": 0.8347,
+      "step": 20183
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.453001996621242e-07,
+      "loss": 0.8772,
+      "step": 20184
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.438328618042942e-07,
+      "loss": 0.8737,
+      "step": 20185
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.4236794012145754e-07,
+      "loss": 0.8189,
+      "step": 20186
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.409054346491748e-07,
+      "loss": 0.7733,
+      "step": 20187
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.394453454229508e-07,
+      "loss": 0.7385,
+      "step": 20188
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.379876724782128e-07,
+      "loss": 0.8841,
+      "step": 20189
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.3653241585033254e-07,
+      "loss": 0.839,
+      "step": 20190
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.350795755746373e-07,
+      "loss": 0.8536,
+      "step": 20191
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.336291516863766e-07,
+      "loss": 0.7655,
+      "step": 20192
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.3218114422075576e-07,
+      "loss": 0.8078,
+      "step": 20193
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.3073555321292425e-07,
+      "loss": 0.8157,
+      "step": 20194
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.2929237869796526e-07,
+      "loss": 0.7804,
+      "step": 20195
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.278516207108951e-07,
+      "loss": 0.8135,
+      "step": 20196
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.264132792866859e-07,
+      "loss": 0.776,
+      "step": 20197
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.2497735446023204e-07,
+      "loss": 0.7946,
+      "step": 20198
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.2354384626640544e-07,
+      "loss": 0.9328,
+      "step": 20199
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.2211275473998947e-07,
+      "loss": 0.8282,
+      "step": 20200
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.206840799157008e-07,
+      "loss": 0.8622,
+      "step": 20201
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.1925782182823394e-07,
+      "loss": 0.8543,
+      "step": 20202
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.1783398051218337e-07,
+      "loss": 0.8022,
+      "step": 20203
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.1641255600212146e-07,
+      "loss": 0.8272,
+      "step": 20204
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.149935483325318e-07,
+      "loss": 0.7913,
+      "step": 20205
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.135769575378645e-07,
+      "loss": 0.8529,
+      "step": 20206
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.1216278365250324e-07,
+      "loss": 1.2039,
+      "step": 20207
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.107510267107428e-07,
+      "loss": 1.1646,
+      "step": 20208
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.09341686746878e-07,
+      "loss": 0.7932,
+      "step": 20209
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.0793476379510363e-07,
+      "loss": 0.7739,
+      "step": 20210
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.06530257889548e-07,
+      "loss": 0.9141,
+      "step": 20211
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.0512816906430606e-07,
+      "loss": 0.8835,
+      "step": 20212
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.037284973534172e-07,
+      "loss": 0.8453,
+      "step": 20213
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.023312427908432e-07,
+      "loss": 0.8123,
+      "step": 20214
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.0093640541049024e-07,
+      "loss": 0.761,
+      "step": 20215
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.995439852462202e-07,
+      "loss": 0.9142,
+      "step": 20216
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.98153982331817e-07,
+      "loss": 1.1591,
+      "step": 20217
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.9676639670102043e-07,
+      "loss": 1.3358,
+      "step": 20218
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.953812283875036e-07,
+      "loss": 1.0099,
+      "step": 20219
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.9399847742488394e-07,
+      "loss": 0.8418,
+      "step": 20220
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.926181438467236e-07,
+      "loss": 0.9449,
+      "step": 20221
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.912402276865179e-07,
+      "loss": 0.6585,
+      "step": 20222
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.8986472897770686e-07,
+      "loss": 0.8636,
+      "step": 20223
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.8849164775368594e-07,
+      "loss": 0.9088,
+      "step": 20224
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.8712098404776187e-07,
+      "loss": 0.7621,
+      "step": 20225
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.8575273789321906e-07,
+      "loss": 0.9153,
+      "step": 20226
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.8438690932324215e-07,
+      "loss": 0.8999,
+      "step": 20227
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.830234983709935e-07,
+      "loss": 0.7823,
+      "step": 20228
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.816625050695577e-07,
+      "loss": 0.7539,
+      "step": 20229
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.803039294519639e-07,
+      "loss": 0.856,
+      "step": 20230
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.789477715511858e-07,
+      "loss": 0.9068,
+      "step": 20231
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.775940314001414e-07,
+      "loss": 0.798,
+      "step": 20232
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7624270903168223e-07,
+      "loss": 0.8666,
+      "step": 20233
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7489380447859323e-07,
+      "loss": 0.7826,
+      "step": 20234
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7354731777362596e-07,
+      "loss": 0.9339,
+      "step": 20235
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7220324894946537e-07,
+      "loss": 0.8084,
+      "step": 20236
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7086159803870755e-07,
+      "loss": 0.8088,
+      "step": 20237
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.695223650739266e-07,
+      "loss": 0.7938,
+      "step": 20238
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.6818555008764077e-07,
+      "loss": 0.688,
+      "step": 20239
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.668511531122576e-07,
+      "loss": 0.9017,
+      "step": 20240
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.6551917418019554e-07,
+      "loss": 0.8548,
+      "step": 20241
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.641896133237621e-07,
+      "loss": 0.6759,
+      "step": 20242
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.628624705752315e-07,
+      "loss": 0.7048,
+      "step": 20243
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.615377459668112e-07,
+      "loss": 0.6913,
+      "step": 20244
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.602154395306534e-07,
+      "loss": 0.8208,
+      "step": 20245
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.5889555129884347e-07,
+      "loss": 0.8215,
+      "step": 20246
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.575780813034335e-07,
+      "loss": 0.7713,
+      "step": 20247
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.562630295763758e-07,
+      "loss": 0.7577,
+      "step": 20248
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.5495039614958924e-07,
+      "loss": 0.8951,
+      "step": 20249
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.5364018105494835e-07,
+      "loss": 0.9247,
+      "step": 20250
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.5233238432422764e-07,
+      "loss": 0.8115,
+      "step": 20251
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.510270059891796e-07,
+      "loss": 0.7225,
+      "step": 20252
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.497240460814788e-07,
+      "loss": 0.6774,
+      "step": 20253
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.4842350463276664e-07,
+      "loss": 0.6711,
+      "step": 20254
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.4712538167458454e-07,
+      "loss": 0.8351,
+      "step": 20255
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.4582967723844063e-07,
+      "loss": 0.8328,
+      "step": 20256
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.4453639135578753e-07,
+      "loss": 0.7777,
+      "step": 20257
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.4324552405801124e-07,
+      "loss": 0.7603,
+      "step": 20258
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.419570753764312e-07,
+      "loss": 0.8322,
