{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.9998118355442656,
  "eval_steps": 500,
  "global_step": 5314,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 3.7593984962406015e-08,
      "loss": 3.8124,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.518796992481203e-08,
      "loss": 3.8379,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1278195488721805e-07,
      "loss": 3.747,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.5037593984962406e-07,
      "loss": 3.8622,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.8796992481203008e-07,
      "loss": 3.9582,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.255639097744361e-07,
      "loss": 3.8242,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.6315789473684213e-07,
      "loss": 3.8843,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.007518796992481e-07,
      "loss": 3.8447,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.3834586466165416e-07,
      "loss": 3.8244,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.7593984962406015e-07,
      "loss": 3.7997,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.135338345864662e-07,
      "loss": 3.8608,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.511278195488722e-07,
      "loss": 3.8211,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.887218045112782e-07,
      "loss": 3.7187,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.263157894736843e-07,
      "loss": 3.6495,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.639097744360903e-07,
      "loss": 3.7025,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.015037593984962e-07,
      "loss": 3.7721,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.390977443609022e-07,
      "loss": 3.6057,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.766917293233083e-07,
      "loss": 3.5782,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.142857142857143e-07,
      "loss": 3.6682,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.518796992481203e-07,
      "loss": 3.5243,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.894736842105263e-07,
      "loss": 3.5923,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.270676691729324e-07,
      "loss": 3.4933,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.646616541353384e-07,
      "loss": 3.3022,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.022556390977444e-07,
      "loss": 3.3384,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.398496240601504e-07,
      "loss": 3.2371,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.774436090225563e-07,
      "loss": 3.1935,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.0150375939849624e-06,
      "loss": 3.1365,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.0526315789473685e-06,
      "loss": 2.9064,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.0902255639097744e-06,
      "loss": 2.9232,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1278195488721805e-06,
      "loss": 2.774,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1654135338345866e-06,
      "loss": 2.679,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2030075187969925e-06,
      "loss": 2.5642,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2406015037593986e-06,
      "loss": 2.5722,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2781954887218045e-06,
      "loss": 2.5736,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3157894736842106e-06,
      "loss": 2.4032,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3533834586466167e-06,
      "loss": 2.4277,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3909774436090225e-06,
      "loss": 2.4344,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4285714285714286e-06,
      "loss": 2.1502,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4661654135338347e-06,
      "loss": 2.1301,
      "step": 39
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5037593984962406e-06,
      "loss": 2.2598,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5413533834586467e-06,
      "loss": 2.0539,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5789473684210526e-06,
      "loss": 1.9843,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6165413533834587e-06,
      "loss": 1.7659,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6541353383458648e-06,
      "loss": 1.8645,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6917293233082707e-06,
      "loss": 1.8412,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.7293233082706767e-06,
      "loss": 1.6839,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.7669172932330828e-06,
      "loss": 1.5168,
      "step": 47
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.8045112781954887e-06,
      "loss": 1.434,
      "step": 48
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.8421052631578948e-06,
      "loss": 1.5035,
      "step": 49
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.8796992481203007e-06,
      "loss": 1.4518,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.917293233082707e-06,
      "loss": 1.4297,
      "step": 51
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9548872180451127e-06,
      "loss": 1.2547,
      "step": 52
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9924812030075188e-06,
      "loss": 1.2797,
      "step": 53
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.030075187969925e-06,
      "loss": 1.0799,
      "step": 54
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.067669172932331e-06,
      "loss": 1.0939,
      "step": 55
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.105263157894737e-06,
      "loss": 1.2042,
      "step": 56
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.1428571428571427e-06,
      "loss": 1.1174,
      "step": 57
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.180451127819549e-06,
      "loss": 0.9656,
      "step": 58
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.218045112781955e-06,
      "loss": 0.9868,
      "step": 59
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.255639097744361e-06,
      "loss": 1.0085,
      "step": 60
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.293233082706767e-06,
      "loss": 0.8955,
      "step": 61
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.330827067669173e-06,
      "loss": 0.9749,
      "step": 62
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.368421052631579e-06,
      "loss": 1.0838,
      "step": 63
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.406015037593985e-06,
      "loss": 0.9319,
      "step": 64
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.443609022556391e-06,
      "loss": 0.8328,
      "step": 65
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.481203007518797e-06,
      "loss": 0.8296,
      "step": 66
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.5187969924812033e-06,
      "loss": 1.0123,
      "step": 67
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.556390977443609e-06,
      "loss": 0.8237,
      "step": 68
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.5939849624060154e-06,
      "loss": 0.8056,
      "step": 69
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.631578947368421e-06,
      "loss": 0.801,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.669172932330827e-06,
      "loss": 0.7226,
      "step": 71
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.7067669172932333e-06,
      "loss": 0.8122,
      "step": 72
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.7443609022556394e-06,
      "loss": 0.7645,
      "step": 73
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.781954887218045e-06,
      "loss": 0.7054,
      "step": 74
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.8195488721804516e-06,
      "loss": 0.7672,
      "step": 75
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.8571428571428573e-06,
      "loss": 0.6448,
      "step": 76
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.8947368421052634e-06,
      "loss": 0.7311,
      "step": 77
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.9323308270676694e-06,
      "loss": 0.7694,
      "step": 78
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.9699248120300755e-06,
      "loss": 0.665,
      "step": 79
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.007518796992481e-06,
      "loss": 0.7295,
      "step": 80
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.0451127819548877e-06,
      "loss": 0.7084,
      "step": 81
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.0827067669172934e-06,
      "loss": 0.7341,
      "step": 82
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.1203007518796995e-06,
      "loss": 0.6022,
      "step": 83
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.157894736842105e-06,
      "loss": 0.5784,
      "step": 84
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.1954887218045117e-06,
      "loss": 0.6094,
      "step": 85
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.2330827067669174e-06,
      "loss": 0.576,
      "step": 86
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.270676691729324e-06,
      "loss": 0.6607,
      "step": 87
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.3082706766917295e-06,
      "loss": 0.5783,
      "step": 88
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.3458646616541356e-06,
      "loss": 0.6067,
      "step": 89
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.3834586466165413e-06,
      "loss": 0.6504,
      "step": 90
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.421052631578948e-06,
      "loss": 0.572,
      "step": 91
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.4586466165413535e-06,
      "loss": 0.5327,
      "step": 92
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.4962406015037596e-06,
      "loss": 0.5118,
      "step": 93
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.5338345864661657e-06,
      "loss": 0.5211,
      "step": 94
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.5714285714285718e-06,
      "loss": 0.6022,
      "step": 95
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.6090225563909775e-06,
      "loss": 0.5538,
      "step": 96
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.646616541353384e-06,
      "loss": 0.5254,
      "step": 97
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.6842105263157896e-06,
      "loss": 0.5527,
      "step": 98
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.7218045112781957e-06,
      "loss": 0.5036,
      "step": 99
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.7593984962406014e-06,
      "loss": 0.5012,
      "step": 100
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.796992481203008e-06,
      "loss": 0.4962,
      "step": 101
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.834586466165414e-06,
      "loss": 0.522,
      "step": 102
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.87218045112782e-06,
      "loss": 0.4526,
      "step": 103
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.909774436090225e-06,
      "loss": 0.4576,
      "step": 104
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.947368421052632e-06,
      "loss": 0.5031,
      "step": 105
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.9849624060150376e-06,
      "loss": 0.5256,
      "step": 106
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.022556390977444e-06,
      "loss": 0.4607,
      "step": 107
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.06015037593985e-06,
      "loss": 0.4241,
      "step": 108
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.097744360902256e-06,
      "loss": 0.4645,
      "step": 109
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.135338345864662e-06,
      "loss": 0.4345,
      "step": 110
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.172932330827068e-06,
      "loss": 0.4664,
      "step": 111
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.210526315789474e-06,
      "loss": 0.3866,
      "step": 112
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.24812030075188e-06,
      "loss": 0.5332,
      "step": 113
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.2857142857142855e-06,
      "loss": 0.3868,
      "step": 114
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.323308270676692e-06,
      "loss": 0.4167,
      "step": 115
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.360902255639098e-06,
      "loss": 0.4143,
      "step": 116
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.398496240601504e-06,
      "loss": 0.4094,
      "step": 117
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.43609022556391e-06,
      "loss": 0.3819,
      "step": 118
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.473684210526316e-06,
      "loss": 0.4106,
      "step": 119
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.511278195488722e-06,
      "loss": 0.421,
      "step": 120
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.548872180451128e-06,
      "loss": 0.3777,
      "step": 121
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.586466165413534e-06,
      "loss": 0.4167,
      "step": 122
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.62406015037594e-06,
      "loss": 0.423,
      "step": 123
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.661654135338346e-06,
      "loss": 0.3979,
      "step": 124
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.6992481203007525e-06,
      "loss": 0.3865,
      "step": 125
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.736842105263158e-06,
      "loss": 0.3838,
      "step": 126
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.774436090225565e-06,
      "loss": 0.3824,
      "step": 127
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.81203007518797e-06,
      "loss": 0.4438,
      "step": 128
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.849624060150376e-06,
      "loss": 0.359,
      "step": 129
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.887218045112782e-06,
      "loss": 0.3582,
      "step": 130
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.924812030075188e-06,
      "loss": 0.4314,
      "step": 131
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.962406015037594e-06,
      "loss": 0.3402,
      "step": 132
    },
    {
      "epoch": 0.05,
      "learning_rate": 5e-06,
      "loss": 0.4071,
      "step": 133
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.0375939849624065e-06,
      "loss": 0.3506,
      "step": 134
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.075187969924813e-06,
      "loss": 0.3841,
      "step": 135
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.112781954887218e-06,
      "loss": 0.3405,
      "step": 136
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.150375939849625e-06,
      "loss": 0.4197,
      "step": 137
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.187969924812031e-06,
      "loss": 0.3365,
      "step": 138
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.225563909774437e-06,
      "loss": 0.3408,
      "step": 139
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.263157894736842e-06,
      "loss": 0.363,
      "step": 140
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.300751879699248e-06,
      "loss": 0.3616,
      "step": 141
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.338345864661654e-06,
      "loss": 0.3875,
      "step": 142
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.375939849624061e-06,
      "loss": 0.3251,
      "step": 143
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.413533834586467e-06,
      "loss": 0.39,
      "step": 144
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.451127819548873e-06,
      "loss": 0.3339,
      "step": 145
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.488721804511279e-06,
      "loss": 0.3384,
      "step": 146
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.526315789473685e-06,
      "loss": 0.3722,
      "step": 147
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.56390977443609e-06,
      "loss": 0.3241,
      "step": 148
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.601503759398497e-06,
      "loss": 0.3309,
      "step": 149
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.639097744360903e-06,
      "loss": 0.36,
      "step": 150
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.676691729323309e-06,
      "loss": 0.3188,
      "step": 151
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.7142857142857145e-06,
      "loss": 0.3347,
      "step": 152
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.751879699248121e-06,
      "loss": 0.3016,
      "step": 153
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.789473684210527e-06,
      "loss": 0.3217,
      "step": 154
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.827067669172934e-06,
      "loss": 0.3487,
      "step": 155
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.864661654135339e-06,
      "loss": 0.3153,
      "step": 156
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.902255639097745e-06,
      "loss": 0.3189,
      "step": 157
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.939849624060151e-06,
      "loss": 0.3654,
      "step": 158
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.977443609022557e-06,
      "loss": 0.3146,
      "step": 159
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.015037593984962e-06,
      "loss": 0.2973,
      "step": 160
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.0526315789473685e-06,
      "loss": 0.311,
      "step": 161
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.0902255639097755e-06,
      "loss": 0.3149,
      "step": 162
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.1278195488721816e-06,
      "loss": 0.3272,
      "step": 163
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.165413533834587e-06,
      "loss": 0.3521,
      "step": 164
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.203007518796993e-06,
      "loss": 0.3229,
      "step": 165
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.240601503759399e-06,
      "loss": 0.2524,
      "step": 166
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.278195488721806e-06,
      "loss": 0.332,
      "step": 167
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.31578947368421e-06,
      "loss": 0.3117,
      "step": 168
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.353383458646617e-06,
      "loss": 0.2854,
      "step": 169
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.390977443609023e-06,
      "loss": 0.2758,
      "step": 170
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.4285714285714295e-06,
      "loss": 0.2723,
      "step": 171
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.466165413533835e-06,
      "loss": 0.2833,
      "step": 172
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.503759398496241e-06,
      "loss": 0.2971,
      "step": 173
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.541353383458648e-06,
      "loss": 0.3246,
      "step": 174
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.578947368421054e-06,
      "loss": 0.2649,
      "step": 175
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.616541353383459e-06,
      "loss": 0.3183,
      "step": 176
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.654135338345865e-06,
      "loss": 0.3066,
      "step": 177
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.691729323308271e-06,
      "loss": 0.2997,
      "step": 178
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.729323308270677e-06,
      "loss": 0.2896,
      "step": 179
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.766917293233083e-06,
      "loss": 0.3057,
      "step": 180
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.8045112781954896e-06,
      "loss": 0.2622,
      "step": 181
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.842105263157896e-06,
      "loss": 0.2992,
      "step": 182
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.879699248120302e-06,
      "loss": 0.2964,
      "step": 183
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.917293233082707e-06,
      "loss": 0.2562,
      "step": 184
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.954887218045113e-06,
      "loss": 0.2687,
      "step": 185
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.992481203007519e-06,
      "loss": 0.2864,
      "step": 186
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.030075187969926e-06,
      "loss": 0.2962,
      "step": 187
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.067669172932331e-06,
      "loss": 0.2924,
      "step": 188
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.1052631578947375e-06,
      "loss": 0.2419,
      "step": 189
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.1428571428571436e-06,
      "loss": 0.3047,
      "step": 190
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.18045112781955e-06,
      "loss": 0.2453,
      "step": 191
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.218045112781955e-06,
      "loss": 0.2854,
      "step": 192
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.255639097744361e-06,
      "loss": 0.288,
      "step": 193
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.293233082706768e-06,
      "loss": 0.2648,
      "step": 194
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.330827067669174e-06,
      "loss": 0.2683,
      "step": 195
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.368421052631579e-06,
      "loss": 0.2801,
      "step": 196
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.406015037593985e-06,
      "loss": 0.286,
      "step": 197
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.4436090225563915e-06,
      "loss": 0.3148,
      "step": 198
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.481203007518798e-06,
      "loss": 0.3088,
      "step": 199
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.518796992481203e-06,
      "loss": 0.2654,
      "step": 200
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.55639097744361e-06,
      "loss": 0.2576,
      "step": 201
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.593984962406016e-06,
      "loss": 0.2846,
      "step": 202
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.631578947368423e-06,
      "loss": 0.2835,
      "step": 203
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.669172932330828e-06,
      "loss": 0.2665,
      "step": 204
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.706766917293233e-06,
      "loss": 0.2785,
      "step": 205
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.74436090225564e-06,
      "loss": 0.2625,
      "step": 206
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.781954887218045e-06,
      "loss": 0.2795,
      "step": 207
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.81954887218045e-06,
      "loss": 0.2997,
      "step": 208
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.857142857142858e-06,
      "loss": 0.2871,
      "step": 209
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.894736842105265e-06,
      "loss": 0.269,
      "step": 210
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.93233082706767e-06,
      "loss": 0.2804,
      "step": 211
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.969924812030075e-06,
      "loss": 0.2653,
      "step": 212
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.007518796992482e-06,
      "loss": 0.2759,
      "step": 213
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.045112781954887e-06,
      "loss": 0.2594,
      "step": 214
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.082706766917294e-06,
      "loss": 0.2166,
      "step": 215
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.1203007518797e-06,
      "loss": 0.2891,
      "step": 216
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.157894736842106e-06,
      "loss": 0.283,
      "step": 217
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.195488721804512e-06,
      "loss": 0.2819,
      "step": 218
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.233082706766919e-06,
      "loss": 0.2997,
      "step": 219
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.270676691729324e-06,
      "loss": 0.2766,
      "step": 220
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.308270676691729e-06,
      "loss": 0.2791,
      "step": 221
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.345864661654136e-06,
      "loss": 0.2825,
      "step": 222
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.383458646616543e-06,
      "loss": 0.3174,
      "step": 223
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.421052631578948e-06,
      "loss": 0.2798,
      "step": 224
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.458646616541353e-06,
      "loss": 0.2557,
      "step": 225
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.49624060150376e-06,
      "loss": 0.2899,
      "step": 226
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.533834586466166e-06,
      "loss": 0.263,
      "step": 227
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.571428571428571e-06,
      "loss": 0.2542,
      "step": 228
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.609022556390978e-06,
      "loss": 0.2682,
      "step": 229
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.646616541353385e-06,