+      "step": 20259
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.406710453423334e-07,
+      "loss": 0.84,
+      "step": 20260
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.393874339869152e-07,
+      "loss": 0.8324,
+      "step": 20261
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.381062413413405e-07,
+      "loss": 0.7441,
+      "step": 20262
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.3682746743669557e-07,
+      "loss": 1.1512,
+      "step": 20263
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.3555111230401116e-07,
+      "loss": 1.0043,
+      "step": 20264
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.3427717597427357e-07,
+      "loss": 0.8647,
+      "step": 20265
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.330056584784025e-07,
+      "loss": 0.7774,
+      "step": 20266
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.3173655984725105e-07,
+      "loss": 0.8453,
+      "step": 20267
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.304698801116168e-07,
+      "loss": 0.7971,
+      "step": 20268
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.292056193022419e-07,
+      "loss": 0.8545,
+      "step": 20269
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.279437774498129e-07,
+      "loss": 1.2386,
+      "step": 20270
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.266843545849496e-07,
+      "loss": 1.4731,
+      "step": 20271
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.2542735073822775e-07,
+      "loss": 0.9606,
+      "step": 20272
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.241727659401561e-07,
+      "loss": 0.838,
+      "step": 20273
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.229206002211549e-07,
+      "loss": 1.3082,
+      "step": 20274
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.2167085361164415e-07,
+      "loss": 1.0528,
+      "step": 20275
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.204235261419442e-07,
+      "loss": 0.7962,
+      "step": 20276
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.191786178423195e-07,
+      "loss": 0.8783,
+      "step": 20277
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.179361287429905e-07,
+      "loss": 0.8726,
+      "step": 20278
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.166960588741108e-07,
+      "loss": 0.6188,
+      "step": 20279
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.1545840826577855e-07,
+      "loss": 0.8016,
+      "step": 20280
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.142231769480364e-07,
+      "loss": 0.8079,
+      "step": 20281
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.129903649508492e-07,
+      "loss": 0.7566,
+      "step": 20282
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.1175997230413754e-07,
+      "loss": 0.8049,
+      "step": 20283
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.105319990377664e-07,
+      "loss": 0.775,
+      "step": 20284
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.093064451815453e-07,
+      "loss": 0.7861,
+      "step": 20285
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.08083310765217e-07,
+      "loss": 0.7568,
+      "step": 20286
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.068625958184468e-07,
+      "loss": 0.8799,
+      "step": 20287
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.0564430037088863e-07,
+      "loss": 0.8613,
+      "step": 20288
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.0442842445208566e-07,
+      "loss": 0.8661,
+      "step": 20289
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.032149680915586e-07,
+      "loss": 0.9089,
+      "step": 20290
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.020039313187617e-07,
+      "loss": 0.7338,
+      "step": 20291
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.007953141630715e-07,
+      "loss": 0.9536,
+      "step": 20292
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.995891166538423e-07,
+      "loss": 1.1435,
+      "step": 20293
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.9838533882031725e-07,
+      "loss": 1.1165,
+      "step": 20294
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.971839806917398e-07,
+      "loss": 0.8418,
+      "step": 20295
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.959850422972532e-07,
+      "loss": 0.7299,
+      "step": 20296
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.9478852366595646e-07,
+      "loss": 1.211,
+      "step": 20297
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.935944248268929e-07,
+      "loss": 1.1351,
+      "step": 20298
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.924027458090395e-07,
+      "loss": 0.7741,
+      "step": 20299
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.912134866413174e-07,
+      "loss": 0.9303,
+      "step": 20300
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.9002664735258146e-07,
+      "loss": 0.8622,
+      "step": 20301
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8884222797165295e-07,
+      "loss": 0.8137,
+      "step": 20302
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.876602285272534e-07,
+      "loss": 0.7829,
+      "step": 20303
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8648064904809313e-07,
+      "loss": 1.3125,
+      "step": 20304
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8530348956279373e-07,
+      "loss": 1.1077,
+      "step": 20305
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8412875009992126e-07,
+      "loss": 0.8448,
+      "step": 20306
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8295643068797504e-07,
+      "loss": 0.8183,
+      "step": 20307
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.817865313554213e-07,
+      "loss": 0.8492,
+      "step": 20308
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.806190521306484e-07,
+      "loss": 1.2904,
+      "step": 20309
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.794539930419893e-07,
+      "loss": 1.4956,
+      "step": 20310
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7829135411772125e-07,
+      "loss": 0.9958,
+      "step": 20311
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7713113538606616e-07,
+      "loss": 0.8304,
+      "step": 20312
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7597333687516824e-07,
+      "loss": 0.6714,
+      "step": 20313
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.748179586131383e-07,
+      "loss": 0.8237,
+      "step": 20314
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7366500062802057e-07,
+      "loss": 0.9522,
+      "step": 20315
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.725144629477705e-07,
+      "loss": 0.6971,
+      "step": 20316
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7136634560034345e-07,
+      "loss": 0.9863,
+      "step": 20317
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7022064861358385e-07,
+      "loss": 0.8481,
+      "step": 20318
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.690773720153028e-07,
+      "loss": 0.774,
+      "step": 20319
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6793651583325586e-07,
+      "loss": 0.859,
+      "step": 20320
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6679808009510973e-07,
+      "loss": 0.801,
+      "step": 20321
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6566206482850906e-07,
+      "loss": 0.7877,
+      "step": 20322
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.6452847006103175e-07,
+      "loss": 0.8102,
+      "step": 20323
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.6339729582016694e-07,
+      "loss": 0.66,
+      "step": 20324
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.622685421333926e-07,
+      "loss": 0.748,
+      "step": 20325
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.6114220902807574e-07,
+      "loss": 0.8657,
+      "step": 20326
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.600182965315723e-07,
+      "loss": 0.8326,
+      "step": 20327
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.588968046711604e-07,
+      "loss": 0.7986,
+      "step": 20328
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.5777773347404056e-07,
+      "loss": 0.7312,
+      "step": 20329
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.56661082967391e-07,
+      "loss": 0.9017,
+      "step": 20330
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.555468531783012e-07,
+      "loss": 0.8828,
+      "step": 20331
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.54435044133805e-07,
+      "loss": 0.8113,
+      "step": 20332
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.533256558609143e-07,
+      "loss": 0.8546,
+      "step": 20333
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.522186883865185e-07,
+      "loss": 0.8611,
+      "step": 20334
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.5111414173750737e-07,
+      "loss": 0.7914,
+      "step": 20335