      "loss": 0.2839,
      "step": 230
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.68421052631579e-06,
      "loss": 0.2405,
      "step": 231
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.721804511278195e-06,
      "loss": 0.2685,
      "step": 232
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.759398496240602e-06,
      "loss": 0.2399,
      "step": 233
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.796992481203007e-06,
      "loss": 0.2849,
      "step": 234
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.834586466165414e-06,
      "loss": 0.2529,
      "step": 235
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.87218045112782e-06,
      "loss": 0.2572,
      "step": 236
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.909774436090227e-06,
      "loss": 0.2428,
      "step": 237
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.947368421052632e-06,
      "loss": 0.2929,
      "step": 238
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.984962406015039e-06,
      "loss": 0.2522,
      "step": 239
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.022556390977444e-06,
      "loss": 0.2663,
      "step": 240
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.06015037593985e-06,
      "loss": 0.2696,
      "step": 241
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.097744360902256e-06,
      "loss": 0.2287,
      "step": 242
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.135338345864663e-06,
      "loss": 0.3045,
      "step": 243
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.172932330827068e-06,
      "loss": 0.2291,
      "step": 244
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.210526315789474e-06,
      "loss": 0.274,
      "step": 245
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.24812030075188e-06,
      "loss": 0.3251,
      "step": 246
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.285714285714288e-06,
      "loss": 0.2509,
      "step": 247
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.323308270676693e-06,
      "loss": 0.2385,
      "step": 248
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.360902255639098e-06,
      "loss": 0.2837,
      "step": 249
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.398496240601505e-06,
      "loss": 0.2653,
      "step": 250
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.43609022556391e-06,
      "loss": 0.3059,
      "step": 251
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.473684210526315e-06,
      "loss": 0.2792,
      "step": 252
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.511278195488722e-06,
      "loss": 0.2537,
      "step": 253
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.54887218045113e-06,
      "loss": 0.2403,
      "step": 254
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.586466165413535e-06,
      "loss": 0.2577,
      "step": 255
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.62406015037594e-06,
      "loss": 0.275,
      "step": 256
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.661654135338347e-06,
      "loss": 0.209,
      "step": 257
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.699248120300752e-06,
      "loss": 0.2827,
      "step": 258
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.736842105263159e-06,
      "loss": 0.2865,
      "step": 259
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.774436090225564e-06,
      "loss": 0.2627,
      "step": 260
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.812030075187971e-06,
      "loss": 0.2514,
      "step": 261
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.849624060150376e-06,
      "loss": 0.2403,
      "step": 262
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.887218045112783e-06,
      "loss": 0.264,
      "step": 263
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.924812030075189e-06,
      "loss": 0.215,
      "step": 264
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.962406015037594e-06,
      "loss": 0.2291,
      "step": 265
    },
    {
      "epoch": 0.1,
      "learning_rate": 1e-05,
      "loss": 0.2292,
      "step": 266
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0037593984962406e-05,
      "loss": 0.2373,
      "step": 267
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0075187969924813e-05,
      "loss": 0.2598,
      "step": 268
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0112781954887218e-05,
      "loss": 0.2665,
      "step": 269
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0150375939849625e-05,
      "loss": 0.2627,
      "step": 270
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.018796992481203e-05,
      "loss": 0.2538,
      "step": 271
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0225563909774436e-05,
      "loss": 0.248,
      "step": 272
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0263157894736844e-05,
      "loss": 0.2344,
      "step": 273
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.030075187969925e-05,
      "loss": 0.2275,
      "step": 274
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0338345864661655e-05,
      "loss": 0.2241,
      "step": 275
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0375939849624062e-05,
      "loss": 0.2244,
      "step": 276
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0413533834586467e-05,
      "loss": 0.2471,
      "step": 277
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0451127819548874e-05,
      "loss": 0.2622,
      "step": 278
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.048872180451128e-05,
      "loss": 0.2697,
      "step": 279
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0526315789473684e-05,
      "loss": 0.222,
      "step": 280
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0563909774436091e-05,
      "loss": 0.2347,
      "step": 281
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0601503759398497e-05,
      "loss": 0.2451,
      "step": 282
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0639097744360902e-05,
      "loss": 0.2631,
      "step": 283
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0676691729323309e-05,
      "loss": 0.2456,
      "step": 284
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0714285714285714e-05,
      "loss": 0.2465,
      "step": 285
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0751879699248123e-05,
      "loss": 0.2829,
      "step": 286
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0789473684210528e-05,
      "loss": 0.2096,
      "step": 287
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0827067669172933e-05,
      "loss": 0.2283,
      "step": 288
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.086466165413534e-05,
      "loss": 0.2284,
      "step": 289
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0902255639097745e-05,
      "loss": 0.2154,
      "step": 290
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.093984962406015e-05,
      "loss": 0.2399,
      "step": 291
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0977443609022558e-05,
      "loss": 0.2143,
      "step": 292
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1015037593984963e-05,
      "loss": 0.221,
      "step": 293
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.105263157894737e-05,
      "loss": 0.2032,
      "step": 294
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1090225563909775e-05,
      "loss": 0.2075,
      "step": 295
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.112781954887218e-05,
      "loss": 0.2108,
      "step": 296
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1165413533834589e-05,
      "loss": 0.2603,
      "step": 297
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1203007518796994e-05,
      "loss": 0.2126,
      "step": 298
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1240601503759398e-05,
      "loss": 0.2062,
      "step": 299
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1278195488721806e-05,
      "loss": 0.2108,
      "step": 300
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1315789473684212e-05,
      "loss": 0.1967,
      "step": 301
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1353383458646619e-05,
      "loss": 0.2508,
      "step": 302
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1390977443609024e-05,
      "loss": 0.2303,
      "step": 303
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1428571428571429e-05,
      "loss": 0.227,
      "step": 304
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1466165413533836e-05,
      "loss": 0.23,
      "step": 305
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1503759398496241e-05,
      "loss": 0.2477,
      "step": 306
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1541353383458646e-05,
      "loss": 0.2777,
      "step": 307
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1578947368421053e-05,
      "loss": 0.2286,
      "step": 308
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1616541353383459e-05,
      "loss": 0.2508,
      "step": 309
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1654135338345867e-05,
      "loss": 0.2524,
      "step": 310
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1691729323308273e-05,
      "loss": 0.1947,
      "step": 311
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1729323308270678e-05,
      "loss": 0.2212,
      "step": 312
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1766917293233085e-05,
      "loss": 0.2324,
      "step": 313
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.180451127819549e-05,
      "loss": 0.1924,
      "step": 314
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1842105263157895e-05,
      "loss": 0.2064,
      "step": 315
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1879699248120302e-05,
      "loss": 0.23,
      "step": 316
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1917293233082707e-05,
      "loss": 0.2148,
      "step": 317
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1954887218045114e-05,
      "loss": 0.2573,
      "step": 318
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.199248120300752e-05,
      "loss": 0.238,
      "step": 319
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2030075187969925e-05,
      "loss": 0.2369,
      "step": 320
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2067669172932332e-05,
      "loss": 0.2323,
      "step": 321
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2105263157894737e-05,
      "loss": 0.2183,
      "step": 322
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2142857142857142e-05,
      "loss": 0.1754,
      "step": 323
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2180451127819551e-05,
      "loss": 0.2031,
      "step": 324
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2218045112781956e-05,
      "loss": 0.1861,
      "step": 325
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2255639097744363e-05,
      "loss": 0.2222,
      "step": 326
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2293233082706768e-05,
      "loss": 0.2267,
      "step": 327
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2330827067669174e-05,
      "loss": 0.2453,
      "step": 328
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.236842105263158e-05,
      "loss": 0.2201,
      "step": 329
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2406015037593986e-05,
      "loss": 0.1929,
      "step": 330
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2443609022556391e-05,
      "loss": 0.2176,
      "step": 331
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2481203007518798e-05,
      "loss": 0.26,
      "step": 332
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2518796992481203e-05,
      "loss": 0.2407,
      "step": 333
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2556390977443612e-05,
      "loss": 0.2156,
      "step": 334
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2593984962406015e-05,
      "loss": 0.2585,
      "step": 335
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.263157894736842e-05,
      "loss": 0.2134,
      "step": 336
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.266917293233083e-05,
      "loss": 0.2533,
      "step": 337
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2706766917293235e-05,
      "loss": 0.2201,
      "step": 338
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.274436090225564e-05,
      "loss": 0.2182,
      "step": 339
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2781954887218047e-05,
      "loss": 0.2185,
      "step": 340
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2819548872180452e-05,
      "loss": 0.2175,
      "step": 341
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2857142857142859e-05,
      "loss": 0.2627,
      "step": 342
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2894736842105264e-05,
      "loss": 0.1952,
      "step": 343
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.293233082706767e-05,
      "loss": 0.2795,
      "step": 344
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2969924812030076e-05,
      "loss": 0.207,
      "step": 345
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3007518796992482e-05,
      "loss": 0.2254,
      "step": 346
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3045112781954887e-05,
      "loss": 0.2203,
      "step": 347
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3082706766917295e-05,
      "loss": 0.1979,
      "step": 348
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3120300751879699e-05,
      "loss": 0.2507,
      "step": 349
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3157894736842108e-05,
      "loss": 0.2005,
      "step": 350
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3195488721804513e-05,
      "loss": 0.2143,
      "step": 351
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3233082706766918e-05,
      "loss": 0.2539,
      "step": 352
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3270676691729325e-05,
      "loss": 0.2099,
      "step": 353
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.330827067669173e-05,
      "loss": 0.1941,
      "step": 354
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3345864661654136e-05,
      "loss": 0.2001,
      "step": 355
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3383458646616543e-05,
      "loss": 0.1863,
      "step": 356
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3421052631578948e-05,
      "loss": 0.1913,
      "step": 357
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3458646616541355e-05,
      "loss": 0.224,
      "step": 358
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.349624060150376e-05,
      "loss": 0.217,
      "step": 359
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3533834586466165e-05,
      "loss": 0.2283,
      "step": 360
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3571428571428574e-05,
      "loss": 0.1818,
      "step": 361
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3609022556390979e-05,
      "loss": 0.2512,
      "step": 362
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3646616541353384e-05,
      "loss": 0.2114,
      "step": 363
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3684210526315791e-05,
      "loss": 0.1776,
      "step": 364
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3721804511278197e-05,
      "loss": 0.2589,
      "step": 365
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3759398496240603e-05,
      "loss": 0.2103,
      "step": 366
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3796992481203009e-05,
      "loss": 0.2015,
      "step": 367
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3834586466165414e-05,
      "loss": 0.2057,
      "step": 368
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3872180451127821e-05,
      "loss": 0.2432,
      "step": 369
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3909774436090226e-05,
      "loss": 0.2272,
      "step": 370
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3947368421052631e-05,
      "loss": 0.205,
      "step": 371
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3984962406015038e-05,
      "loss": 0.1966,
      "step": 372
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4022556390977444e-05,
      "loss": 0.2377,
      "step": 373
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4060150375939852e-05,
      "loss": 0.2156,
      "step": 374
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4097744360902257e-05,
      "loss": 0.219,
      "step": 375
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4135338345864663e-05,
      "loss": 0.2197,
      "step": 376
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.417293233082707e-05,
      "loss": 0.207,
      "step": 377
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4210526315789475e-05,
      "loss": 0.2529,
      "step": 378
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.424812030075188e-05,
      "loss": 0.2484,
      "step": 379
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4285714285714287e-05,
      "loss": 0.2185,
      "step": 380
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4323308270676692e-05,
      "loss": 0.212,
      "step": 381
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.43609022556391e-05,
      "loss": 0.2188,
      "step": 382
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4398496240601505e-05,
      "loss": 0.1897,
      "step": 383
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.443609022556391e-05,
      "loss": 0.2042,
      "step": 384
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4473684210526317e-05,
      "loss": 0.2162,
      "step": 385
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4511278195488722e-05,
      "loss": 0.2116,
      "step": 386
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4548872180451127e-05,
      "loss": 0.2021,
      "step": 387
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4586466165413536e-05,
      "loss": 0.2,
      "step": 388
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4624060150375941e-05,
      "loss": 0.2078,
      "step": 389
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4661654135338348e-05,
      "loss": 0.1886,
      "step": 390
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4699248120300753e-05,
      "loss": 0.2177,
      "step": 391
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4736842105263159e-05,
      "loss": 0.2344,
      "step": 392
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4774436090225566e-05,
      "loss": 0.1799,
      "step": 393
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.481203007518797e-05,
      "loss": 0.215,
      "step": 394
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4849624060150376e-05,
      "loss": 0.2086,
      "step": 395
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4887218045112783e-05,
      "loss": 0.1858,
      "step": 396
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4924812030075188e-05,
      "loss": 0.1837,
      "step": 397
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4962406015037597e-05,
      "loss": 0.219,
      "step": 398
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 0.1863,
      "step": 399
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5037593984962406e-05,
      "loss": 0.2307,
      "step": 400
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5075187969924814e-05,
      "loss": 0.1815,
      "step": 401
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.511278195488722e-05,
      "loss": 0.2229,
      "step": 402
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5150375939849626e-05,
      "loss": 0.2299,
      "step": 403
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5187969924812032e-05,
      "loss": 0.2126,
      "step": 404
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5225563909774437e-05,
      "loss": 0.2236,
      "step": 405
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5263157894736846e-05,
      "loss": 0.1931,
      "step": 406
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.530075187969925e-05,
      "loss": 0.1931,
      "step": 407
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5338345864661656e-05,
      "loss": 0.2114,
      "step": 408
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.537593984962406e-05,
      "loss": 0.1716,
      "step": 409
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5413533834586467e-05,
      "loss": 0.2028,
      "step": 410
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5451127819548875e-05,
      "loss": 0.2347,
      "step": 411
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.548872180451128e-05,
      "loss": 0.2203,
      "step": 412
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5526315789473686e-05,
      "loss": 0.1986,
      "step": 413
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.556390977443609e-05,
      "loss": 0.1767,
      "step": 414
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5601503759398496e-05,
      "loss": 0.1972,
      "step": 415
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.56390977443609e-05,
      "loss": 0.2092,
      "step": 416
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.567669172932331e-05,
      "loss": 0.2131,
      "step": 417
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5714285714285715e-05,
      "loss": 0.2232,
      "step": 418
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5751879699248124e-05,
      "loss": 0.1998,
      "step": 419
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.578947368421053e-05,
      "loss": 0.2033,
      "step": 420
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5827067669172934e-05,
      "loss": 0.2296,
      "step": 421
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.586466165413534e-05,
      "loss": 0.228,
      "step": 422
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5902255639097745e-05,
      "loss": 0.2078,
      "step": 423
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.593984962406015e-05,
      "loss": 0.1908,
      "step": 424
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.597744360902256e-05,
      "loss": 0.2122,
      "step": 425
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6015037593984964e-05,
      "loss": 0.1817,
      "step": 426
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.605263157894737e-05,
      "loss": 0.1789,
      "step": 427
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6090225563909775e-05,
      "loss": 0.1833,
      "step": 428
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.612781954887218e-05,
      "loss": 0.1779,
      "step": 429
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.616541353383459e-05,
      "loss": 0.1863,
      "step": 430
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6203007518796994e-05,
      "loss": 0.1832,
      "step": 431
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.62406015037594e-05,
      "loss": 0.2044,
      "step": 432
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6278195488721808e-05,
      "loss": 0.1874,
      "step": 433
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6315789473684213e-05,
      "loss": 0.2036,
      "step": 434
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6353383458646618e-05,
      "loss": 0.1697,
      "step": 435
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6390977443609023e-05,
      "loss": 0.2454,
      "step": 436
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.642857142857143e-05,
      "loss": 0.1898,
      "step": 437
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6466165413533837e-05,
      "loss": 0.2256,
      "step": 438
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6503759398496242e-05,
      "loss": 0.2308,
      "step": 439
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6541353383458648e-05,
      "loss": 0.1806,
      "step": 440
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6578947368421053e-05,
      "loss": 0.184,
      "step": 441
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6616541353383458e-05,
      "loss": 0.1946,
      "step": 442
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6654135338345867e-05,
      "loss": 0.1964,
      "step": 443
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6691729323308272e-05,
      "loss": 0.1942,
      "step": 444
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6729323308270677e-05,
      "loss": 0.1986,
      "step": 445
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6766917293233086e-05,
      "loss": 0.1633,
      "step": 446
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.680451127819549e-05,
      "loss": 0.2064,
      "step": 447
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6842105263157896e-05,
      "loss": 0.1963,
      "step": 448
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6879699248120302e-05,
      "loss": 0.2085,
      "step": 449
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6917293233082707e-05,
      "loss": 0.1828,
      "step": 450
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6954887218045116e-05,
      "loss": 0.1856,
      "step": 451
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.699248120300752e-05,
      "loss": 0.225,
      "step": 452
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7030075187969926e-05,
      "loss": 0.2228,