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.500120159406816e-07,
+      "loss": 0.7777,
+      "step": 20336
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4891231102277533e-07,
+      "loss": 0.6934,
+      "step": 20337
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4781502701050063e-07,
+      "loss": 0.8589,
+      "step": 20338
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.467201639304806e-07,
+      "loss": 0.6459,
+      "step": 20339
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4562772180927174e-07,
+      "loss": 0.8293,
+      "step": 20340
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.445377006734084e-07,
+      "loss": 0.7863,
+      "step": 20341
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4345010054933614e-07,
+      "loss": 0.8857,
+      "step": 20342
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4236492146344494e-07,
+      "loss": 0.7437,
+      "step": 20343
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412821634420692e-07,
+      "loss": 0.7355,
+      "step": 20344
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4020182651149916e-07,
+      "loss": 0.7807,
+      "step": 20345
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3912391069794703e-07,
+      "loss": 0.7466,
+      "step": 20346
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.380484160275809e-07,
+      "loss": 0.7887,
+      "step": 20347
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.369753425264909e-07,
+      "loss": 0.7849,
+      "step": 20348
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3590469022073401e-07,
+      "loss": 0.8291,
+      "step": 20349
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3483645913627839e-07,
+      "loss": 0.7585,
+      "step": 20350
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3377064929905878e-07,
+      "loss": 0.763,
+      "step": 20351
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3270726073493233e-07,
+      "loss": 0.7978,
+      "step": 20352
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3164629346972277e-07,
+      "loss": 0.8291,
+      "step": 20353
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.3058774752917622e-07,
+      "loss": 0.7931,
+      "step": 20354
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2953162293897213e-07,
+      "loss": 0.7337,
+      "step": 20355
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2847791972474552e-07,
+      "loss": 0.8481,
+      "step": 20356
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2742663791206488e-07,
+      "loss": 0.8566,
+      "step": 20357
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2637777752646528e-07,
+      "loss": 0.847,
+      "step": 20358
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2533133859338195e-07,
+      "loss": 0.8074,
+      "step": 20359
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2428732113821683e-07,
+      "loss": 0.8319,
+      "step": 20360
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2324572518630516e-07,
+      "loss": 0.8569,
+      "step": 20361
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2220655076292674e-07,
+      "loss": 0.7681,
+      "step": 20362
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2116979789330582e-07,
+      "loss": 0.856,
+      "step": 20363
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.201354666026001e-07,
+      "loss": 1.2882,
+      "step": 20364
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1910355691592278e-07,
+      "loss": 1.1339,
+      "step": 20365
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1807406885829828e-07,
+      "loss": 0.8365,
+      "step": 20366
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1704700245472888e-07,
+      "loss": 0.7908,
+      "step": 20367
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1602235773012792e-07,
+      "loss": 0.8115,
+      "step": 20368
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1500013470937552e-07,
+      "loss": 0.8421,
+      "step": 20369
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1398033341726297e-07,
+      "loss": 0.8497,
+      "step": 20370
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1296295387855937e-07,
+      "loss": 0.7845,
+      "step": 20371
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.11947996117956e-07,
+      "loss": 0.8217,
+      "step": 20372
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.109354601600666e-07,
+      "loss": 1.3948,
+      "step": 20373
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0992534602948255e-07,
+      "loss": 1.1113,
+      "step": 20374
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0891765375070648e-07,
+      "loss": 0.7467,
+      "step": 20375
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0791238334820773e-07,
+      "loss": 0.8375,
+      "step": 20376
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0690953484636678e-07,
+      "loss": 0.8829,
+      "step": 20377
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0590910826954191e-07,
+      "loss": 1.2768,
+      "step": 20378
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.049111036419915e-07,
+      "loss": 0.9945,
+      "step": 20379
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0391552098795174e-07,
+      "loss": 0.803,
+      "step": 20380
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0292236033158108e-07,
+      "loss": 0.715,
+      "step": 20381
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0193162169698242e-07,
+      "loss": 0.749,
+      "step": 20382
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0094330510820324e-07,
+      "loss": 0.8382,
+      "step": 20383
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9995741058922433e-07,
+      "loss": 0.7322,
+      "step": 20384
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9897393816397103e-07,
+      "loss": 0.6139,
+      "step": 20385
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.979928878563242e-07,
+      "loss": 0.8367,
+      "step": 20386
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.97014259690087e-07,
+      "loss": 0.9191,
+      "step": 20387
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9603805368899609e-07,
+      "loss": 0.8398,
+      "step": 20388
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9506426987676574e-07,
+      "loss": 0.705,
+      "step": 20389
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9409290827701043e-07,
+      "loss": 0.7285,
+      "step": 20390
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.931239689133224e-07,
+      "loss": 0.8243,
+      "step": 20391
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9215745180920508e-07,
+      "loss": 0.8886,
+      "step": 20392
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9119335698811746e-07,
+      "loss": 0.7471,
+      "step": 20393
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9023168447346307e-07,
+      "loss": 0.7441,
+      "step": 20394
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8927243428857877e-07,
+      "loss": 0.7739,
+      "step": 20395
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8831560645673486e-07,
+      "loss": 0.7409,
+      "step": 20396
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8736120100116826e-07,
+      "loss": 0.6729,
+      "step": 20397
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.864092179450272e-07,
+      "loss": 0.8403,
+      "step": 20398
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8545965731142645e-07,
+      "loss": 1.0112,
+      "step": 20399
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8451251912341427e-07,
+      "loss": 0.8195,
+      "step": 20400
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8356780340396118e-07,
+      "loss": 0.7724,
+      "step": 20401
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.826255101760155e-07,
+      "loss": 0.7874,
+      "step": 20402
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8168563946243668e-07,
+      "loss": 0.8222,
+      "step": 20403
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.807481912860287e-07,
+      "loss": 0.7804,
+      "step": 20404
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.798131656695401e-07,
+      "loss": 0.6232,
+      "step": 20405
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7888056263568597e-07,
+      "loss": 0.931,
+      "step": 20406
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.779503822070705e-07,
+      "loss": 1.2626,
+      "step": 20407
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.770226244062978e-07,
+      "loss": 1.1889,
+      "step": 20408
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.76097289255861e-07,
+      "loss": 0.8587,
+      "step": 20409
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7517437677823102e-07,
+      "loss": 0.9594,
+      "step": 20410
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7425388699580103e-07,
+      "loss": 0.8036,
+      "step": 20411
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7333581993090876e-07,