      "step": 453
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.706766917293233e-05,
      "loss": 0.2368,
      "step": 454
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7105263157894737e-05,
      "loss": 0.1903,
      "step": 455
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7142857142857142e-05,
      "loss": 0.1776,
      "step": 456
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.718045112781955e-05,
      "loss": 0.1687,
      "step": 457
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7218045112781956e-05,
      "loss": 0.1858,
      "step": 458
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7255639097744364e-05,
      "loss": 0.1923,
      "step": 459
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.729323308270677e-05,
      "loss": 0.1695,
      "step": 460
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7330827067669175e-05,
      "loss": 0.1702,
      "step": 461
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.736842105263158e-05,
      "loss": 0.2131,
      "step": 462
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7406015037593985e-05,
      "loss": 0.1965,
      "step": 463
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.744360902255639e-05,
      "loss": 0.163,
      "step": 464
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.74812030075188e-05,
      "loss": 0.172,
      "step": 465
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7518796992481204e-05,
      "loss": 0.194,
      "step": 466
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7556390977443613e-05,
      "loss": 0.2191,
      "step": 467
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7593984962406015e-05,
      "loss": 0.2074,
      "step": 468
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.763157894736842e-05,
      "loss": 0.2043,
      "step": 469
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.766917293233083e-05,
      "loss": 0.2151,
      "step": 470
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7706766917293234e-05,
      "loss": 0.1798,
      "step": 471
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.774436090225564e-05,
      "loss": 0.2345,
      "step": 472
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7781954887218048e-05,
      "loss": 0.1929,
      "step": 473
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7819548872180453e-05,
      "loss": 0.1881,
      "step": 474
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.785714285714286e-05,
      "loss": 0.196,
      "step": 475
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7894736842105264e-05,
      "loss": 0.2126,
      "step": 476
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.793233082706767e-05,
      "loss": 0.1898,
      "step": 477
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7969924812030078e-05,
      "loss": 0.1898,
      "step": 478
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8007518796992483e-05,
      "loss": 0.2029,
      "step": 479
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8045112781954888e-05,
      "loss": 0.2131,
      "step": 480
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8082706766917297e-05,
      "loss": 0.2086,
      "step": 481
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.81203007518797e-05,
      "loss": 0.2152,
      "step": 482
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8157894736842107e-05,
      "loss": 0.196,
      "step": 483
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8195488721804512e-05,
      "loss": 0.2046,
      "step": 484
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8233082706766918e-05,
      "loss": 0.1652,
      "step": 485
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8270676691729326e-05,
      "loss": 0.1968,
      "step": 486
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.830827067669173e-05,
      "loss": 0.19,
      "step": 487
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8345864661654137e-05,
      "loss": 0.1998,
      "step": 488
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8383458646616542e-05,
      "loss": 0.2276,
      "step": 489
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8421052631578947e-05,
      "loss": 0.2135,
      "step": 490
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8458646616541356e-05,
      "loss": 0.2048,
      "step": 491
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.849624060150376e-05,
      "loss": 0.1805,
      "step": 492
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8533834586466166e-05,
      "loss": 0.1964,
      "step": 493
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8571428571428575e-05,
      "loss": 0.1933,
      "step": 494
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.860902255639098e-05,
      "loss": 0.1691,
      "step": 495
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8646616541353386e-05,
      "loss": 0.176,
      "step": 496
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.868421052631579e-05,
      "loss": 0.1908,
      "step": 497
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8721804511278196e-05,
      "loss": 0.2153,
      "step": 498
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8759398496240605e-05,
      "loss": 0.2101,
      "step": 499
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.879699248120301e-05,
      "loss": 0.1645,
      "step": 500
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8834586466165415e-05,
      "loss": 0.1727,
      "step": 501
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.887218045112782e-05,
      "loss": 0.1858,
      "step": 502
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8909774436090226e-05,
      "loss": 0.2101,
      "step": 503
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.894736842105263e-05,
      "loss": 0.2446,
      "step": 504
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.898496240601504e-05,
      "loss": 0.1965,
      "step": 505
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9022556390977445e-05,
      "loss": 0.1717,
      "step": 506
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9060150375939854e-05,
      "loss": 0.1803,
      "step": 507
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.909774436090226e-05,
      "loss": 0.1778,
      "step": 508
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9135338345864664e-05,
      "loss": 0.1863,
      "step": 509
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.917293233082707e-05,
      "loss": 0.1943,
      "step": 510
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9210526315789474e-05,
      "loss": 0.174,
      "step": 511
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.924812030075188e-05,
      "loss": 0.1731,
      "step": 512
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.928571428571429e-05,
      "loss": 0.2278,
      "step": 513
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9323308270676694e-05,
      "loss": 0.2369,
      "step": 514
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.93609022556391e-05,
      "loss": 0.21,
      "step": 515
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9398496240601504e-05,
      "loss": 0.195,
      "step": 516
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.943609022556391e-05,
      "loss": 0.2,
      "step": 517
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9473684210526318e-05,
      "loss": 0.1854,
      "step": 518
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9511278195488723e-05,
      "loss": 0.1899,
      "step": 519
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.954887218045113e-05,
      "loss": 0.1911,
      "step": 520
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9586466165413537e-05,
      "loss": 0.1784,
      "step": 521
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9624060150375942e-05,
      "loss": 0.1885,
      "step": 522
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9661654135338348e-05,
      "loss": 0.1986,
      "step": 523
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9699248120300753e-05,
      "loss": 0.1918,
      "step": 524
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9736842105263158e-05,
      "loss": 0.2215,
      "step": 525
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9774436090225567e-05,
      "loss": 0.2094,
      "step": 526
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9812030075187972e-05,
      "loss": 0.1919,
      "step": 527
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9849624060150377e-05,
      "loss": 0.2078,
      "step": 528
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9887218045112783e-05,
      "loss": 0.2101,
      "step": 529
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9924812030075188e-05,
      "loss": 0.1997,
      "step": 530
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9962406015037596e-05,
      "loss": 0.1796,
      "step": 531
    },
    {
      "epoch": 0.2,
      "learning_rate": 2e-05,
      "loss": 0.1989,
      "step": 532
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999997842004225e-05,
      "loss": 0.1912,
      "step": 533
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999991368017825e-05,
      "loss": 0.19,
      "step": 534
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999980578043595e-05,
      "loss": 0.1756,
      "step": 535
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999965472086192e-05,
      "loss": 0.1978,
      "step": 536
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999946050152137e-05,
      "loss": 0.2317,
      "step": 537
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999922312249814e-05,
      "loss": 0.1832,
      "step": 538
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999894258389462e-05,
      "loss": 0.2086,
      "step": 539
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999861888583198e-05,
      "loss": 0.1945,
      "step": 540
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999825202844983e-05,
      "loss": 0.1688,
      "step": 541
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999784201190658e-05,
      "loss": 0.202,
      "step": 542
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999973888363792e-05,
      "loss": 0.1687,
      "step": 543
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9999689250206324e-05,
      "loss": 0.2258,
      "step": 544
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999963530091729e-05,
      "loss": 0.1737,
      "step": 545
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999577035794106e-05,
      "loss": 0.1542,
      "step": 546
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999951445486192e-05,
      "loss": 0.1724,
      "step": 547
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999447558147738e-05,
      "loss": 0.1736,
      "step": 548
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999376345680435e-05,
      "loss": 0.1795,
      "step": 549
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999300817490745e-05,
      "loss": 0.1821,
      "step": 550
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999220973611272e-05,
      "loss": 0.2053,
      "step": 551
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999136814076465e-05,
      "loss": 0.1849,
      "step": 552
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9999048338922657e-05,
      "loss": 0.1461,
      "step": 553
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998955548188027e-05,
      "loss": 0.2237,
      "step": 554
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999885844191263e-05,
      "loss": 0.2208,
      "step": 555
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998757020138372e-05,
      "loss": 0.1528,
      "step": 556
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998651282909033e-05,
      "loss": 0.2288,
      "step": 557
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998541230270237e-05,
      "loss": 0.1946,
      "step": 558
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998426862269497e-05,
      "loss": 0.187,
      "step": 559
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998308178956164e-05,
      "loss": 0.2207,
      "step": 560
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998185180381467e-05,
      "loss": 0.1743,
      "step": 561
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9998057866598487e-05,
      "loss": 0.1795,
      "step": 562
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997926237662176e-05,
      "loss": 0.2085,
      "step": 563
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997790293629344e-05,
      "loss": 0.176,
      "step": 564
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997650034558666e-05,
      "loss": 0.2204,
      "step": 565
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997505460510675e-05,
      "loss": 0.2319,
      "step": 566
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997356571547772e-05,
      "loss": 0.182,
      "step": 567
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997203367734218e-05,
      "loss": 0.1951,
      "step": 568
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9997045849136128e-05,
      "loss": 0.2011,
      "step": 569
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9996884015821492e-05,
      "loss": 0.2067,
      "step": 570
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9996717867860164e-05,
      "loss": 0.1551,
      "step": 571
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999654740532384e-05,
      "loss": 0.1824,
      "step": 572
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9996372628286106e-05,
      "loss": 0.1841,
      "step": 573
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9996193536822378e-05,
      "loss": 0.1755,
      "step": 574
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999601013100997e-05,
      "loss": 0.2037,
      "step": 575
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999582241092803e-05,
      "loss": 0.1926,
      "step": 576
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9995630376657575e-05,
      "loss": 0.1838,
      "step": 577
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9995434028281493e-05,
      "loss": 0.189,
      "step": 578
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9995233365884526e-05,
      "loss": 0.1826,
      "step": 579
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999502838955328e-05,
      "loss": 0.1967,
      "step": 580
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9994819099376222e-05,
      "loss": 0.2108,
      "step": 581
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999460549544368e-05,
      "loss": 0.1963,
      "step": 582
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999438757784785e-05,
      "loss": 0.1976,
      "step": 583
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999416534668278e-05,
      "loss": 0.1721,
      "step": 584
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9993938802044387e-05,
      "loss": 0.2119,
      "step": 585
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9993707944030448e-05,
      "loss": 0.1846,
      "step": 586
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.99934727727406e-05,
      "loss": 0.1787,
      "step": 587
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999323328827634e-05,
      "loss": 0.2017,
      "step": 588
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9992989490741037e-05,
      "loss": 0.1815,
      "step": 589
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9992741380239905e-05,
      "loss": 0.1784,
      "step": 590
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9992488956880035e-05,
      "loss": 0.1689,
      "step": 591
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999223222077037e-05,
      "loss": 0.1871,
      "step": 592
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9991971172021717e-05,
      "loss": 0.2017,
      "step": 593
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9991705810746748e-05,
      "loss": 0.2089,
      "step": 594
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999143613705998e-05,
      "loss": 0.1823,
      "step": 595
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9991162151077822e-05,
      "loss": 0.2198,
      "step": 596
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9990883852918514e-05,
      "loss": 0.1838,
      "step": 597
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9990601242702174e-05,
      "loss": 0.1767,
      "step": 598
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9990314320550778e-05,
      "loss": 0.1738,
      "step": 599
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.999002308658815e-05,
      "loss": 0.193,
      "step": 600
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998972754094e-05,
      "loss": 0.1835,
      "step": 601
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998942768373388e-05,
      "loss": 0.1853,
      "step": 602
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998912351509921e-05,
      "loss": 0.2123,
      "step": 603
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998881503516726e-05,
      "loss": 0.1782,
      "step": 604
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9988502244071184e-05,
      "loss": 0.1806,
      "step": 605
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9988185141945973e-05,
      "loss": 0.1364,
      "step": 606
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9987863728928488e-05,
      "loss": 0.1861,
      "step": 607
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9987538005157453e-05,
      "loss": 0.2132,
      "step": 608
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9987207970773453e-05,
      "loss": 0.2037,
      "step": 609
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9986873625918923e-05,
      "loss": 0.1917,
      "step": 610
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9986534970738174e-05,
      "loss": 0.234,
      "step": 611
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998619200537736e-05,
      "loss": 0.2076,
      "step": 612
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9985844729984515e-05,
      "loss": 0.177,
      "step": 613
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998549314470952e-05,
      "loss": 0.2019,
      "step": 614
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998513724970411e-05,
      "loss": 0.1943,
      "step": 615
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.99847770451219e-05,
      "loss": 0.1944,
      "step": 616
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998441253111835e-05,
      "loss": 0.2035,
      "step": 617
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9984043707850782e-05,
      "loss": 0.2011,
      "step": 618
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9983670575478383e-05,
      "loss": 0.1665,
      "step": 619
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9983293134162195e-05,
      "loss": 0.1793,
      "step": 620
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998291138406512e-05,
      "loss": 0.1894,
      "step": 621
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9982525325351924e-05,
      "loss": 0.1915,
      "step": 622
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998213495818923e-05,
      "loss": 0.1879,
      "step": 623
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998174028274551e-05,
      "loss": 0.1665,
      "step": 624
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9981341299191122e-05,
      "loss": 0.1764,
      "step": 625
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9980938007698257e-05,
      "loss": 0.1991,
      "step": 626
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9980530408440976e-05,
      "loss": 0.1924,
      "step": 627
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.99801185015952e-05,
      "loss": 0.2007,
      "step": 628
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9979702287338705e-05,
      "loss": 0.179,
      "step": 629
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9979281765851133e-05,
      "loss": 0.1931,
      "step": 630
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.997885693731398e-05,
      "loss": 0.1624,
      "step": 631
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.99784278019106e-05,
      "loss": 0.1414,
      "step": 632
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9977994359826204e-05,
      "loss": 0.2151,
      "step": 633
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9977556611247872e-05,
      "loss": 0.1772,
      "step": 634
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9977114556364535e-05,
      "loss": 0.1808,
      "step": 635
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.997666819536698e-05,
      "loss": 0.1674,
      "step": 636
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9976217528447854e-05,
      "loss": 0.1864,
      "step": 637
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9975762555801674e-05,
      "loss": 0.1699,
      "step": 638
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9975303277624793e-05,
      "loss": 0.1835,
      "step": 639
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9974839694115444e-05,
      "loss": 0.1826,
      "step": 640
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.997437180547371e-05,
      "loss": 0.1979,
      "step": 641
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9973899611901527e-05,
      "loss": 0.1483,
      "step": 642
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9973423113602694e-05,
      "loss": 0.1504,
      "step": 643
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9972942310782868e-05,
      "loss": 0.2081,
      "step": 644
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.997245720364956e-05,
      "loss": 0.2133,
      "step": 645
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9971967792412154e-05,
      "loss": 0.1831,
      "step": 646
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9971474077281862e-05,
      "loss": 0.1824,
      "step": 647
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9970976058471782e-05,
      "loss": 0.2033,
      "step": 648
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9970473736196857e-05,
      "loss": 0.2514,
      "step": 649
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9969967110673887e-05,
      "loss": 0.1731,
      "step": 650
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9969456182121532e-05,
      "loss": 0.1743,
      "step": 651
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9968940950760308e-05,
      "loss": 0.1786,
      "step": 652
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9968421416812588e-05,
      "loss": 0.1903,
      "step": 653
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9967897580502606e-05,
      "loss": 0.1606,
      "step": 654
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9967369442056442e-05,
      "loss": 0.171,
      "step": 655
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.996683700170205e-05,
      "loss": 0.2083,
      "step": 656
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.996630025966922e-05,
      "loss": 0.1837,
      "step": 657
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9965759216189617e-05,
      "loss": 0.2001,
      "step": 658
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9965213871496755e-05,
      "loss": 0.1535,
      "step": 659
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9964664225825997e-05,
      "loss": 0.1873,
      "step": 660
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9964110279414578e-05,
      "loss": 0.1331,
      "step": 661
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9963552032501578e-05,
      "loss": 0.1755,
      "step": 662
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9962989485327932e-05,
      "loss": 0.1733,
      "step": 663
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.996242263813644e-05,
      "loss": 0.1841,
      "step": 664
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9961851491171752e-05,
      "loss": 0.1819,
      "step": 665
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9961276044680374e-05,
      "loss": 0.1793,
      "step": 666
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9960696298910666e-05,
      "loss": 0.18,
      "step": 667
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9960112254112847e-05,
      "loss": 0.2051,
      "step": 668
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9959523910538992e-05,
      "loss": 0.1591,
      "step": 669
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9958931268443028e-05,
      "loss": 0.1994,
      "step": 670
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9958334328080737e-05,
      "loss": 0.1597,
      "step": 671
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.995773308970976e-05,
      "loss": 0.1757,
      "step": 672
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9957127553589597e-05,
      "loss": 0.1427,
      "step": 673
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9956517719981587e-05,
      "loss": 0.1544,
      "step": 674
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.995590358914894e-05,
      "loss": 0.1717,
      "step": 675
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9955285161356708e-05,
      "loss": 0.172,
      "step": 676