+      "loss": 0.7741,
+      "step": 20412
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7242017560584744e-07,
+      "loss": 0.7598,
+      "step": 20413
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7150695404282159e-07,
+      "loss": 1.1332,
+      "step": 20414
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.705961552640023e-07,
+      "loss": 1.0932,
+      "step": 20415
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6968777929150525e-07,
+      "loss": 0.8006,
+      "step": 20416
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6878182614735728e-07,
+      "loss": 0.9992,
+      "step": 20417
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6787829585355185e-07,
+      "loss": 0.7453,
+      "step": 20418
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6697718843202703e-07,
+      "loss": 0.8986,
+      "step": 20419
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6607850390464308e-07,
+      "loss": 0.6988,
+      "step": 20420
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6518224229320478e-07,
+      "loss": 0.8015,
+      "step": 20421
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.642884036194836e-07,
+      "loss": 0.7326,
+      "step": 20422
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6339698790515113e-07,
+      "loss": 0.7455,
+      "step": 20423
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6250799517185666e-07,
+      "loss": 1.2161,
+      "step": 20424
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6162142544116077e-07,
+      "loss": 1.0087,
+      "step": 20425
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6073727873459067e-07,
+      "loss": 0.6941,
+      "step": 20426
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5985555507360695e-07,
+      "loss": 0.7079,
+      "step": 20427
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5897625447960364e-07,
+      "loss": 0.8536,
+      "step": 20428
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5809937697391918e-07,
+      "loss": 0.8916,
+      "step": 20429
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.572249225778366e-07,
+      "loss": 0.8122,
+      "step": 20430
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5635289131258335e-07,
+      "loss": 0.8534,
+      "step": 20431
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5548328319930917e-07,
+      "loss": 0.8552,
+      "step": 20432
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5461609825913048e-07,
+      "loss": 0.8413,
+      "step": 20433
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5375133651309716e-07,
+      "loss": 0.7802,
+      "step": 20434
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5288899798219236e-07,
+      "loss": 0.8902,
+      "step": 20435
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5202908268733274e-07,
+      "loss": 0.842,
+      "step": 20436
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5117159064940157e-07,
+      "loss": 0.6684,
+      "step": 20437
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5031652188920443e-07,
+      "loss": 0.9976,
+      "step": 20438
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4946387642750247e-07,
+      "loss": 0.8489,
+      "step": 20439
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.486136542849792e-07,
+      "loss": 0.7654,
+      "step": 20440
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4776585548227362e-07,
+      "loss": 0.7659,
+      "step": 20441
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4692048003994707e-07,
+      "loss": 0.817,
+      "step": 20442
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4607752797853867e-07,
+      "loss": 0.791,
+      "step": 20443
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4523699931849878e-07,
+      "loss": 0.7348,
+      "step": 20444
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.443988940802221e-07,
+      "loss": 0.7182,
+      "step": 20445
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.43563212284048e-07,
+      "loss": 0.7866,
+      "step": 20446
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4272995395026024e-07,
+      "loss": 1.2174,
+      "step": 20447
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4189911909907594e-07,
+      "loss": 0.9175,
+      "step": 20448
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4107070775067898e-07,
+      "loss": 0.944,
+      "step": 20449
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4024471992515332e-07,
+      "loss": 0.71,
+      "step": 20450
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3942115564254953e-07,
+      "loss": 0.7865,
+      "step": 20451
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3860001492286279e-07,
+      "loss": 0.8068,
+      "step": 20452
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3778129778602156e-07,
+      "loss": 0.8238,
+      "step": 20453
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3696500425188774e-07,
+      "loss": 0.7557,
+      "step": 20454
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.361511343402788e-07,
+      "loss": 0.801,
+      "step": 20455
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3533968807093456e-07,
+      "loss": 0.7999,
+      "step": 20456
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3453066546356142e-07,
+      "loss": 0.8065,
+      "step": 20457
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3372406653779923e-07,
+      "loss": 0.7805,
+      "step": 20458
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3291989131321016e-07,
+      "loss": 0.6931,
+      "step": 20459
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3211813980931186e-07,
+      "loss": 0.679,
+      "step": 20460
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.313188120455666e-07,
+      "loss": 0.9059,
+      "step": 20461
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.305219080413811e-07,
+      "loss": 1.3638,
+      "step": 20462
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2972742781608426e-07,
+      "loss": 0.9621,
+      "step": 20463
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.289353713889607e-07,
+      "loss": 0.8475,
+      "step": 20464
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2814573877922843e-07,
+      "loss": 1.2518,
+      "step": 20465
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.273585300060609e-07,
+      "loss": 0.9955,
+      "step": 20466
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.265737450885651e-07,
+      "loss": 0.7843,
+      "step": 20467
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2579138404577028e-07,
+      "loss": 0.8707,
+      "step": 20468
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2501144689668343e-07,
+      "loss": 0.9305,
+      "step": 20469
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2423393366022274e-07,
+      "loss": 0.8959,
+      "step": 20470
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2345884435526201e-07,
+      "loss": 0.7454,
+      "step": 20471
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.226861790006084e-07,
+      "loss": 0.8754,
+      "step": 20472
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2191593761502472e-07,
+      "loss": 0.8071,
+      "step": 20473
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2114812021718492e-07,
+      "loss": 1.2738,
+      "step": 20474
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.203827268257407e-07,
+      "loss": 1.0567,
+      "step": 20475
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1961975745926613e-07,
+      "loss": 0.8045,
+      "step": 20476
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.188592121362686e-07,
+      "loss": 0.7286,
+      "step": 20477
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1810109087520005e-07,
+      "loss": 0.7354,
+      "step": 20478
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1734539369447905e-07,
+      "loss": 0.7805,
+      "step": 20479
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1659212061244651e-07,
+      "loss": 0.8635,
+      "step": 20480
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1584127164736558e-07,
+      "loss": 0.6821,
+      "step": 20481
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1509284681746613e-07,
+      "loss": 1.1183,
+      "step": 20482
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1434684614092251e-07,
+      "loss": 1.1174,
+      "step": 20483
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1360326963582024e-07,
+      "loss": 0.7661,
+      "step": 20484
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1286211732022267e-07,
+      "loss": 0.7797,
+      "step": 20485
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.121233892121154e-07,
+      "loss": 0.881,
+      "step": 20486
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1138708532942854e-07,
+      "loss": 0.8317,
+      "step": 20487
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.1065320569001447e-07,
+      "loss": 0.8252,
+      "step": 20488
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0992175031169227e-07,
+      "loss": 0.7872,
+      "step": 20489
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0919271921222551e-07,
+      "loss": 0.869,
+      "step": 20490
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0846611240930004e-07,
+      "loss": 0.812,
+      "step": 20491
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0774192992054622e-07,
+      "loss": 0.8676,
+      "step": 20492
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0702017176354995e-07,
+      "loss": 0.8529,
+      "step": 20493
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0630083795580836e-07,
+      "loss": 0.6917,
+      "step": 20494
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0558392851479637e-07,
+      "loss": 0.8181,
+      "step": 20495
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0486944345791116e-07,
+      "loss": 0.7863,
+      "step": 20496
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0415738280249443e-07,
+      "loss": 0.862,
+      "step": 20497
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0344774656581013e-07,
+      "loss": 0.7012,
+      "step": 20498
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0274053476511115e-07,
+      "loss": 0.9112,
+      "step": 20499
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.020357474175282e-07,
+      "loss": 0.828,
+      "step": 20500
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0133338454019203e-07,
+      "loss": 0.9009,
+      "step": 20501
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0063344615013348e-07,
+      "loss": 0.7892,
+      "step": 20502
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.993593226433895e-08,
+      "loss": 0.7721,
+      "step": 20503
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.924084289975044e-08,
+      "loss": 0.9349,
+      "step": 20504
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.854817807323225e-08,
+      "loss": 0.7613,
+      "step": 20505
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.785793780158203e-08,
+      "loss": 0.8347,
+      "step": 20506
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.717012210156418e-08,
+      "loss": 0.7248,
+      "step": 20507
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.648473098987643e-08,
+      "loss": 0.7814,
+      "step": 20508
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.580176448313882e-08,
+      "loss": 0.8356,
+      "step": 20509
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.51212225979381e-08,
+      "loss": 0.6976,
+      "step": 20510
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.444310535077216e-08,
+      "loss": 0.9274,
+      "step": 20511
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.376741275812783e-08,
+      "loss": 0.8559,
+      "step": 20512
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.30941448363809e-08,
+      "loss": 0.747,
+      "step": 20513
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.242330160187385e-08,
+      "loss": 0.8799,
+      "step": 20514
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.175488307090475e-08,
+      "loss": 0.9013,
+      "step": 20515
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.108888925967174e-08,
+      "loss": 0.8879,
+      "step": 20516
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.042532018435079e-08,
+      "loss": 0.8662,
+      "step": 20517
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.976417586105123e-08,
+      "loss": 0.8078,
+      "step": 20518
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.91054563058158e-08,
+      "loss": 0.7107,
+      "step": 20519
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.844916153462057e-08,
+      "loss": 0.8771,
+      "step": 20520
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.779529156339728e-08,
+      "loss": 1.0905,
+      "step": 20521
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.714384640802208e-08,
+      "loss": 1.0508,
+      "step": 20522
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.649482608430459e-08,
+      "loss": 0.8956,
+      "step": 20523
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.584823060799885e-08,
+      "loss": 0.8464,
+      "step": 20524
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.520405999478121e-08,
+      "loss": 0.8476,
+      "step": 20525
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.456231426029471e-08,
+      "loss": 0.8163,
+      "step": 20526
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.392299342012688e-08,
+      "loss": 0.8176,
+      "step": 20527
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.328609748977645e-08,
+      "loss": 0.8374,
+      "step": 20528
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.26516264847088e-08,
+      "loss": 0.7679,
+      "step": 20529
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.201958042031166e-08,
+      "loss": 0.8028,
+      "step": 20530
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.138995931193938e-08,
+      "loss": 0.8499,
+      "step": 20531
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.076276317486864e-08,
+      "loss": 0.8065,
+      "step": 20532
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.01379920243095e-08,
+      "loss": 0.8333,
+      "step": 20533
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.95156458754498e-08,
+      "loss": 0.7676,
+      "step": 20534
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.889572474335527e-08,
+      "loss": 0.8817,
+      "step": 20535
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.827822864311384e-08,
+      "loss": 0.8067,
+      "step": 20536
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.766315758968024e-08,
+      "loss": 0.8187,
+      "step": 20537
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.705051159798694e-08,
+      "loss": 0.7758,
+      "step": 20538
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.644029068292202e-08,
+      "loss": 0.8279,
+      "step": 20539
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.583249485927368e-08,
+      "loss": 0.9961,
+      "step": 20540
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.522712414179679e-08,
+      "loss": 0.722,
+      "step": 20541
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.462417854520176e-08,
+      "loss": 0.8175,
+      "step": 20542
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.402365808409917e-08,
+      "loss": 0.7569,
+      "step": 20543
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.342556277306623e-08,
+      "loss": 0.8265,
+      "step": 20544
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.282989262662465e-08,
+      "loss": 0.7833,
+      "step": 20545
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.223664765924065e-08,
+      "loss": 0.7854,
+      "step": 20546
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.164582788529161e-08,
+      "loss": 0.811,
+      "step": 20547
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.105743331913272e-08,
+      "loss": 0.8245,
+      "step": 20548
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.047146397503035e-08,
+      "loss": 0.8197,
+      "step": 20549
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.988791986721754e-08,
+      "loss": 0.8101,
+      "step": 20550
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.930680100984965e-08,
+      "loss": 0.8066,
+      "step": 20551
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.872810741702652e-08,
+      "loss": 0.7856,
+      "step": 20552
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.815183910281464e-08,
+      "loss": 0.797,
+      "step": 20553
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.757799608116955e-08,
+      "loss": 0.9116,
+      "step": 20554
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.700657836603563e-08,
+      "loss": 0.8212,
+      "step": 20555
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.643758597127958e-08,
+      "loss": 0.797,
+      "step": 20556
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.587101891070147e-08,
+      "loss": 0.7407,
+      "step": 20557
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.530687719806805e-08,
+      "loss": 0.8313,
+      "step": 20558
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.474516084705729e-08,
+      "loss": 0.7984,
+      "step": 20559
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.418586987130271e-08,
+      "loss": 0.7057,
+      "step": 20560
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.362900428438235e-08,
+      "loss": 0.8546,
+      "step": 20561
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.307456409980761e-08,
+      "loss": 0.8776,
+      "step": 20562
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.25225493310455e-08,
+      "loss": 0.7297,
+      "step": 20563
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.197295999147424e-08,
+      "loss": 0.7724,
+      "step": 20564
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.142579609444976e-08,