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9954662436871808e-05,
      "loss": 0.2103,
      "step": 677
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9954035415963008e-05,
      "loss": 0.1759,
      "step": 678
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.995340409890093e-05,
      "loss": 0.1762,
      "step": 679
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.995276848595805e-05,
      "loss": 0.1711,
      "step": 680
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9952128577408697e-05,
      "loss": 0.2027,
      "step": 681
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.995148437352905e-05,
      "loss": 0.2189,
      "step": 682
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.995083587459716e-05,
      "loss": 0.1824,
      "step": 683
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.99501830808929e-05,
      "loss": 0.1826,
      "step": 684
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9949525992698034e-05,
      "loss": 0.1801,
      "step": 685
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.994886461029615e-05,
      "loss": 0.1647,
      "step": 686
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.99481989339727e-05,
      "loss": 0.179,
      "step": 687
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9947528964014992e-05,
      "loss": 0.1845,
      "step": 688
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9946854700712184e-05,
      "loss": 0.1519,
      "step": 689
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9946176144355292e-05,
      "loss": 0.1994,
      "step": 690
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.994549329523717e-05,
      "loss": 0.1671,
      "step": 691
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9944806153652545e-05,
      "loss": 0.1722,
      "step": 692
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.994411471989798e-05,
      "loss": 0.1612,
      "step": 693
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.99434189942719e-05,
      "loss": 0.1717,
      "step": 694
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9942718977074582e-05,
      "loss": 0.1826,
      "step": 695
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.994201466860815e-05,
      "loss": 0.1793,
      "step": 696
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9941306069176583e-05,
      "loss": 0.2396,
      "step": 697
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.994059317908571e-05,
      "loss": 0.1739,
      "step": 698
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9939875998643222e-05,
      "loss": 0.2794,
      "step": 699
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.993915452815864e-05,
      "loss": 0.1925,
      "step": 700
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9938428767943362e-05,
      "loss": 0.1566,
      "step": 701
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.993769871831062e-05,
      "loss": 0.1983,
      "step": 702
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9936964379575505e-05,
      "loss": 0.1963,
      "step": 703
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9936225752054956e-05,
      "loss": 0.1874,
      "step": 704
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.993548283606776e-05,
      "loss": 0.2023,
      "step": 705
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9934735631934568e-05,
      "loss": 0.1719,
      "step": 706
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9933984139977865e-05,
      "loss": 0.1614,
      "step": 707
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9933228360521997e-05,
      "loss": 0.1824,
      "step": 708
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9932468293893157e-05,
      "loss": 0.1921,
      "step": 709
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.993170394041939e-05,
      "loss": 0.1801,
      "step": 710
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9930935300430587e-05,
      "loss": 0.1933,
      "step": 711
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9930162374258496e-05,
      "loss": 0.1963,
      "step": 712
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9929385162236715e-05,
      "loss": 0.1905,
      "step": 713
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.992860366470068e-05,
      "loss": 0.1737,
      "step": 714
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.992781788198769e-05,
      "loss": 0.1789,
      "step": 715
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9927027814436885e-05,
      "loss": 0.1711,
      "step": 716
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.992623346238926e-05,
      "loss": 0.1915,
      "step": 717
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9925434826187657e-05,
      "loss": 0.1816,
      "step": 718
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9924631906176762e-05,
      "loss": 0.1843,
      "step": 719
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.992382470270312e-05,
      "loss": 0.2008,
      "step": 720
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9923013216115118e-05,
      "loss": 0.1929,
      "step": 721
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.992219744676299e-05,
      "loss": 0.2049,
      "step": 722
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9921377394998825e-05,
      "loss": 0.171,
      "step": 723
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9920553061176557e-05,
      "loss": 0.1914,
      "step": 724
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9919724445651963e-05,
      "loss": 0.2011,
      "step": 725
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9918891548782677e-05,
      "loss": 0.1932,
      "step": 726
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9918054370928175e-05,
      "loss": 0.161,
      "step": 727
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9917212912449782e-05,
      "loss": 0.2095,
      "step": 728
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9916367173710673e-05,
      "loss": 0.1776,
      "step": 729
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9915517155075865e-05,
      "loss": 0.1738,
      "step": 730
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9914662856912228e-05,
      "loss": 0.1882,
      "step": 731
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9913804279588473e-05,
      "loss": 0.1873,
      "step": 732
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9912941423475165e-05,
      "loss": 0.195,
      "step": 733
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9912074288944708e-05,
      "loss": 0.1913,
      "step": 734
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9911202876371365e-05,
      "loss": 0.1594,
      "step": 735
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9910327186131226e-05,
      "loss": 0.2233,
      "step": 736
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9909447218602244e-05,
      "loss": 0.2041,
      "step": 737
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.990856297416421e-05,
      "loss": 0.1771,
      "step": 738
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9907674453198766e-05,
      "loss": 0.2073,
      "step": 739
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9906781656089395e-05,
      "loss": 0.1909,
      "step": 740
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9905884583221427e-05,
      "loss": 0.1999,
      "step": 741
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.990498323498204e-05,
      "loss": 0.1691,
      "step": 742
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9904077611760254e-05,
      "loss": 0.1604,
      "step": 743
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9903167713946937e-05,
      "loss": 0.1519,
      "step": 744
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9902253541934792e-05,
      "loss": 0.1669,
      "step": 745
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9901335096118386e-05,
      "loss": 0.186,
      "step": 746
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9900412376894113e-05,
      "loss": 0.1573,
      "step": 747
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.989948538466022e-05,
      "loss": 0.1602,
      "step": 748
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9898554119816795e-05,
      "loss": 0.1625,
      "step": 749
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9897618582765773e-05,
      "loss": 0.1901,
      "step": 750
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9896678773910926e-05,
      "loss": 0.2013,
      "step": 751
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.989573469365788e-05,
      "loss": 0.1755,
      "step": 752
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.98947863424141e-05,
      "loss": 0.1914,
      "step": 753
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9893833720588885e-05,
      "loss": 0.1628,
      "step": 754
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9892876828593394e-05,
      "loss": 0.1913,
      "step": 755
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9891915666840623e-05,
      "loss": 0.1713,
      "step": 756
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.98909502357454e-05,
      "loss": 0.1499,
      "step": 757
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.988998053572441e-05,
      "loss": 0.1968,
      "step": 758
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.988900656719617e-05,
      "loss": 0.1806,
      "step": 759
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9888028330581052e-05,
      "loss": 0.1528,
      "step": 760
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9887045826301254e-05,
      "loss": 0.2,
      "step": 761
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.988605905478083e-05,
      "loss": 0.1864,
      "step": 762
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9885068016445666e-05,
      "loss": 0.1579,
      "step": 763
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.988407271172349e-05,
      "loss": 0.1814,
      "step": 764
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9883073141043887e-05,
      "loss": 0.1757,
      "step": 765
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.988206930483826e-05,
      "loss": 0.2028,
      "step": 766
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.988106120353986e-05,
      "loss": 0.1748,
      "step": 767
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9880048837583797e-05,
      "loss": 0.2242,
      "step": 768
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9879032207407e-05,
      "loss": 0.1902,
      "step": 769
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9878011313448244e-05,
      "loss": 0.1822,
      "step": 770
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9876986156148143e-05,
      "loss": 0.1458,
      "step": 771
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.987595673594916e-05,
      "loss": 0.1692,
      "step": 772
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9874923053295596e-05,
      "loss": 0.168,
      "step": 773
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9873885108633577e-05,
      "loss": 0.1721,
      "step": 774
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9872842902411086e-05,
      "loss": 0.1847,
      "step": 775
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9871796435077932e-05,
      "loss": 0.1848,
      "step": 776
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9870745707085777e-05,
      "loss": 0.1691,
      "step": 777
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.986969071888811e-05,
      "loss": 0.2181,
      "step": 778
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9868631470940266e-05,
      "loss": 0.1578,
      "step": 779
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.986756796369941e-05,
      "loss": 0.1861,
      "step": 780
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9866500197624557e-05,
      "loss": 0.1471,
      "step": 781
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.986542817317655e-05,
      "loss": 0.1911,
      "step": 782
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9864351890818078e-05,
      "loss": 0.1806,
      "step": 783
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.986327135101366e-05,
      "loss": 0.1796,
      "step": 784
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9862186554229652e-05,
      "loss": 0.1671,
      "step": 785
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.986109750093426e-05,
      "loss": 0.1709,
      "step": 786
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.986000419159751e-05,
      "loss": 0.1627,
      "step": 787
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9858906626691285e-05,
      "loss": 0.1743,
      "step": 788
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9857804806689282e-05,
      "loss": 0.1811,
      "step": 789
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.985669873206705e-05,
      "loss": 0.1856,
      "step": 790
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9855588403301972e-05,
      "loss": 0.1902,
      "step": 791
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.985447382087326e-05,
      "loss": 0.1926,
      "step": 792
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.985335498526197e-05,
      "loss": 0.1642,
      "step": 793
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.985223189695099e-05,
      "loss": 0.1545,
      "step": 794
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.985110455642504e-05,
      "loss": 0.1912,
      "step": 795
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9849972964170687e-05,
      "loss": 0.1673,
      "step": 796
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.984883712067632e-05,
      "loss": 0.1541,
      "step": 797
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.984769702643217e-05,
      "loss": 0.2043,
      "step": 798
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.98465526819303e-05,
      "loss": 0.1632,
      "step": 799
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.984540408766461e-05,
      "loss": 0.1843,
      "step": 800
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9844251244130825e-05,
      "loss": 0.1779,
      "step": 801
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9843094151826522e-05,
      "loss": 0.1883,
      "step": 802
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9841932811251096e-05,
      "loss": 0.16,
      "step": 803
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9840767222905778e-05,
      "loss": 0.1813,
      "step": 804
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9839597387293636e-05,
      "loss": 0.1634,
      "step": 805
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9838423304919575e-05,
      "loss": 0.1725,
      "step": 806
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.983724497629032e-05,
      "loss": 0.1591,
      "step": 807
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9836062401914447e-05,
      "loss": 0.1676,
      "step": 808
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9834875582302344e-05,
      "loss": 0.2117,
      "step": 809
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9833684517966247e-05,
      "loss": 0.1925,
      "step": 810
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9832489209420216e-05,
      "loss": 0.1761,
      "step": 811
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.983128965718015e-05,
      "loss": 0.1709,
      "step": 812
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9830085861763762e-05,
      "loss": 0.1899,
      "step": 813
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9828877823690623e-05,
      "loss": 0.1667,
      "step": 814
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.982766554348212e-05,
      "loss": 0.1642,
      "step": 815
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9826449021661466e-05,
      "loss": 0.1524,
      "step": 816
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.982522825875371e-05,
      "loss": 0.1812,
      "step": 817
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9824003255285735e-05,
      "loss": 0.2015,
      "step": 818
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9822774011786255e-05,
      "loss": 0.1603,
      "step": 819
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9821540528785802e-05,
      "loss": 0.1718,
      "step": 820
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9820302806816757e-05,
      "loss": 0.1672,
      "step": 821
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9819060846413314e-05,
      "loss": 0.1712,
      "step": 822
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.98178146481115e-05,
      "loss": 0.1584,
      "step": 823
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9816564212449173e-05,
      "loss": 0.1861,
      "step": 824
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9815309539966026e-05,
      "loss": 0.1537,
      "step": 825
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.981405063120357e-05,
      "loss": 0.1544,
      "step": 826
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9812787486705147e-05,
      "loss": 0.182,
      "step": 827
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9811520107015935e-05,
      "loss": 0.1725,
      "step": 828
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.981024849268293e-05,
      "loss": 0.1651,
      "step": 829
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.980897264425496e-05,
      "loss": 0.1622,
      "step": 830
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9807692562282685e-05,
      "loss": 0.1993,
      "step": 831
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9806408247318576e-05,
      "loss": 0.1462,
      "step": 832
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9805119699916955e-05,
      "loss": 0.174,
      "step": 833
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9803826920633947e-05,
      "loss": 0.1735,
      "step": 834
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9802529910027524e-05,
      "loss": 0.1786,
      "step": 835
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.980122866865747e-05,
      "loss": 0.1622,
      "step": 836
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.97999231970854e-05,
      "loss": 0.1801,
      "step": 837
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9798613495874754e-05,
      "loss": 0.1729,
      "step": 838
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9797299565590795e-05,
      "loss": 0.1696,
      "step": 839
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.979598140680062e-05,
      "loss": 0.1817,
      "step": 840
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9794659020073146e-05,
      "loss": 0.1833,
      "step": 841
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9793332405979107e-05,
      "loss": 0.1823,
      "step": 842
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9792001565091074e-05,
      "loss": 0.1522,
      "step": 843
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9790666497983433e-05,
      "loss": 0.1791,
      "step": 844
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.97893272052324e-05,
      "loss": 0.1913,
      "step": 845
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9787983687416017e-05,
      "loss": 0.1782,
      "step": 846
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9786635945114135e-05,
      "loss": 0.1973,
      "step": 847
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.978528397890845e-05,
      "loss": 0.1819,
      "step": 848
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.978392778938246e-05,
      "loss": 0.1727,
      "step": 849
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.97825673771215e-05,
      "loss": 0.1677,
      "step": 850
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.978120274271272e-05,
      "loss": 0.1742,
      "step": 851
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9779833886745098e-05,
      "loss": 0.1541,
      "step": 852
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.977846080980943e-05,
      "loss": 0.1625,
      "step": 853
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9777083512498336e-05,
      "loss": 0.2079,
      "step": 854
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9775701995406252e-05,
      "loss": 0.1617,
      "step": 855
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9774316259129444e-05,
      "loss": 0.1796,
      "step": 856
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9772926304265994e-05,
      "loss": 0.1668,
      "step": 857
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9771532131415806e-05,
      "loss": 0.1522,
      "step": 858
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.97701337411806e-05,
      "loss": 0.1614,
      "step": 859
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9768731134163924e-05,
      "loss": 0.1735,
      "step": 860
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.976732431097114e-05,
      "loss": 0.1876,
      "step": 861
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.976591327220943e-05,
      "loss": 0.1525,
      "step": 862
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.97644980184878e-05,
      "loss": 0.1975,
      "step": 863
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9763078550417073e-05,
      "loss": 0.1981,
      "step": 864
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.976165486860989e-05,
      "loss": 0.1722,
      "step": 865
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9760226973680706e-05,
      "loss": 0.1797,
      "step": 866
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9758794866245808e-05,
      "loss": 0.1552,
      "step": 867
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9757358546923286e-05,
      "loss": 0.1691,
      "step": 868
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9755918016333053e-05,
      "loss": 0.1642,
      "step": 869
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9754473275096845e-05,
      "loss": 0.1185,
      "step": 870
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9753024323838207e-05,
      "loss": 0.1776,
      "step": 871
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9751571163182506e-05,
      "loss": 0.1722,
      "step": 872
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.975011379375693e-05,
      "loss": 0.178,
      "step": 873
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9748652216190477e-05,
      "loss": 0.1566,
      "step": 874
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9747186431113957e-05,
      "loss": 0.1922,
      "step": 875
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9745716439160007e-05,
      "loss": 0.1754,
      "step": 876
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.974424224096307e-05,
      "loss": 0.1837,
      "step": 877
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9742763837159412e-05,
      "loss": 0.1874,
      "step": 878
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.974128122838711e-05,
      "loss": 0.1696,
      "step": 879
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.973979441528606e-05,
      "loss": 0.1922,
      "step": 880
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.973830339849796e-05,
      "loss": 0.1544,
      "step": 881
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9736808178666342e-05,
      "loss": 0.1744,
      "step": 882
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9735308756436535e-05,
      "loss": 0.1761,
      "step": 883
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9733805132455693e-05,
      "loss": 0.1712,
      "step": 884
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9732297307372775e-05,
      "loss": 0.1401,
      "step": 885
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9730785281838554e-05,
      "loss": 0.1583,
      "step": 886
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.972926905650563e-05,
      "loss": 0.1781,
      "step": 887
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9727748632028394e-05,
      "loss": 0.1891,
      "step": 888
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9726224009063063e-05,
      "loss": 0.1849,
      "step": 889
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.972469518826767e-05,
      "loss": 0.1971,
      "step": 890
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.972316217030204e-05,
      "loss": 0.1805,
      "step": 891
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.972162495582783e-05,
      "loss": 0.1528,
      "step": 892
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.97200835455085e-05,
      "loss": 0.1899,
      "step": 893
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.971853794000932e-05,
      "loss": 0.1754,
      "step": 894
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9716988139997373e-05,
      "loss": 0.2447,
      "step": 895
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9715434146141553e-05,
      "loss": 0.1835,
      "step": 896
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.971387595911256e-05,
      "loss": 0.156,
      "step": 897
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9712313579582907e-05,
      "loss": 0.1878,
      "step": 898
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9710747008226915e-05,
      "loss": 0.1673,