+      "loss": 0.7262,
+      "step": 20565
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.088105765323926e-08,
+      "loss": 0.79,
+      "step": 20566
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.033874468106549e-08,
+      "loss": 0.8358,
+      "step": 20567
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.979885719109568e-08,
+      "loss": 0.6522,
+      "step": 20568
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.926139519643048e-08,
+      "loss": 0.7677,
+      "step": 20569
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.8726358710103905e-08,
+      "loss": 0.7401,
+      "step": 20570
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.819374774511665e-08,
+      "loss": 0.8498,
+      "step": 20571
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.766356231439174e-08,
+      "loss": 0.832,
+      "step": 20572
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.713580243078554e-08,
+      "loss": 0.7372,
+      "step": 20573
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.6610468107110015e-08,
+      "loss": 0.8517,
+      "step": 20574
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.608755935612164e-08,
+      "loss": 0.8212,
+      "step": 20575
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.556707619051027e-08,
+      "loss": 0.875,
+      "step": 20576
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.504901862291023e-08,
+      "loss": 0.8475,
+      "step": 20577
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.453338666587815e-08,
+      "loss": 0.8554,
+      "step": 20578
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.402018033194845e-08,
+      "loss": 0.8686,
+      "step": 20579
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.350939963356671e-08,
+      "loss": 0.8528,
+      "step": 20580
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.300104458312305e-08,
+      "loss": 0.8069,
+      "step": 20581
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.2495115192963126e-08,
+      "loss": 0.9498,
+      "step": 20582
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.199161147536602e-08,
+      "loss": 0.7883,
+      "step": 20583
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.149053344255528e-08,
+      "loss": 0.6227,
+      "step": 20584
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.0991881106687846e-08,
+      "loss": 0.7127,
+      "step": 20585
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.049565447985405e-08,
+      "loss": 0.8595,
+      "step": 20586
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.0001853574110914e-08,
+      "loss": 0.6829,
+      "step": 20587
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.9510478401437744e-08,
+      "loss": 0.6947,
+      "step": 20588
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.9021528973758336e-08,
+      "loss": 0.8675,
+      "step": 20589
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8535005302952076e-08,
+      "loss": 0.8617,
+      "step": 20590
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.805090740080953e-08,
+      "loss": 0.7584,
+      "step": 20591
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7569235279087965e-08,
+      "loss": 0.858,
+      "step": 20592
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7089988949478025e-08,
+      "loss": 0.6719,
+      "step": 20593
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.661316842361485e-08,
+      "loss": 0.7888,
+      "step": 20594
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.613877371305586e-08,
+      "loss": 0.7919,
+      "step": 20595
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.566680482932517e-08,
+      "loss": 0.6959,
+      "step": 20596
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.519726178386918e-08,
+      "loss": 0.7925,
+      "step": 20597
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.4730144588089884e-08,
+      "loss": 0.8526,
+      "step": 20598
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.426545325333375e-08,
+      "loss": 0.8845,
+      "step": 20599
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.380318779085846e-08,
+      "loss": 0.7911,
+      "step": 20600
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.334334821189945e-08,
+      "loss": 0.7641,
+      "step": 20601
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.2885934527603366e-08,
+      "loss": 0.6998,
+      "step": 20602
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.243094674908355e-08,
+      "loss": 0.7431,
+      "step": 20603
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.19783848873756e-08,
+      "loss": 0.8253,
+      "step": 20604
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.1528248953470736e-08,
+      "loss": 0.779,
+      "step": 20605
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.108053895827135e-08,
+      "loss": 1.1658,
+      "step": 20606
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.063525491266873e-08,
+      "loss": 1.0955,
+      "step": 20607
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.019239682746534e-08,
+      "loss": 0.8606,
+      "step": 20608
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.975196471339704e-08,
+      "loss": 0.7148,
+      "step": 20609
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.931395858115528e-08,
+      "loss": 0.8336,
+      "step": 20610
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.8878378441375985e-08,
+      "loss": 0.8197,
+      "step": 20611
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.8445224304628494e-08,
+      "loss": 0.9209,
+      "step": 20612
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.801449618143771e-08,
+      "loss": 0.9338,
+      "step": 20613
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.7586194082228634e-08,
+      "loss": 0.839,
+      "step": 20614
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.7160318017415155e-08,
+      "loss": 0.8145,
+      "step": 20615
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.673686799734455e-08,
+      "loss": 1.2952,
+      "step": 20616
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.631584403226418e-08,
+      "loss": 1.0186,
+      "step": 20617
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.58972461324103e-08,
+      "loss": 0.7647,
+      "step": 20618
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.548107430794145e-08,
+      "loss": 0.7237,
+      "step": 20619
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.5067328568949567e-08,
+      "loss": 0.8707,
+      "step": 20620
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.465600892549326e-08,
+      "loss": 0.8045,
+      "step": 20621
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.424711538753123e-08,
+      "loss": 0.6617,
+      "step": 20622
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.384064796501107e-08,
+      "loss": 0.8124,
+      "step": 20623
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.343660666778048e-08,
+      "loss": 0.9156,
+      "step": 20624
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.303499150564271e-08,
+      "loss": 0.9809,
+      "step": 20625
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.263580248836773e-08,
+      "loss": 0.6168,
+      "step": 20626
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.2239039625614476e-08,
+      "loss": 0.7215,
+      "step": 20627
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.184470292704189e-08,
+      "loss": 0.9084,
+      "step": 20628
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.1452792402197895e-08,
+      "loss": 0.9548,
+      "step": 20629
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.10633080605971e-08,
+      "loss": 0.9252,
+      "step": 20630
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.06762499116986e-08,
+      "loss": 0.772,
+      "step": 20631
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.029161796489488e-08,
+      "loss": 0.9426,
+      "step": 20632
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.990941222951182e-08,
+      "loss": 0.6993,
+      "step": 20633
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.9529632714841993e-08,
+      "loss": 0.8801,
+      "step": 20634
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.9152279430089136e-08,
+      "loss": 0.7807,
+      "step": 20635
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8777352384423696e-08,
+      "loss": 0.7992,
+      "step": 20636
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8404851586927295e-08,
+      "loss": 0.614,
+      "step": 20637
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8034777046648253e-08,
+      "loss": 0.7822,
+      "step": 20638
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.7667128772579376e-08,
+      "loss": 0.7202,
+      "step": 20639
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.7301906773624653e-08,
+      "loss": 1.2559,
+      "step": 20640
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.6939111058665865e-08,
+      "loss": 1.1316,