      "step": 899
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9709176245720713e-05,
      "loss": 0.1674,
      "step": 900
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.970760129274224e-05,
      "loss": 0.1825,
      "step": 901
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9706022149971257e-05,
      "loss": 0.194,
      "step": 902
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9704438818089304e-05,
      "loss": 0.1681,
      "step": 903
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9702851297779755e-05,
      "loss": 0.1531,
      "step": 904
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.970125958972778e-05,
      "loss": 0.1701,
      "step": 905
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9699663694620355e-05,
      "loss": 0.1797,
      "step": 906
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.969806361314628e-05,
      "loss": 0.1775,
      "step": 907
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.969645934599613e-05,
      "loss": 0.1714,
      "step": 908
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9694850893862322e-05,
      "loss": 0.1494,
      "step": 909
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9693238257439056e-05,
      "loss": 0.1917,
      "step": 910
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9691621437422343e-05,
      "loss": 0.1697,
      "step": 911
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.969000043451e-05,
      "loss": 0.1632,
      "step": 912
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9688375249401656e-05,
      "loss": 0.1836,
      "step": 913
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9686745882798733e-05,
      "loss": 0.1565,
      "step": 914
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.968511233540447e-05,
      "loss": 0.1743,
      "step": 915
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9683474607923902e-05,
      "loss": 0.1531,
      "step": 916
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.968183270106387e-05,
      "loss": 0.162,
      "step": 917
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9680186615533022e-05,
      "loss": 0.1511,
      "step": 918
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9678536352041807e-05,
      "loss": 0.1633,
      "step": 919
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9676881911302476e-05,
      "loss": 0.1791,
      "step": 920
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9675223294029084e-05,
      "loss": 0.1897,
      "step": 921
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.967356050093749e-05,
      "loss": 0.1783,
      "step": 922
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.967189353274535e-05,
      "loss": 0.1685,
      "step": 923
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9670222390172134e-05,
      "loss": 0.1699,
      "step": 924
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.96685470739391e-05,
      "loss": 0.2324,
      "step": 925
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.966686758476931e-05,
      "loss": 0.1619,
      "step": 926
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.966518392338764e-05,
      "loss": 0.1544,
      "step": 927
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.966349609052075e-05,
      "loss": 0.1543,
      "step": 928
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9661804086897106e-05,
      "loss": 0.2165,
      "step": 929
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9660107913246977e-05,
      "loss": 0.1801,
      "step": 930
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9658407570302435e-05,
      "loss": 0.1556,
      "step": 931
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.965670305879734e-05,
      "loss": 0.1962,
      "step": 932
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9654994379467363e-05,
      "loss": 0.1884,
      "step": 933
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.965328153304996e-05,
      "loss": 0.1561,
      "step": 934
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9651564520284406e-05,
      "loss": 0.2165,
      "step": 935
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.964984334191175e-05,
      "loss": 0.1566,
      "step": 936
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9648117998674863e-05,
      "loss": 0.1474,
      "step": 937
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9646388491318394e-05,
      "loss": 0.143,
      "step": 938
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9644654820588798e-05,
      "loss": 0.1696,
      "step": 939
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9642916987234327e-05,
      "loss": 0.1766,
      "step": 940
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9641174992005028e-05,
      "loss": 0.1681,
      "step": 941
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.963942883565274e-05,
      "loss": 0.1783,
      "step": 942
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9637678518931116e-05,
      "loss": 0.1599,
      "step": 943
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9635924042595578e-05,
      "loss": 0.1772,
      "step": 944
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9634165407403363e-05,
      "loss": 0.1939,
      "step": 945
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.963240261411349e-05,
      "loss": 0.191,
      "step": 946
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9630635663486786e-05,
      "loss": 0.1969,
      "step": 947
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.962886455628587e-05,
      "loss": 0.1878,
      "step": 948
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9627089293275133e-05,
      "loss": 0.2037,
      "step": 949
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.962530987522079e-05,
      "loss": 0.1433,
      "step": 950
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9623526302890834e-05,
      "loss": 0.1447,
      "step": 951
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9621738577055056e-05,
      "loss": 0.1898,
      "step": 952
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9619946698485033e-05,
      "loss": 0.1613,
      "step": 953
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9618150667954137e-05,
      "loss": 0.1519,
      "step": 954
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9616350486237536e-05,
      "loss": 0.1652,
      "step": 955
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.961454615411219e-05,
      "loss": 0.1797,
      "step": 956
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9612737672356838e-05,
      "loss": 0.1638,
      "step": 957
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.961092504175203e-05,
      "loss": 0.166,
      "step": 958
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9609108263080085e-05,
      "loss": 0.1789,
      "step": 959
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9607287337125133e-05,
      "loss": 0.1597,
      "step": 960
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.960546226467308e-05,
      "loss": 0.1686,
      "step": 961
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9603633046511623e-05,
      "loss": 0.2049,
      "step": 962
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9601799683430255e-05,
      "loss": 0.2059,
      "step": 963
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9599962176220256e-05,
      "loss": 0.1809,
      "step": 964
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9598120525674687e-05,
      "loss": 0.1626,
      "step": 965
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9596274732588404e-05,
      "loss": 0.1829,
      "step": 966
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.959442479775805e-05,
      "loss": 0.1496,
      "step": 967
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.959257072198206e-05,
      "loss": 0.1984,
      "step": 968
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9590712506060642e-05,
      "loss": 0.1669,
      "step": 969
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.958885015079581e-05,
      "loss": 0.1674,
      "step": 970
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.958698365699135e-05,
      "loss": 0.1942,
      "step": 971
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9585113025452837e-05,
      "loss": 0.1912,
      "step": 972
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.958323825698764e-05,
      "loss": 0.1379,
      "step": 973
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9581359352404907e-05,
      "loss": 0.1744,
      "step": 974
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9579476312515563e-05,
      "loss": 0.1614,
      "step": 975
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9577589138132338e-05,
      "loss": 0.1901,
      "step": 976
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9575697830069728e-05,
      "loss": 0.1864,
      "step": 977
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9573802389144023e-05,
      "loss": 0.1829,
      "step": 978
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.957190281617329e-05,
      "loss": 0.1732,
      "step": 979
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9569999111977384e-05,
      "loss": 0.1809,
      "step": 980
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.956809127737795e-05,
      "loss": 0.1814,
      "step": 981
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9566179313198393e-05,
      "loss": 0.1712,
      "step": 982
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.956426322026393e-05,
      "loss": 0.1911,
      "step": 983
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9562342999401532e-05,
      "loss": 0.1954,
      "step": 984
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9560418651439974e-05,
      "loss": 0.131,
      "step": 985
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.95584901772098e-05,
      "loss": 0.1807,
      "step": 986
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.955655757754334e-05,
      "loss": 0.1754,
      "step": 987
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9554620853274698e-05,
      "loss": 0.1731,
      "step": 988
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9552680005239768e-05,
      "loss": 0.1991,
      "step": 989
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9550735034276213e-05,
      "loss": 0.1665,
      "step": 990
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9548785941223483e-05,
      "loss": 0.1697,
      "step": 991
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9546832726922803e-05,
      "loss": 0.1786,
      "step": 992
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9544875392217185e-05,
      "loss": 0.1773,
      "step": 993
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9542913937951405e-05,
      "loss": 0.1592,
      "step": 994
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.954094836497203e-05,
      "loss": 0.1765,
      "step": 995
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9538978674127397e-05,
      "loss": 0.195,
      "step": 996
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9537004866267626e-05,
      "loss": 0.1642,
      "step": 997
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.953502694224461e-05,
      "loss": 0.1679,
      "step": 998
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9533044902912017e-05,
      "loss": 0.1655,
      "step": 999
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9531058749125295e-05,
      "loss": 0.1839,
      "step": 1000
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9529068481741666e-05,
      "loss": 0.1519,
      "step": 1001
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.952707410162013e-05,
      "loss": 0.1512,
      "step": 1002
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9525075609621452e-05,
      "loss": 0.1896,
      "step": 1003
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9523073006608187e-05,
      "loss": 0.1954,
      "step": 1004
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9521066293444656e-05,
      "loss": 0.1518,
      "step": 1005
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9519055470996954e-05,
      "loss": 0.179,
      "step": 1006
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.951704054013295e-05,
      "loss": 0.1953,
      "step": 1007
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9515021501722286e-05,
      "loss": 0.1771,
      "step": 1008
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9512998356636377e-05,
      "loss": 0.1544,
      "step": 1009
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9510971105748412e-05,
      "loss": 0.1878,
      "step": 1010
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9508939749933348e-05,
      "loss": 0.1555,
      "step": 1011
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9506904290067922e-05,
      "loss": 0.1763,
      "step": 1012
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9504864727030634e-05,
      "loss": 0.1634,
      "step": 1013
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9502821061701752e-05,
      "loss": 0.1727,
      "step": 1014
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9500773294963328e-05,
      "loss": 0.1629,
      "step": 1015
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9498721427699174e-05,
      "loss": 0.1631,
      "step": 1016
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.949666546079487e-05,
      "loss": 0.1835,
      "step": 1017
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9494605395137778e-05,
      "loss": 0.2096,
      "step": 1018
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9492541231617015e-05,
      "loss": 0.1956,
      "step": 1019
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.949047297112347e-05,
      "loss": 0.1705,
      "step": 1020
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.948840061454981e-05,
      "loss": 0.1704,
      "step": 1021
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9486324162790455e-05,
      "loss": 0.156,
      "step": 1022
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9484243616741606e-05,
      "loss": 0.1764,
      "step": 1023
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9482158977301217e-05,
      "loss": 0.1517,
      "step": 1024
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9480070245369024e-05,
      "loss": 0.1802,
      "step": 1025
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.947797742184652e-05,
      "loss": 0.1712,
      "step": 1026
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9475880507636965e-05,
      "loss": 0.1813,
      "step": 1027
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9473779503645386e-05,
      "loss": 0.166,
      "step": 1028
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9471674410778573e-05,
      "loss": 0.1425,
      "step": 1029
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9469565229945083e-05,
      "loss": 0.1817,
      "step": 1030
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9467451962055236e-05,
      "loss": 0.1547,
      "step": 1031
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.946533460802112e-05,
      "loss": 0.1911,
      "step": 1032
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9463213168756582e-05,
      "loss": 0.1853,
      "step": 1033
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9461087645177227e-05,
      "loss": 0.1863,
      "step": 1034
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.945895803820044e-05,
      "loss": 0.1742,
      "step": 1035
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.945682434874535e-05,
      "loss": 0.1354,
      "step": 1036
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.945468657773286e-05,
      "loss": 0.1577,
      "step": 1037
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9452544726085625e-05,
      "loss": 0.1485,
      "step": 1038
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9450398794728067e-05,
      "loss": 0.1539,
      "step": 1039
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9448248784586374e-05,
      "loss": 0.1722,
      "step": 1040
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.944609469658848e-05,
      "loss": 0.1883,
      "step": 1041
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.94439365316641e-05,
      "loss": 0.2049,
      "step": 1042
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9441774290744682e-05,
      "loss": 0.2012,
      "step": 1043
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.943960797476346e-05,
      "loss": 0.194,
      "step": 1044
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9437437584655404e-05,
      "loss": 0.1695,
      "step": 1045
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.943526312135726e-05,
      "loss": 0.1758,
      "step": 1046
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9433084585807517e-05,
      "loss": 0.1956,
      "step": 1047
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9430901978946436e-05,
      "loss": 0.1697,
      "step": 1048
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9428715301716026e-05,
      "loss": 0.2028,
      "step": 1049
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9426524555060053e-05,
      "loss": 0.163,
      "step": 1050
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9424329739924044e-05,
      "loss": 0.1491,
      "step": 1051
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9422130857255277e-05,
      "loss": 0.187,
      "step": 1052
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.941992790800279e-05,
      "loss": 0.1702,
      "step": 1053
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.941772089311737e-05,
      "loss": 0.2022,
      "step": 1054
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9415509813551573e-05,
      "loss": 0.1532,
      "step": 1055
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9413294670259688e-05,
      "loss": 0.1372,
      "step": 1056
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.941107546419777e-05,
      "loss": 0.2006,
      "step": 1057
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9408852196323638e-05,
      "loss": 0.1851,
      "step": 1058
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.940662486759684e-05,
      "loss": 0.1628,
      "step": 1059
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9404393478978688e-05,
      "loss": 0.1706,
      "step": 1060
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.940215803143226e-05,
      "loss": 0.1647,
      "step": 1061
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9399918525922364e-05,
      "loss": 0.1777,
      "step": 1062
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.939767496341557e-05,
      "loss": 0.1769,
      "step": 1063
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.93954273448802e-05,
      "loss": 0.1496,
      "step": 1064
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.939317567128632e-05,
      "loss": 0.1688,
      "step": 1065
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.939091994360575e-05,
      "loss": 0.1785,
      "step": 1066
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9388660162812072e-05,
      "loss": 0.1819,
      "step": 1067
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9386396329880593e-05,
      "loss": 0.1566,
      "step": 1068
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.938412844578838e-05,
      "loss": 0.149,
      "step": 1069
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.938185651151426e-05,
      "loss": 0.1625,
      "step": 1070
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9379580528038793e-05,
      "loss": 0.1955,
      "step": 1071
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.937730049634429e-05,
      "loss": 0.1578,
      "step": 1072
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9375016417414815e-05,
      "loss": 0.1537,
      "step": 1073
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.937272829223617e-05,
      "loss": 0.1526,
      "step": 1074
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.937043612179591e-05,
      "loss": 0.1891,
      "step": 1075
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9368139907083333e-05,
      "loss": 0.1776,
      "step": 1076
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9365839649089483e-05,
      "loss": 0.1501,
      "step": 1077
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.936353534880715e-05,
      "loss": 0.1967,
      "step": 1078
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9361227007230867e-05,
      "loss": 0.1952,
      "step": 1079
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9358914625356914e-05,
      "loss": 0.1853,
      "step": 1080
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9356598204183314e-05,
      "loss": 0.2016,
      "step": 1081
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.935427774470983e-05,
      "loss": 0.201,
      "step": 1082
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9351953247937977e-05,
      "loss": 0.1626,
      "step": 1083
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9349624714870994e-05,
      "loss": 0.1573,
      "step": 1084
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.934729214651388e-05,
      "loss": 0.1963,
      "step": 1085
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9344955543873373e-05,
      "loss": 0.2013,
      "step": 1086
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9342614907957942e-05,
      "loss": 0.176,
      "step": 1087
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.934027023977781e-05,
      "loss": 0.1489,
      "step": 1088
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9337921540344927e-05,
      "loss": 0.1518,
      "step": 1089
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9335568810672995e-05,
      "loss": 0.156,
      "step": 1090
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.933321205177745e-05,
      "loss": 0.1637,
      "step": 1091
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9330851264675462e-05,
      "loss": 0.1427,
      "step": 1092
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.932848645038595e-05,
      "loss": 0.1517,
      "step": 1093
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9326117609929568e-05,
      "loss": 0.1698,
      "step": 1094
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9323744744328698e-05,
      "loss": 0.1701,
      "step": 1095
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9321367854607472e-05,
      "loss": 0.156,
      "step": 1096
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.931898694179175e-05,
      "loss": 0.1541,
      "step": 1097
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9316602006909132e-05,
      "loss": 0.1913,
      "step": 1098
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9314213050988962e-05,
      "loss": 0.1731,
      "step": 1099
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9311820075062303e-05,
      "loss": 0.2109,
      "step": 1100
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9309423080161963e-05,
      "loss": 0.2151,
      "step": 1101
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9307022067322484e-05,
      "loss": 0.1785,
      "step": 1102
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.930461703758014e-05,
      "loss": 0.1679,
      "step": 1103
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.930220799197294e-05,
      "loss": 0.1717,
      "step": 1104
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.929979493154063e-05,
      "loss": 0.176,
      "step": 1105
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.929737785732468e-05,
      "loss": 0.1697,
      "step": 1106
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9294956770368297e-05,
      "loss": 0.1684,
      "step": 1107
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9292531671716423e-05,
      "loss": 0.1753,
      "step": 1108
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.929010256241573e-05,
      "loss": 0.1638,
      "step": 1109
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.928766944351461e-05,
      "loss": 0.1345,
      "step": 1110
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9285232316063206e-05,
      "loss": 0.1533,
      "step": 1111
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9282791181113377e-05,
      "loss": 0.1693,
      "step": 1112
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9280346039718715e-05,
      "loss": 0.1661,
      "step": 1113
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9277896892934536e-05,
      "loss": 0.1744,
      "step": 1114
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9275443741817895e-05,
      "loss": 0.1732,
      "step": 1115
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.927298658742757e-05,
      "loss": 0.1581,
      "step": 1116
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.927052543082406e-05,
      "loss": 0.1878,
      "step": 1117
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.926806027306961e-05,
      "loss": 0.1463,
      "step": 1118
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.926559111522817e-05,
      "loss": 0.1913,
      "step": 1119
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9263117958365428e-05,
      "loss": 0.1657,
      "step": 1120
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.92606408035488e-05,