+      "step": 20641
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.6578741636484882e-08,
+      "loss": 0.8622,
+      "step": 20642
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.6220798515852464e-08,
+      "loss": 0.8661,
+      "step": 20643
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.586528170543945e-08,
+      "loss": 0.7206,
+      "step": 20644
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.5512191213883375e-08,
+      "loss": 0.8805,
+      "step": 20645
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.5161527049755164e-08,
+      "loss": 0.854,
+      "step": 20646
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.481328922154802e-08,
+      "loss": 0.8723,
+      "step": 20647
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4467477737732947e-08,
+      "loss": 0.7627,
+      "step": 20648
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4124092606692127e-08,
+      "loss": 0.8823,
+      "step": 20649
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.378313383676334e-08,
+      "loss": 0.7947,
+      "step": 20650
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.344460143621774e-08,
+      "loss": 0.834,
+      "step": 20651
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.310849541327098e-08,
+      "loss": 0.8062,
+      "step": 20652
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.2774815776083202e-08,
+      "loss": 0.8655,
+      "step": 20653
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.2443562532759032e-08,
+      "loss": 0.7871,
+      "step": 20654
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.2114735691325383e-08,
+      "loss": 0.6134,
+      "step": 20655
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.178833525977586e-08,
+      "loss": 0.6729,
+      "step": 20656
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.1464361246015252e-08,
+      "loss": 0.8547,
+      "step": 20657
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.114281365792614e-08,
+      "loss": 0.821,
+      "step": 20658
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.082369250329119e-08,
+      "loss": 0.8018,
+      "step": 20659
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.0506997789870865e-08,
+      "loss": 0.8147,
+      "step": 20660
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.0192729525347897e-08,
+      "loss": 0.9029,
+      "step": 20661
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.9880887717338427e-08,
+      "loss": 0.7155,
+      "step": 20662
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.9571472373436374e-08,
+      "loss": 0.7658,
+      "step": 20663
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.926448350112464e-08,
+      "loss": 0.8226,
+      "step": 20664
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.8959921107863932e-08,
+      "loss": 0.8055,
+      "step": 20665
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.8657785201048327e-08,
+      "loss": 0.7116,
+      "step": 20666
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.83580757880053e-08,
+      "loss": 0.8245,
+      "step": 20667
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.8060792876006816e-08,
+      "loss": 0.7616,
+      "step": 20668
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.776593647228042e-08,
+      "loss": 1.1423,
+      "step": 20669
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.747350658396485e-08,
+      "loss": 1.0526,
+      "step": 20670
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.7183503218176633e-08,
+      "loss": 0.8902,
+      "step": 20671
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.689592638193238e-08,
+      "loss": 0.8259,
+      "step": 20672
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.6610776082226498e-08,
+      "loss": 0.6999,
+      "step": 20673
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.6328052325975672e-08,
+      "loss": 0.8414,
+      "step": 20674
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.6047755120052187e-08,
+      "loss": 0.7728,
+      "step": 20675
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.5769884471239505e-08,
+      "loss": 0.8062,
+      "step": 20676
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.549444038628778e-08,
+      "loss": 0.8103,
+      "step": 20677
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.5221422871880553e-08,
+      "loss": 0.8898,
+      "step": 20678
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.4950831934645858e-08,
+      "loss": 0.841,
+      "step": 20679
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.4682667581167319e-08,
+      "loss": 0.9226,
+      "step": 20680
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.4416929817928637e-08,
+      "loss": 0.7784,
+      "step": 20681
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.4153618651380208e-08,
+      "loss": 0.8085,
+      "step": 20682
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.3892734087928017e-08,
+      "loss": 0.8505,
+      "step": 20683
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.3634276133900337e-08,
+      "loss": 0.6808,
+      "step": 20684
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.3378244795558825e-08,
+      "loss": 0.8619,
+      "step": 20685
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.3124640079131834e-08,
+      "loss": 0.7998,
+      "step": 20686
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.2873461990758894e-08,
+      "loss": 0.7196,
+      "step": 20687
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.2624710536546237e-08,
+      "loss": 0.8493,
+      "step": 20688
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.2378385722522368e-08,
+      "loss": 0.7603,
+      "step": 20689
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.2134487554682494e-08,
+      "loss": 0.8367,
+      "step": 20690
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.18930160389219e-08,
+      "loss": 0.6662,
+      "step": 20691
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.1653971181124768e-08,
+      "loss": 0.9005,
+      "step": 20692
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.1417352987075358e-08,
+      "loss": 0.7422,
+      "step": 20693
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.1183161462524627e-08,
+      "loss": 0.7554,
+      "step": 20694
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0951396613156917e-08,
+      "loss": 0.7968,
+      "step": 20695
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0722058444589955e-08,
+      "loss": 0.8701,
+      "step": 20696
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0495146962397062e-08,
+      "loss": 0.83,
+      "step": 20697
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0270662172073841e-08,
+      "loss": 0.7998,
+      "step": 20698
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0048604079082592e-08,
+      "loss": 0.8318,
+      "step": 20699
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 9.828972688807892e-09,
+      "loss": 0.7679,
+      "step": 20700
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 9.611768006578815e-09,
+      "loss": 0.8076,
+      "step": 20701
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 9.396990037668918e-09,
+      "loss": 1.3003,
+      "step": 20702
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 9.184638787285149e-09,
+      "loss": 1.0226,
+      "step": 20703
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.974714260578942e-09,
+      "loss": 0.9222,
+      "step": 20704
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.767216462657325e-09,
+      "loss": 0.8558,
+      "step": 20705
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.562145398549604e-09,
+      "loss": 0.7425,
+      "step": 20706
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.359501073229581e-09,
+      "loss": 1.3551,
+      "step": 20707
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.159283491626646e-09,
+      "loss": 0.9801,
+      "step": 20708
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.96149265858137e-09,
+      "loss": 1.3932,
+      "step": 20709
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.766128578912123e-09,
+      "loss": 1.0598,
+      "step": 20710
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.573191257348456e-09,
+      "loss": 0.716,
+      "step": 20711
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.382680698586608e-09,
+      "loss": 1.3127,
+      "step": 20712
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.194596907222906e-09,
+      "loss": 1.0084,
+      "step": 20713
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.0089398878536716e-09,
+      "loss": 0.8699,
+      "step": 20714
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.825709644964207e-09,
+      "loss": 0.7072,
+      "step": 20715
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.644906183017607e-09,
+      "loss": 0.8143,
+      "step": 20716
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.46652950637705e-09,
+      "loss": 0.6975,
+      "step": 20717