      "loss": 0.1602,
      "step": 1121
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9258159651847424e-05,
      "loss": 0.1407,
      "step": 1122
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9255674504332156e-05,
      "loss": 0.1551,
      "step": 1123
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9253185362075593e-05,
      "loss": 0.1743,
      "step": 1124
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9250692226152044e-05,
      "loss": 0.1686,
      "step": 1125
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9248195097637538e-05,
      "loss": 0.142,
      "step": 1126
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9245693977609837e-05,
      "loss": 0.1741,
      "step": 1127
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.924318886714843e-05,
      "loss": 0.1549,
      "step": 1128
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9240679767334512e-05,
      "loss": 0.21,
      "step": 1129
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9238166679251006e-05,
      "loss": 0.1488,
      "step": 1130
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9235649603982566e-05,
      "loss": 0.1734,
      "step": 1131
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9233128542615558e-05,
      "loss": 0.1769,
      "step": 1132
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9230603496238067e-05,
      "loss": 0.1461,
      "step": 1133
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9228074465939904e-05,
      "loss": 0.1597,
      "step": 1134
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9225541452812595e-05,
      "loss": 0.1517,
      "step": 1135
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9223004457949382e-05,
      "loss": 0.1661,
      "step": 1136
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.922046348244524e-05,
      "loss": 0.1535,
      "step": 1137
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9217918527396842e-05,
      "loss": 0.1846,
      "step": 1138
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9215369593902593e-05,
      "loss": 0.1506,
      "step": 1139
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9212816683062612e-05,
      "loss": 0.1632,
      "step": 1140
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.921025979597873e-05,
      "loss": 0.1393,
      "step": 1141
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9207698933754498e-05,
      "loss": 0.1559,
      "step": 1142
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9205134097495183e-05,
      "loss": 0.1695,
      "step": 1143
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9202565288307764e-05,
      "loss": 0.1831,
      "step": 1144
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9199992507300938e-05,
      "loss": 0.1673,
      "step": 1145
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9197415755585117e-05,
      "loss": 0.1559,
      "step": 1146
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9194835034272424e-05,
      "loss": 0.157,
      "step": 1147
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9192250344476694e-05,
      "loss": 0.1874,
      "step": 1148
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9189661687313477e-05,
      "loss": 0.1785,
      "step": 1149
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.918706906390004e-05,
      "loss": 0.1485,
      "step": 1150
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9184472475355353e-05,
      "loss": 0.202,
      "step": 1151
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.91818719228001e-05,
      "loss": 0.1882,
      "step": 1152
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9179267407356682e-05,
      "loss": 0.1796,
      "step": 1153
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.91766589301492e-05,
      "loss": 0.1526,
      "step": 1154
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9174046492303476e-05,
      "loss": 0.1532,
      "step": 1155
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9171430094947036e-05,
      "loss": 0.1781,
      "step": 1156
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9168809739209112e-05,
      "loss": 0.1467,
      "step": 1157
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9166185426220645e-05,
      "loss": 0.1568,
      "step": 1158
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9163557157114288e-05,
      "loss": 0.1833,
      "step": 1159
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9160924933024403e-05,
      "loss": 0.1524,
      "step": 1160
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9158288755087056e-05,
      "loss": 0.1838,
      "step": 1161
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9155648624440012e-05,
      "loss": 0.1779,
      "step": 1162
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9153004542222756e-05,
      "loss": 0.1666,
      "step": 1163
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.915035650957647e-05,
      "loss": 0.2293,
      "step": 1164
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9147704527644044e-05,
      "loss": 0.1655,
      "step": 1165
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9145048597570067e-05,
      "loss": 0.15,
      "step": 1166
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.914238872050084e-05,
      "loss": 0.1489,
      "step": 1167
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.913972489758436e-05,
      "loss": 0.1702,
      "step": 1168
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9137057129970334e-05,
      "loss": 0.147,
      "step": 1169
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.913438541881017e-05,
      "loss": 0.1861,
      "step": 1170
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.913170976525697e-05,
      "loss": 0.1804,
      "step": 1171
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9129030170465545e-05,
      "loss": 0.1665,
      "step": 1172
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.912634663559241e-05,
      "loss": 0.1677,
      "step": 1173
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.912365916179578e-05,
      "loss": 0.1871,
      "step": 1174
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.912096775023556e-05,
      "loss": 0.1715,
      "step": 1175
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9118272402073353e-05,
      "loss": 0.1853,
      "step": 1176
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9115573118472486e-05,
      "loss": 0.1648,
      "step": 1177
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9112869900597955e-05,
      "loss": 0.165,
      "step": 1178
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9110162749616476e-05,
      "loss": 0.1678,
      "step": 1179
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9107451666696443e-05,
      "loss": 0.1723,
      "step": 1180
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9104736653007964e-05,
      "loss": 0.1852,
      "step": 1181
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.910201770972284e-05,
      "loss": 0.1592,
      "step": 1182
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.909929483801455e-05,
      "loss": 0.2035,
      "step": 1183
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9096568039058297e-05,
      "loss": 0.1595,
      "step": 1184
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.909383731403096e-05,
      "loss": 0.1448,
      "step": 1185
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.909110266411112e-05,
      "loss": 0.1201,
      "step": 1186
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9088364090479046e-05,
      "loss": 0.1737,
      "step": 1187
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.908562159431671e-05,
      "loss": 0.165,
      "step": 1188
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.908287517680776e-05,
      "loss": 0.1715,
      "step": 1189
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9080124839137563e-05,
      "loss": 0.1638,
      "step": 1190
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9077370582493146e-05,
      "loss": 0.1839,
      "step": 1191
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9074612408063257e-05,
      "loss": 0.1563,
      "step": 1192
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9071850317038313e-05,
      "loss": 0.1444,
      "step": 1193
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9069084310610434e-05,
      "loss": 0.1392,
      "step": 1194
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9066314389973426e-05,
      "loss": 0.1809,
      "step": 1195
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9063540556322787e-05,
      "loss": 0.1791,
      "step": 1196
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9060762810855694e-05,
      "loss": 0.1505,
      "step": 1197
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9057981154771026e-05,
      "loss": 0.1628,
      "step": 1198
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9055195589269342e-05,
      "loss": 0.1607,
      "step": 1199
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.905240611555289e-05,
      "loss": 0.1903,
      "step": 1200
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.90496127348256e-05,
      "loss": 0.1293,
      "step": 1201
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9046815448293095e-05,
      "loss": 0.1494,
      "step": 1202
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.904401425716269e-05,
      "loss": 0.1691,
      "step": 1203
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9041209162643365e-05,
      "loss": 0.147,
      "step": 1204
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.90384001659458e-05,
      "loss": 0.1876,
      "step": 1205
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9035587268282357e-05,
      "loss": 0.1709,
      "step": 1206
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9032770470867082e-05,
      "loss": 0.1571,
      "step": 1207
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9029949774915698e-05,
      "loss": 0.1494,
      "step": 1208
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9027125181645617e-05,
      "loss": 0.1661,
      "step": 1209
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.902429669227593e-05,
      "loss": 0.1676,
      "step": 1210
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9021464308027416e-05,
      "loss": 0.1751,
      "step": 1211
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9018628030122526e-05,
      "loss": 0.1384,
      "step": 1212
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.901578785978539e-05,
      "loss": 0.1619,
      "step": 1213
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9012943798241833e-05,
      "loss": 0.1916,
      "step": 1214
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.901009584671934e-05,
      "loss": 0.1547,
      "step": 1215
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9007244006447093e-05,
      "loss": 0.1411,
      "step": 1216
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9004388278655938e-05,
      "loss": 0.1829,
      "step": 1217
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9001528664578402e-05,
      "loss": 0.201,
      "step": 1218
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8998665165448702e-05,
      "loss": 0.1812,
      "step": 1219
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8995797782502715e-05,
      "loss": 0.1462,
      "step": 1220
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8992926516978e-05,
      "loss": 0.212,
      "step": 1221
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8990051370113796e-05,
      "loss": 0.1828,
      "step": 1222
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8987172343151014e-05,
      "loss": 0.136,
      "step": 1223
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.898428943733224e-05,
      "loss": 0.152,
      "step": 1224
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8981402653901727e-05,
      "loss": 0.164,
      "step": 1225
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.897851199410542e-05,
      "loss": 0.1671,
      "step": 1226
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8975617459190914e-05,
      "loss": 0.1806,
      "step": 1227
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8972719050407494e-05,
      "loss": 0.1585,
      "step": 1228
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8969816769006115e-05,
      "loss": 0.2019,
      "step": 1229
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8966910616239387e-05,
      "loss": 0.1706,
      "step": 1230
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8964000593361613e-05,
      "loss": 0.1737,
      "step": 1231
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8961086701628753e-05,
      "loss": 0.1558,
      "step": 1232
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.895816894229844e-05,
      "loss": 0.1512,
      "step": 1233
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8955247316629975e-05,
      "loss": 0.1607,
      "step": 1234
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8952321825884336e-05,
      "loss": 0.1454,
      "step": 1235
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.894939247132415e-05,
      "loss": 0.1557,
      "step": 1236
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8946459254213737e-05,
      "loss": 0.1634,
      "step": 1237
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8943522175819065e-05,
      "loss": 0.1906,
      "step": 1238
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8940581237407774e-05,
      "loss": 0.1531,
      "step": 1239
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.893763644024917e-05,
      "loss": 0.1718,
      "step": 1240
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.893468778561423e-05,
      "loss": 0.1704,
      "step": 1241
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8931735274775587e-05,
      "loss": 0.166,
      "step": 1242
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.892877890900754e-05,
      "loss": 0.1785,
      "step": 1243
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8925818689586057e-05,
      "loss": 0.1662,
      "step": 1244
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8922854617788768e-05,
      "loss": 0.1762,
      "step": 1245
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.891988669489496e-05,
      "loss": 0.1615,
      "step": 1246
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8916914922185586e-05,
      "loss": 0.1555,
      "step": 1247
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8913939300943266e-05,
      "loss": 0.156,
      "step": 1248
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8910959832452268e-05,
      "loss": 0.1581,
      "step": 1249
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8907976517998534e-05,
      "loss": 0.1571,
      "step": 1250
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8904989358869655e-05,
      "loss": 0.1293,
      "step": 1251
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.890199835635489e-05,
      "loss": 0.1415,
      "step": 1252
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8899003511745155e-05,
      "loss": 0.2222,
      "step": 1253
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8896004826333018e-05,
      "loss": 0.1906,
      "step": 1254
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.889300230141271e-05,
      "loss": 0.1522,
      "step": 1255
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8889995938280124e-05,
      "loss": 0.1874,
      "step": 1256
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8886985738232795e-05,
      "loss": 0.1738,
      "step": 1257
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8883971702569924e-05,
      "loss": 0.1727,
      "step": 1258
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8880953832592373e-05,
      "loss": 0.1475,
      "step": 1259
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8877932129602646e-05,
      "loss": 0.1584,
      "step": 1260
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8874906594904914e-05,
      "loss": 0.1838,
      "step": 1261
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.8871877229804987e-05,
      "loss": 0.1787,
      "step": 1262
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8868844035610337e-05,
      "loss": 0.1546,
      "step": 1263
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.88658070136301e-05,
      "loss": 0.1719,
      "step": 1264
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8862766165175035e-05,
      "loss": 0.1398,
      "step": 1265
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8859721491557582e-05,
      "loss": 0.146,
      "step": 1266
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8856672994091814e-05,
      "loss": 0.1542,
      "step": 1267
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.885362067409346e-05,
      "loss": 0.1608,
      "step": 1268
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8850564532879906e-05,
      "loss": 0.1568,
      "step": 1269
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8847504571770167e-05,
      "loss": 0.1587,
      "step": 1270
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8844440792084932e-05,
      "loss": 0.1331,
      "step": 1271
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.884137319514652e-05,
      "loss": 0.1811,
      "step": 1272
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.88383017822789e-05,
      "loss": 0.1405,
      "step": 1273
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.88352265548077e-05,
      "loss": 0.1351,
      "step": 1274
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8832147514060176e-05,
      "loss": 0.1885,
      "step": 1275
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8829064661365247e-05,
      "loss": 0.1558,
      "step": 1276
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8825977998053466e-05,
      "loss": 0.1929,
      "step": 1277
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8822887525457038e-05,
      "loss": 0.1721,
      "step": 1278
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.88197932449098e-05,
      "loss": 0.1417,
      "step": 1279
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8816695157747244e-05,
      "loss": 0.1616,
      "step": 1280
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8813593265306508e-05,
      "loss": 0.1611,
      "step": 1281
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8810487568926362e-05,
      "loss": 0.1661,
      "step": 1282
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.880737806994722e-05,
      "loss": 0.1611,
      "step": 1283
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.8804264769711145e-05,
      "loss": 0.1657,
      "step": 1284
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.880114766956183e-05,
      "loss": 0.1837,
      "step": 1285
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.879802677084461e-05,
      "loss": 0.1534,
      "step": 1286
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.879490207490647e-05,
      "loss": 0.209,
      "step": 1287
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.879177358309602e-05,
      "loss": 0.149,
      "step": 1288
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8788641296763516e-05,
      "loss": 0.1442,
      "step": 1289
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.878550521726085e-05,
      "loss": 0.1398,
      "step": 1290
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8782365345941553e-05,
      "loss": 0.1744,
      "step": 1291
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.877922168416079e-05,
      "loss": 0.1411,
      "step": 1292
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.877607423327536e-05,
      "loss": 0.1633,
      "step": 1293
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8772922994643707e-05,
      "loss": 0.1606,
      "step": 1294
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8769767969625897e-05,
      "loss": 0.153,
      "step": 1295
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8766609159583635e-05,
      "loss": 0.1662,
      "step": 1296
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8763446565880265e-05,
      "loss": 0.1501,
      "step": 1297
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.876028018988076e-05,
      "loss": 0.1743,
      "step": 1298
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.875711003295172e-05,
      "loss": 0.1707,
      "step": 1299
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8753936096461388e-05,
      "loss": 0.1923,
      "step": 1300
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.875075838177963e-05,
      "loss": 0.1614,
      "step": 1301
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8747576890277944e-05,
      "loss": 0.1965,
      "step": 1302
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8744391623329457e-05,
      "loss": 0.17,
      "step": 1303
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8741202582308932e-05,
      "loss": 0.1578,
      "step": 1304
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8738009768592757e-05,
      "loss": 0.1941,
      "step": 1305
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8734813183558942e-05,
      "loss": 0.1629,
      "step": 1306
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8731612828587136e-05,
      "loss": 0.2066,
      "step": 1307
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8728408705058604e-05,
      "loss": 0.18,
      "step": 1308
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.872520081435625e-05,
      "loss": 0.1574,
      "step": 1309
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8721989157864594e-05,
      "loss": 0.1667,
      "step": 1310
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.871877373696978e-05,
      "loss": 0.1351,
      "step": 1311
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8715554553059587e-05,
      "loss": 0.1686,
      "step": 1312
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8712331607523406e-05,
      "loss": 0.1633,
      "step": 1313
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.8709104901752263e-05,
      "loss": 0.1742,
      "step": 1314
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.87058744371388e-05,
      "loss": 0.1474,
      "step": 1315
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8702640215077283e-05,
      "loss": 0.1599,
      "step": 1316
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8699402236963598e-05,
      "loss": 0.1605,
      "step": 1317
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8696160504195253e-05,
      "loss": 0.1308,
      "step": 1318
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8692915018171384e-05,
      "loss": 0.1789,
      "step": 1319
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8689665780292727e-05,
      "loss": 0.1474,
      "step": 1320
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.868641279196166e-05,
      "loss": 0.155,
      "step": 1321
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.868315605458217e-05,
      "loss": 0.1644,
      "step": 1322
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.867989556955986e-05,
      "loss": 0.237,
      "step": 1323
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8676631338301955e-05,
      "loss": 0.154,
      "step": 1324
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8673363362217287e-05,
      "loss": 0.1614,
      "step": 1325
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.867009164271632e-05,
      "loss": 0.1701,
      "step": 1326
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8666816181211118e-05,
      "loss": 0.1545,
      "step": 1327
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8663536979115376e-05,
      "loss": 0.1761,
      "step": 1328
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.866025403784439e-05,
      "loss": 0.1489,
      "step": 1329
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8656967358815073e-05,
      "loss": 0.2062,
      "step": 1330
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8653676943445956e-05,
      "loss": 0.1711,
      "step": 1331
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8650382793157177e-05,
      "loss": 0.1723,
      "step": 1332
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8647084909370492e-05,
      "loss": 0.1537,
      "step": 1333
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8643783293509262e-05,
      "loss": 0.145,
      "step": 1334
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8640477946998464e-05,
      "loss": 0.1561,
      "step": 1335
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.863716887126468e-05,
      "loss": 0.1942,
      "step": 1336
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.86338560677361e-05,
      "loss": 0.1781,
      "step": 1337
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8630539537842544e-05,
      "loss": 0.1614,
      "step": 1338
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8627219283015404e-05,
      "loss": 0.1785,
      "step": 1339
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.862389530468771e-05,
      "loss": 0.1814,
      "step": 1340
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.8620567604294083e-05,
      "loss": 0.182,
      "step": 1341
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8617236183270764e-05,
      "loss": 0.1634,
      "step": 1342
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8613901043055583e-05,
      "loss": 0.1961,
      "step": 1343
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8610562185087985e-05,
      "loss": 0.1457,
      "step": 1344
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.860721961080902e-05,
      "loss": 0.1588,
      "step": 1345
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.860387332166134e-05,
      "loss": 0.1543,
      "step": 1346
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8600523319089202e-05,
      "loss": 0.1517,
      "step": 1347
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.859716960453846e-05,
      "loss": 0.1691,
      "step": 1348
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8593812179456582e-05,
      "loss": 0.1575,
      "step": 1349
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8590451045292624e-05,
      "loss": 0.1676,
      "step": 1350
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8587086203497245e-05,
      "loss": 0.1278,
      "step": 1351
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8583717655522717e-05,
      "loss": 0.1609,
      "step": 1352
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.85803454028229e-05,
      "loss": 0.1675,
      "step": 1353
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8576969446853248e-05,
      "loss": 0.1519,
      "step": 1354
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8573589789070826e-05,
      "loss": 0.1752,
      "step": 1355
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8570206430934297e-05,
      "loss": 0.1808,
      "step": 1356
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.856681937390391e-05,
      "loss": 0.1538,
      "step": 1357
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.856342861944151e-05,
      "loss": 0.212,
      "step": 1358
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8560034169010556e-05,
      "loss": 0.1341,
      "step": 1359
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.855663602407608e-05,
      "loss": 0.1168,
      "step": 1360
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8553234186104726e-05,
      "loss": 0.1582,
      "step": 1361
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8549828656564717e-05,
      "loss": 0.1892,
      "step": 1362
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.854641943692588e-05,
      "loss": 0.1872,
      "step": 1363
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8543006528659634e-05,
      "loss": 0.1357,
      "step": 1364
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8539589933238985e-05,
      "loss": 0.1802,
      "step": 1365
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8536169652138532e-05,
      "loss": 0.1459,
      "step": 1366
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.853274568683447e-05,
      "loss": 0.1683,
      "step": 1367
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.852931803880457e-05,
      "loss": 0.1307,
      "step": 1368
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.852588670952821e-05,
      "loss": 0.1799,
      "step": 1369
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8522451700486348e-05,
      "loss": 0.1505,
      "step": 1370
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8519013013161526e-05,
      "loss": 0.147,
      "step": 1371
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8515570649037885e-05,
      "loss": 0.1422,
      "step": 1372
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8512124609601143e-05,
      "loss": 0.1587,
      "step": 1373
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8508674896338606e-05,
      "loss": 0.1329,
      "step": 1374
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8505221510739174e-05,
      "loss": 0.1841,
      "step": 1375
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8501764454293315e-05,
      "loss": 0.1821,
      "step": 1376
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8498303728493102e-05,
      "loss": 0.1463,
      "step": 1377
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8494839334832176e-05,
      "loss": 0.1529,
      "step": 1378
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8491371274805767e-05,
      "loss": 0.1309,
      "step": 1379
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.848789954991069e-05,
      "loss": 0.1492,
      "step": 1380
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8484424161645333e-05,
      "loss": 0.1691,
      "step": 1381
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.848094511150967e-05,
      "loss": 0.1745,
      "step": 1382
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8477462401005262e-05,
      "loss": 0.1753,
      "step": 1383
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.847397603163524e-05,
      "loss": 0.1625,
      "step": 1384
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.847048600490432e-05,
      "loss": 0.1538,
      "step": 1385
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8466992322318794e-05,
      "loss": 0.1842,
      "step": 1386
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8463494985386534e-05,
      "loss": 0.1577,
      "step": 1387
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8459993995616984e-05,
      "loss": 0.1612,
      "step": 1388
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.845648935452117e-05,
      "loss": 0.1827,
      "step": 1389
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8452981063611696e-05,
      "loss": 0.1536,
      "step": 1390
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.844946912440273e-05,
      "loss": 0.1623,
      "step": 1391
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8445953538410026e-05,
      "loss": 0.1305,
      "step": 1392
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.844243430715091e-05,
      "loss": 0.1677,
      "step": 1393
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8438911432144274e-05,
      "loss": 0.185,
      "step": 1394
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8435384914910592e-05,
      "loss": 0.1846,
      "step": 1395
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8431854756971905e-05,
      "loss": 0.168,
      "step": 1396
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8428320959851823e-05,
      "loss": 0.1755,
      "step": 1397
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8424783525075538e-05,
      "loss": 0.1364,
      "step": 1398
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8421242454169794e-05,
      "loss": 0.1369,
      "step": 1399
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8417697748662918e-05,
      "loss": 0.1733,
      "step": 1400
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8414149410084806e-05,
      "loss": 0.1681,
      "step": 1401
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8410597439966915e-05,
      "loss": 0.1756,
      "step": 1402
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8407041839842272e-05,
      "loss": 0.1582,
      "step": 1403
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8403482611245468e-05,
      "loss": 0.1693,
      "step": 1404
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8399919755712667e-05,
      "loss": 0.1636,
      "step": 1405
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.839635327478159e-05,
      "loss": 0.1845,
      "step": 1406
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8392783169991538e-05,
      "loss": 0.1527,
      "step": 1407
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8389209442883355e-05,
      "loss": 0.1512,
      "step": 1408
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.838563209499946e-05,
      "loss": 0.158,
      "step": 1409
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8382051127883836e-05,
      "loss": 0.1832,
      "step": 1410
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8378466543082022e-05,
      "loss": 0.1352,
      "step": 1411
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8374878342141126e-05,
      "loss": 0.1465,
      "step": 1412
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8371286526609812e-05,
      "loss": 0.1715,
      "step": 1413
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.83676910980383e-05,
      "loss": 0.1695,
      "step": 1414
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8364092057978376e-05,
      "loss": 0.2078,
      "step": 1415
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8360489407983385e-05,
      "loss": 0.17,
      "step": 1416
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8356883149608228e-05,
      "loss": 0.1881,
      "step": 1417
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8353273284409356e-05,
      "loss": 0.136,
      "step": 1418
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8349659813944794e-05,
      "loss": 0.1576,
      "step": 1419
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8346042739774105e-05,
      "loss": 0.1446,
      "step": 1420
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.834242206345842e-05,
      "loss": 0.1336,
      "step": 1421
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8338797786560416e-05,
      "loss": 0.1482,
      "step": 1422
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8335169910644334e-05,
      "loss": 0.188,
      "step": 1423
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8331538437275954e-05,
      "loss": 0.2101,
      "step": 1424
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8327903368022618e-05,
      "loss": 0.158,
      "step": 1425
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8324264704453228e-05,
      "loss": 0.1409,
      "step": 1426
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8320622448138216e-05,
      "loss": 0.1666,
      "step": 1427
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8316976600649584e-05,
      "loss": 0.1884,
      "step": 1428
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8313327163560876e-05,
      "loss": 0.1692,
      "step": 1429
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8309674138447187e-05,
      "loss": 0.154,
      "step": 1430
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8306017526885153e-05,
      "loss": 0.1467,
      "step": 1431
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8302357330452972e-05,
      "loss": 0.1577,
      "step": 1432
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.829869355073038e-05,
      "loss": 0.1678,
      "step": 1433
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8295026189298658e-05,
      "loss": 0.1752,
      "step": 1434
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.829135524774064e-05,
      "loss": 0.1835,
      "step": 1435
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8287680727640695e-05,
      "loss": 0.1351,
      "step": 1436
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.828400263058475e-05,
      "loss": 0.1548,
      "step": 1437
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8280320958160268e-05,
      "loss": 0.1348,
      "step": 1438
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.827663571195625e-05,
      "loss": 0.1563,
      "step": 1439
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8272946893563247e-05,
      "loss": 0.1367,
      "step": 1440
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8269254504573358e-05,
      "loss": 0.1713,
      "step": 1441
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8265558546580205e-05,
      "loss": 0.1931,
      "step": 1442
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.826185902117896e-05,
      "loss": 0.1659,
      "step": 1443
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8258155929966344e-05,
      "loss": 0.1598,
      "step": 1444
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.82544492745406e-05,
      "loss": 0.2016,
      "step": 1445
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.825073905650152e-05,
      "loss": 0.1802,
      "step": 1446
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8247025277450438e-05,
      "loss": 0.1771,
      "step": 1447
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8243307938990203e-05,
      "loss": 0.1509,
      "step": 1448
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8239587042725225e-05,
      "loss": 0.1457,
      "step": 1449
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.823586259026144e-05,
      "loss": 0.1544,
      "step": 1450
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8232134583206312e-05,
      "loss": 0.1701,
      "step": 1451
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8228403023168856e-05,
      "loss": 0.1877,
      "step": 1452
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8224667911759602e-05,
      "loss": 0.1827,
      "step": 1453
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8220929250590625e-05,
      "loss": 0.1499,
      "step": 1454
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8217187041275523e-05,
      "loss": 0.168,
      "step": 1455
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8213441285429435e-05,
      "loss": 0.1617,
      "step": 1456
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8209691984669023e-05,
      "loss": 0.1788,
      "step": 1457
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8205939140612485e-05,
      "loss": 0.159,
      "step": 1458
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8202182754879544e-05,
      "loss": 0.1681,
      "step": 1459
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8198422829091457e-05,
      "loss": 0.1695,
      "step": 1460
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8194659364870997e-05,
      "loss": 0.1619,
      "step": 1461
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8190892363842474e-05,
      "loss": 0.1359,
      "step": 1462
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8187121827631727e-05,
      "loss": 0.1816,
      "step": 1463
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8183347757866112e-05,
      "loss": 0.1778,
      "step": 1464
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8179570156174518e-05,
      "loss": 0.1408,
      "step": 1465
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.817578902418735e-05,
      "loss": 0.1311,
      "step": 1466
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8172004363536545e-05,
      "loss": 0.1532,
      "step": 1467
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8168216175855557e-05,
      "loss": 0.1665,
      "step": 1468
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8164424462779365e-05,
      "loss": 0.1505,
      "step": 1469
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8160629225944473e-05,
      "loss": 0.1667,
      "step": 1470
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.81568304669889e-05,
      "loss": 0.1701,
      "step": 1471
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.815302818755218e-05,
      "loss": 0.1646,
      "step": 1472
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8149222389275386e-05,
      "loss": 0.1756,
      "step": 1473
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.814541307380109e-05,
      "loss": 0.1552,
      "step": 1474
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8141600242773388e-05,
      "loss": 0.1849,
      "step": 1475
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8137783897837897e-05,
      "loss": 0.1548,
      "step": 1476
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8133964040641746e-05,
      "loss": 0.1751,
      "step": 1477
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.813014067283359e-05,
      "loss": 0.1878,
      "step": 1478
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8126313796063585e-05,
      "loss": 0.1385,
      "step": 1479
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8122483411983403e-05,
      "loss": 0.2047,
      "step": 1480
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8118649522246245e-05,
      "loss": 0.162,
      "step": 1481
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.811481212850681e-05,
      "loss": 0.1435,
      "step": 1482
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.811097123242131e-05,
      "loss": 0.1988,
      "step": 1483
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.810712683564748e-05,
      "loss": 0.175,
      "step": 1484
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.810327893984455e-05,
      "loss": 0.1434,
      "step": 1485
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8099427546673277e-05,
      "loss": 0.1656,
      "step": 1486
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8095572657795917e-05,
      "loss": 0.1726,
      "step": 1487
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.809171427487623e-05,
      "loss": 0.1586,
      "step": 1488
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8087852399579494e-05,
      "loss": 0.1609,
      "step": 1489
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.80839870335725e-05,
      "loss": 0.1411,
      "step": 1490
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8080118178523523e-05,
      "loss": 0.1735,
      "step": 1491
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8076245836102368e-05,
      "loss": 0.1497,
      "step": 1492
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.807237000798033e-05,
      "loss": 0.1796,
      "step": 1493
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.806849069583021e-05,
      "loss": 0.1709,
      "step": 1494
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.806460790132632e-05,
      "loss": 0.1641,
      "step": 1495
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8060721626144477e-05,
      "loss": 0.1459,
      "step": 1496
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8056831871961985e-05,
      "loss": 0.1542,
      "step": 1497
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.805293864045766e-05,
      "loss": 0.1444,
      "step": 1498
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.804904193331182e-05,
      "loss": 0.183,
      "step": 1499
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8045141752206277e-05,
      "loss": 0.1418,
      "step": 1500
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.804123809882435e-05,
      "loss": 0.1582,
      "step": 1501
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8037330974850852e-05,
      "loss": 0.1703,
      "step": 1502
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.803342038197209e-05,
      "loss": 0.1667,
      "step": 1503
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8029506321875875e-05,
      "loss": 0.1359,
      "step": 1504
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8025588796251516e-05,
      "loss": 0.2008,
      "step": 1505
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.802166780678981e-05,
      "loss": 0.191,
      "step": 1506
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.801774335518305e-05,
      "loss": 0.1865,
      "step": 1507
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.801381544312503e-05,
      "loss": 0.1465,
      "step": 1508
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8009884072311035e-05,
      "loss": 0.1835,
      "step": 1509
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8005949244437836e-05,
      "loss": 0.159,
      "step": 1510
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8002010961203707e-05,
      "loss": 0.1586,
      "step": 1511
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7998069224308407e-05,
      "loss": 0.1445,
      "step": 1512
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.799412403545318e-05,
      "loss": 0.1686,
      "step": 1513
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.799017539634077e-05,
      "loss": 0.1868,
      "step": 1514
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.798622330867541e-05,
      "loss": 0.1706,
      "step": 1515
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7982267774162813e-05,
      "loss": 0.1317,
      "step": 1516
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7978308794510188e-05,
      "loss": 0.1366,
      "step": 1517
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7974346371426225e-05,
      "loss": 0.139,
      "step": 1518
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.79703805066211e-05,
      "loss": 0.1517,
      "step": 1519
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.796641120180648e-05,
      "loss": 0.1647,
      "step": 1520
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7962438458695514e-05,
      "loss": 0.1326,
      "step": 1521
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7958462279002834e-05,
      "loss": 0.1844,
      "step": 1522
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7954482664444556e-05,
      "loss": 0.1593,
      "step": 1523
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7950499616738276e-05,
      "loss": 0.144,
      "step": 1524
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7946513137603078e-05,
      "loss": 0.1737,
      "step": 1525
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.7942523228759516e-05,
      "loss": 0.1863,
      "step": 1526
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.793852989192964e-05,
      "loss": 0.1789,
      "step": 1527
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.793453312883697e-05,
      "loss": 0.2023,
      "step": 1528
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.79305329412065e-05,
      "loss": 0.1364,
      "step": 1529
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7926529330764707e-05,
      "loss": 0.1645,
      "step": 1530
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7922522299239552e-05,
      "loss": 0.1484,
      "step": 1531
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.791851184836046e-05,
      "loss": 0.1477,
      "step": 1532
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7914497979858346e-05,
      "loss": 0.165,
      "step": 1533
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.791048069546559e-05,
      "loss": 0.1658,
      "step": 1534
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.790645999691604e-05,
      "loss": 0.14,
      "step": 1535
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7902435885945037e-05,
      "loss": 0.1417,
      "step": 1536
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.789840836428938e-05,
      "loss": 0.1544,
      "step": 1537
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7894377433687337e-05,
      "loss": 0.1494,
      "step": 1538
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7890343095878666e-05,
      "loss": 0.1975,
      "step": 1539
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.788630535260458e-05,
      "loss": 0.1663,
      "step": 1540
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.788226420560776e-05,
      "loss": 0.1638,
      "step": 1541
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7878219656632368e-05,
      "loss": 0.1531,
      "step": 1542
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.787417170742402e-05,
      "loss": 0.1954,
      "step": 1543
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7870120359729822e-05,
      "loss": 0.1769,
      "step": 1544
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7866065615298322e-05,
      "loss": 0.1524,
      "step": 1545
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.786200747587954e-05,
      "loss": 0.1444,
      "step": 1546
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7857945943224974e-05,
      "loss": 0.1357,
      "step": 1547
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7853881019087572e-05,
      "loss": 0.1488,
      "step": 1548
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7849812705221762e-05,
      "loss": 0.1966,
      "step": 1549
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7845741003383412e-05,
      "loss": 0.1652,
      "step": 1550
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7841665915329874e-05,
      "loss": 0.1642,
      "step": 1551
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7837587442819946e-05,
      "loss": 0.1885,
      "step": 1552
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.78335055876139e-05,
      "loss": 0.1708,
      "step": 1553
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.7829420351473456e-05,
      "loss": 0.1324,
      "step": 1554
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.78253317361618e-05,
      "loss": 0.1732,
      "step": 1555
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.782123974344358e-05,
      "loss": 0.1769,
      "step": 1556
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7817144375084888e-05,
      "loss": 0.1787,
      "step": 1557
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7813045632853284e-05,
      "loss": 0.2123,
      "step": 1558
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7808943518517787e-05,
      "loss": 0.1864,
      "step": 1559
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7804838033848863e-05,
      "loss": 0.1639,
      "step": 1560
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.780072918061843e-05,
      "loss": 0.1363,
      "step": 1561
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7796616960599874e-05,
      "loss": 0.1455,
      "step": 1562