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.29057961939461e-09,
+      "loss": 0.7715,
+      "step": 20718
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.117056526333542e-09,
+      "loss": 1.1343,
+      "step": 20719
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.945960231401593e-09,
+      "loss": 1.0291,
+      "step": 20720
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.777290738762098e-09,
+      "loss": 0.8383,
+      "step": 20721
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.611048052489576e-09,
+      "loss": 0.7265,
+      "step": 20722
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.447232176636341e-09,
+      "loss": 0.7871,
+      "step": 20723
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.28584311516589e-09,
+      "loss": 0.8422,
+      "step": 20724
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.12688087200841e-09,
+      "loss": 0.756,
+      "step": 20725
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.970345451005276e-09,
+      "loss": 0.6978,
+      "step": 20726
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.816236855964551e-09,
+      "loss": 0.7029,
+      "step": 20727
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.664555090627687e-09,
+      "loss": 0.8164,
+      "step": 20728
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.515300158680624e-09,
+      "loss": 0.6824,
+      "step": 20729
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.3684720637315875e-09,
+      "loss": 1.213,
+      "step": 20730
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.224070809355496e-09,
+      "loss": 0.9303,
+      "step": 20731
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.082096399049551e-09,
+      "loss": 0.805,
+      "step": 20732
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.942548836266546e-09,
+      "loss": 0.7984,
+      "step": 20733
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.8054281243815605e-09,
+      "loss": 0.8819,
+      "step": 20734
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.6707342667363643e-09,
+      "loss": 0.817,
+      "step": 20735
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.5384672665950137e-09,
+      "loss": 0.7329,
+      "step": 20736
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.408627127166053e-09,
+      "loss": 0.8185,
+      "step": 20737
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.281213851602516e-09,
+      "loss": 0.8471,
+      "step": 20738
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.156227442990822e-09,
+      "loss": 0.819,
+      "step": 20739
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.033667904372983e-09,
+      "loss": 0.714,
+      "step": 20740
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.9135352387243962e-09,
+      "loss": 0.8275,
+      "step": 20741
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.7958294489427437e-09,
+      "loss": 0.8518,
+      "step": 20742
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.6805505379035034e-09,
+      "loss": 0.8704,
+      "step": 20743
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.5676985083933347e-09,
+      "loss": 0.7843,
+      "step": 20744
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.457273363165591e-09,
+      "loss": 0.8508,
+      "step": 20745
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.3492751048848072e-09,
+      "loss": 0.842,
+      "step": 20746
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.243703736182212e-09,
+      "loss": 0.8022,
+      "step": 20747
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.140559259611319e-09,
+      "loss": 1.0311,
+      "step": 20748
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.039841677681231e-09,
+      "loss": 0.5716,
+      "step": 20749
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.94155099283444e-09,
+      "loss": 0.7491,
+      "step": 20750
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.8456872074468223e-09,
+      "loss": 0.7768,
+      "step": 20751
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.752250323872051e-09,
+      "loss": 1.1921,
+      "step": 20752
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.6612403443527768e-09,
+      "loss": 1.0776,
+      "step": 20753
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.5726572710983433e-09,
+      "loss": 0.763,
+      "step": 20754
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.4865011062736855e-09,
+      "loss": 0.7834,
+      "step": 20755
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.4027718519660227e-09,
+      "loss": 1.1609,
+      "step": 20756
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.3214695101959606e-09,
+      "loss": 1.1859,
+      "step": 20757
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.2425940829507986e-09,
+      "loss": 0.9064,
+      "step": 20758
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.1661455721401204e-09,
+      "loss": 0.7542,
+      "step": 20759
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.092123979606896e-09,
+      "loss": 0.9217,
+      "step": 20760
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.020529307160789e-09,
+      "loss": 0.8609,
+      "step": 20761
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.513615565448497e-10,
+      "loss": 0.7763,
+      "step": 20762
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.846207294244124e-10,
+      "loss": 0.8734,
+      "step": 20763
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.203068274204029e-10,
+      "loss": 0.8032,
+      "step": 20764
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.584198520982356e-10,
+      "loss": 0.9757,
+      "step": 20765
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.989598049567115e-10,
+      "loss": 0.7565,
+      "step": 20766
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.419266874502228e-10,
+      "loss": 1.1962,
+      "step": 20767
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.873205009443439e-10,
+      "loss": 1.1469,
+      "step": 20768
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.351412467824446e-10,
+      "loss": 0.8459,
+      "step": 20769
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.85388926219077e-10,
+      "loss": 0.7637,
+      "step": 20770
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.3806354046438404e-10,
+      "loss": 1.2958,
+      "step": 20771
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.9316509067299776e-10,
+      "loss": 1.1476,
+      "step": 20772
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.506935779218345e-10,
+      "loss": 0.8009,
+      "step": 20773
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.1064900325450397e-10,
+      "loss": 0.8145,
+      "step": 20774
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.730313676369001e-10,
+      "loss": 0.9113,
+      "step": 20775
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.3784067197940574e-10,
+      "loss": 0.7984,
+      "step": 20776
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.050769171368927e-10,
+      "loss": 0.7583,
+      "step": 20777
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.7474010390872154e-10,
+      "loss": 0.6893,
+      "step": 20778
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.4683023302763944e-10,
+      "loss": 0.9069,
+      "step": 20779
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.2134730518198466e-10,
+      "loss": 0.8857,
+      "step": 20780
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.829132097127769e-11,
+      "loss": 0.8584,
+      "step": 20781
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.766228097283446e-11,
+      "loss": 0.8682,
+      "step": 20782
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.94601856751531e-11,
+      "loss": 0.778,
+      "step": 20783
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.368503552232284e-11,
+      "loss": 0.8273,
+      "step": 20784
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.033683089181949e-11,
+      "loss": 0.6702,
+      "step": 20785
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.94155721278122e-11,
+      "loss": 0.8531,
+      "step": 20786
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0921259474550027e-11,
+      "loss": 0.7797,
+      "step": 20787
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.853893154077583e-12,
+      "loss": 0.9396,
+      "step": 20788
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.2134732996216258e-12,
+      "loss": 0.7952,
+      "step": 20789
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0,
+      "loss": 0.8374,
+      "step": 20790
+    },
+    {
+      "epoch": 2.0,
+      "step": 20790,
+      "total_flos": 7971742088724480.0,
+      "train_loss": 1.102952129064471,
+      "train_runtime": 282403.1125,
+      "train_samples_per_second": 4.712,
+      "train_steps_per_second": 0.074
+    }
+  ],
+  "max_steps": 20790,
+  "num_train_epochs": 2,
+  "total_flos": 7971742088724480.0,
+  "trial_name": null,
+  "trial_params": null
+}