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7792501375568024e-05,
      "loss": 0.152,
      "step": 1563
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7788382427299158e-05,
      "loss": 0.1474,
      "step": 1564
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.778426011757101e-05,
      "loss": 0.1482,
      "step": 1565
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7780134448162777e-05,
      "loss": 0.1441,
      "step": 1566
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7776005420855083e-05,
      "loss": 0.15,
      "step": 1567
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7771873037430018e-05,
      "loss": 0.1592,
      "step": 1568
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7767737299671113e-05,
      "loss": 0.1661,
      "step": 1569
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.776359820936335e-05,
      "loss": 0.2081,
      "step": 1570
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7759455768293156e-05,
      "loss": 0.1628,
      "step": 1571
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7755309978248404e-05,
      "loss": 0.1727,
      "step": 1572
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7751160841018418e-05,
      "loss": 0.1651,
      "step": 1573
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7747008358393958e-05,
      "loss": 0.1633,
      "step": 1574
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.774285253216723e-05,
      "loss": 0.1545,
      "step": 1575
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7738693364131893e-05,
      "loss": 0.157,
      "step": 1576
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7734530856083033e-05,
      "loss": 0.1992,
      "step": 1577
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7730365009817186e-05,
      "loss": 0.1549,
      "step": 1578
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.772619582713233e-05,
      "loss": 0.1853,
      "step": 1579
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7722023309827885e-05,
      "loss": 0.148,
      "step": 1580
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.7717847459704696e-05,
      "loss": 0.2105,
      "step": 1581
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7713668278565062e-05,
      "loss": 0.1465,
      "step": 1582
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7709485768212716e-05,
      "loss": 0.1752,
      "step": 1583
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.770529993045282e-05,
      "loss": 0.1845,
      "step": 1584
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7701110767091983e-05,
      "loss": 0.1457,
      "step": 1585
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7696918279938242e-05,
      "loss": 0.1543,
      "step": 1586
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7692722470801073e-05,
      "loss": 0.1719,
      "step": 1587
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.768852334149138e-05,
      "loss": 0.1413,
      "step": 1588
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7684320893821506e-05,
      "loss": 0.1515,
      "step": 1589
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7680115129605228e-05,
      "loss": 0.1342,
      "step": 1590
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7675906050657738e-05,
      "loss": 0.1677,
      "step": 1591
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7671693658795687e-05,
      "loss": 0.1726,
      "step": 1592
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7667477955837128e-05,
      "loss": 0.1557,
      "step": 1593
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.766325894360156e-05,
      "loss": 0.1582,
      "step": 1594
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7659036623909903e-05,
      "loss": 0.1257,
      "step": 1595
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.765481099858451e-05,
      "loss": 0.188,
      "step": 1596
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.765058206944915e-05,
      "loss": 0.1921,
      "step": 1597
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7646349838329035e-05,
      "loss": 0.1399,
      "step": 1598
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7642114307050785e-05,
      "loss": 0.1676,
      "step": 1599
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.763787547744245e-05,
      "loss": 0.1495,
      "step": 1600
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7633633351333515e-05,
      "loss": 0.1607,
      "step": 1601
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.762938793055487e-05,
      "loss": 0.1919,
      "step": 1602
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.762513921693884e-05,
      "loss": 0.1779,
      "step": 1603
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7620887212319163e-05,
      "loss": 0.1414,
      "step": 1604
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7616631918530996e-05,
      "loss": 0.1825,
      "step": 1605
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.761237333741093e-05,
      "loss": 0.1905,
      "step": 1606
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.7608111470796963e-05,
      "loss": 0.1905,
      "step": 1607
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7603846320528507e-05,
      "loss": 0.1477,
      "step": 1608
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.75995778884464e-05,
      "loss": 0.1274,
      "step": 1609
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7595306176392894e-05,
      "loss": 0.1853,
      "step": 1610
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7591031186211662e-05,
      "loss": 0.1917,
      "step": 1611
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7586752919747776e-05,
      "loss": 0.1559,
      "step": 1612
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7582471378847738e-05,
      "loss": 0.1448,
      "step": 1613
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7578186565359458e-05,
      "loss": 0.1871,
      "step": 1614
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7573898481132253e-05,
      "loss": 0.1529,
      "step": 1615
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.756960712801686e-05,
      "loss": 0.1848,
      "step": 1616
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7565312507865426e-05,
      "loss": 0.185,
      "step": 1617
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.75610146225315e-05,
      "loss": 0.1965,
      "step": 1618
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7556713473870054e-05,
      "loss": 0.1509,
      "step": 1619
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.755240906373745e-05,
      "loss": 0.1531,
      "step": 1620
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7548101393991475e-05,
      "loss": 0.156,
      "step": 1621
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.754379046649131e-05,
      "loss": 0.1804,
      "step": 1622
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7539476283097552e-05,
      "loss": 0.1914,
      "step": 1623
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.75351588456722e-05,
      "loss": 0.1591,
      "step": 1624
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.753083815607865e-05,
      "loss": 0.1553,
      "step": 1625
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.752651421618171e-05,
      "loss": 0.151,
      "step": 1626
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7522187027847595e-05,
      "loss": 0.1445,
      "step": 1627
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.751785659294391e-05,
      "loss": 0.1832,
      "step": 1628
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.751352291333967e-05,
      "loss": 0.1613,
      "step": 1629
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7509185990905285e-05,
      "loss": 0.1168,
      "step": 1630
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7504845827512567e-05,
      "loss": 0.1444,
      "step": 1631
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7500502425034728e-05,
      "loss": 0.1225,
      "step": 1632
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.7496155785346377e-05,
      "loss": 0.1607,
      "step": 1633
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.749180591032352e-05,
      "loss": 0.1395,
      "step": 1634
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.748745280184356e-05,
      "loss": 0.1859,
      "step": 1635
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7483096461785295e-05,
      "loss": 0.164,
      "step": 1636
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7478736892028917e-05,
      "loss": 0.1597,
      "step": 1637
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7474374094456007e-05,
      "loss": 0.1386,
      "step": 1638
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7470008070949552e-05,
      "loss": 0.181,
      "step": 1639
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7465638823393926e-05,
      "loss": 0.1605,
      "step": 1640
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.746126635367489e-05,
      "loss": 0.1579,
      "step": 1641
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7456890663679593e-05,
      "loss": 0.176,
      "step": 1642
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.745251175529658e-05,
      "loss": 0.169,
      "step": 1643
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7448129630415793e-05,
      "loss": 0.1712,
      "step": 1644
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7443744290928547e-05,
      "loss": 0.1709,
      "step": 1645
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.743935573872755e-05,
      "loss": 0.2108,
      "step": 1646
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.74349639757069e-05,
      "loss": 0.1563,
      "step": 1647
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7430569003762077e-05,
      "loss": 0.1687,
      "step": 1648
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7426170824789944e-05,
      "loss": 0.201,
      "step": 1649
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7421769440688757e-05,
      "loss": 0.125,
      "step": 1650
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7417364853358147e-05,
      "loss": 0.157,
      "step": 1651
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7412957064699134e-05,
      "loss": 0.1674,
      "step": 1652
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7408546076614107e-05,
      "loss": 0.1613,
      "step": 1653
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7404131891006852e-05,
      "loss": 0.1904,
      "step": 1654
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7399714509782528e-05,
      "loss": 0.1518,
      "step": 1655
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.739529393484767e-05,
      "loss": 0.1369,
      "step": 1656
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7390870168110193e-05,
      "loss": 0.1577,
      "step": 1657
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7386443211479395e-05,
      "loss": 0.146,
      "step": 1658
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.7382013066865946e-05,
      "loss": 0.1442,
      "step": 1659
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.737757973618189e-05,
      "loss": 0.1749,
      "step": 1660
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.737314322134065e-05,
      "loss": 0.1618,
      "step": 1661
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7368703524257023e-05,
      "loss": 0.1453,
      "step": 1662
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.736426064684718e-05,
      "loss": 0.1264,
      "step": 1663
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7359814591028658e-05,
      "loss": 0.1386,
      "step": 1664
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.735536535872038e-05,
      "loss": 0.1663,
      "step": 1665
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.735091295184262e-05,
      "loss": 0.1743,
      "step": 1666
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7346457372317035e-05,
      "loss": 0.1513,
      "step": 1667
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7341998622066657e-05,
      "loss": 0.1656,
      "step": 1668
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7337536703015872e-05,
      "loss": 0.171,
      "step": 1669
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7333071617090444e-05,
      "loss": 0.1971,
      "step": 1670
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7328603366217497e-05,
      "loss": 0.1477,
      "step": 1671
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.732413195232553e-05,
      "loss": 0.1597,
      "step": 1672
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7319657377344398e-05,
      "loss": 0.1687,
      "step": 1673
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7315179643205323e-05,
      "loss": 0.1624,
      "step": 1674
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.731069875184089e-05,
      "loss": 0.1587,
      "step": 1675
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7306214705185048e-05,
      "loss": 0.1336,
      "step": 1676
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7301727505173114e-05,
      "loss": 0.1718,
      "step": 1677
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7297237153741754e-05,
      "loss": 0.1408,
      "step": 1678
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7292743652829e-05,
      "loss": 0.1481,
      "step": 1679
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7288247004374244e-05,
      "loss": 0.181,
      "step": 1680
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7283747210318237e-05,
      "loss": 0.1686,
      "step": 1681
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7279244272603083e-05,
      "loss": 0.1738,
      "step": 1682
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.727473819317225e-05,
      "loss": 0.1617,
      "step": 1683
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.727022897397056e-05,
      "loss": 0.1523,
      "step": 1684
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7265716616944182e-05,
      "loss": 0.158,
      "step": 1685
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7261201124040646e-05,
      "loss": 0.1606,
      "step": 1686
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7256682497208843e-05,
      "loss": 0.188,
      "step": 1687
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7252160738398998e-05,
      "loss": 0.1672,
      "step": 1688
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7247635849562706e-05,
      "loss": 0.1488,
      "step": 1689
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.72431078326529e-05,
      "loss": 0.1838,
      "step": 1690
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.723857668962387e-05,
      "loss": 0.1648,
      "step": 1691
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7234042422431258e-05,
      "loss": 0.1531,
      "step": 1692
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7229505033032037e-05,
      "loss": 0.1622,
      "step": 1693
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.722496452338456e-05,
      "loss": 0.1425,
      "step": 1694
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.722042089544849e-05,
      "loss": 0.1904,
      "step": 1695
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.721587415118486e-05,
      "loss": 0.1531,
      "step": 1696
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.721132429255604e-05,
      "loss": 0.1571,
      "step": 1697
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7206771321525745e-05,
      "loss": 0.1618,
      "step": 1698
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7202215240059038e-05,
      "loss": 0.1442,
      "step": 1699
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.719765605012231e-05,
      "loss": 0.1565,
      "step": 1700
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7193093753683314e-05,
      "loss": 0.1452,
      "step": 1701
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.718852835271113e-05,
      "loss": 0.168,
      "step": 1702
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7183959849176176e-05,
      "loss": 0.1566,
      "step": 1703
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.717938824505022e-05,
      "loss": 0.156,
      "step": 1704
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.717481354230636e-05,
      "loss": 0.1512,
      "step": 1705
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.717023574291903e-05,
      "loss": 0.1806,
      "step": 1706
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7165654848864014e-05,
      "loss": 0.1893,
      "step": 1707
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.716107086211842e-05,
      "loss": 0.1622,
      "step": 1708
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.715648378466068e-05,
      "loss": 0.1597,
      "step": 1709
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.715189361847059e-05,
      "loss": 0.155,
      "step": 1710
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7147300365529256e-05,
      "loss": 0.1304,
      "step": 1711
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.7142704027819115e-05,
      "loss": 0.1111,
      "step": 1712
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.713810460732395e-05,
      "loss": 0.1446,
      "step": 1713
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7133502106028864e-05,
      "loss": 0.1656,
      "step": 1714
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7128896525920294e-05,
      "loss": 0.1668,
      "step": 1715
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7124287868986004e-05,
      "loss": 0.1657,
      "step": 1716
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7119676137215088e-05,
      "loss": 0.1631,
      "step": 1717
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.711506133259796e-05,
      "loss": 0.2102,
      "step": 1718
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.711044345712637e-05,
      "loss": 0.1625,
      "step": 1719
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7105822512793392e-05,
      "loss": 0.1388,
      "step": 1720
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.710119850159342e-05,
      "loss": 0.1753,
      "step": 1721
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.709657142552217e-05,
      "loss": 0.1437,
      "step": 1722
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7091941286576684e-05,
      "loss": 0.1744,
      "step": 1723
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.708730808675533e-05,
      "loss": 0.176,
      "step": 1724
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.708267182805779e-05,
      "loss": 0.1823,
      "step": 1725
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7078032512485073e-05,
      "loss": 0.1625,
      "step": 1726
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7073390142039498e-05,
      "loss": 0.1959,
      "step": 1727
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7068744718724713e-05,
      "loss": 0.165,
      "step": 1728
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7064096244545677e-05,
      "loss": 0.1586,
      "step": 1729
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7059444721508667e-05,
      "loss": 0.1655,
      "step": 1730
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7054790151621277e-05,
      "loss": 0.1591,
      "step": 1731
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7050132536892415e-05,
      "loss": 0.137,
      "step": 1732
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7045471879332308e-05,
      "loss": 0.1834,
      "step": 1733
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7040808180952485e-05,
      "loss": 0.1731,
      "step": 1734
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7036141443765796e-05,
      "loss": 0.1805,
      "step": 1735
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7031471669786402e-05,
      "loss": 0.1922,
      "step": 1736
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.7026798861029776e-05,
      "loss": 0.1977,
      "step": 1737
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.70221230195127e-05,
      "loss": 0.1474,
      "step": 1738
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.701744414725325e-05,
      "loss": 0.16,
      "step": 1739
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.701276224627084e-05,
      "loss": 0.1759,
      "step": 1740
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.7008077318586162e-05,
      "loss": 0.18,
      "step": 1741
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.700338936622123e-05,
      "loss": 0.1722,
      "step": 1742
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.699869839119937e-05,
      "loss": 0.1488,
      "step": 1743
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.699400439554519e-05,
      "loss": 0.1567,
      "step": 1744
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.698930738128462e-05,
      "loss": 0.1575,
      "step": 1745
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6984607350444885e-05,
      "loss": 0.1555,
      "step": 1746
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.697990430505452e-05,
      "loss": 0.1571,
      "step": 1747
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.697519824714335e-05,
      "loss": 0.16,
      "step": 1748
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6970489178742505e-05,
      "loss": 0.1865,
      "step": 1749
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6965777101884423e-05,
      "loss": 0.1693,
      "step": 1750
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6961062018602824e-05,
      "loss": 0.1808,
      "step": 1751
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6956343930932733e-05,
      "loss": 0.1625,
      "step": 1752
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.695162284091048e-05,
      "loss": 0.1797,
      "step": 1753
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.694689875057368e-05,
      "loss": 0.1492,
      "step": 1754
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.694217166196124e-05,
      "loss": 0.1763,
      "step": 1755
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.693744157711338e-05,
      "loss": 0.151,
      "step": 1756
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6932708498071595e-05,
      "loss": 0.1764,
      "step": 1757
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6927972426878674e-05,
      "loss": 0.1498,
      "step": 1758
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6923233365578706e-05,
      "loss": 0.1783,
      "step": 1759
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6918491316217063e-05,
      "loss": 0.1491,
      "step": 1760
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.691374628084041e-05,
      "loss": 0.1397,
      "step": 1761
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6908998261496704e-05,
      "loss": 0.1594,
      "step": 1762
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.690424726023518e-05,
      "loss": 0.1582,
      "step": 1763
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6899493279106373e-05,
      "loss": 0.1479,
      "step": 1764
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6894736320162087e-05,
      "loss": 0.1683,
      "step": 1765
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.6889976385455435e-05,
      "loss": 0.1492,
      "step": 1766
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.688521347704079e-05,
      "loss": 0.1649,
      "step": 1767
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6880447596973826e-05,
      "loss": 0.1575,
      "step": 1768
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6875678747311486e-05,
      "loss": 0.1399,
      "step": 1769
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.687090693011201e-05,
      "loss": 0.161,
      "step": 1770
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6866132147434902e-05,
      "loss": 0.1387,
      "step": 1771
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.686135440134096e-05,
      "loss": 0.1353,
      "step": 1772
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.685657369389225e-05,
      "loss": 0.1957,
      "step": 1773
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.685179002715213e-05,
      "loss": 0.166,
      "step": 1774
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6847003403185214e-05,
      "loss": 0.1352,
      "step": 1775
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6842213824057414e-05,
      "loss": 0.153,
      "step": 1776
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6837421291835908e-05,
      "loss": 0.1543,
      "step": 1777
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6832625808589144e-05,
      "loss": 0.1333,
      "step": 1778
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6827827376386855e-05,
      "loss": 0.1726,
      "step": 1779
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6823025997300034e-05,
      "loss": 0.1722,
      "step": 1780
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6818221673400956e-05,
      "loss": 0.1395,
      "step": 1781
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.681341440676316e-05,
      "loss": 0.1416,
      "step": 1782
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.680860419946146e-05,
      "loss": 0.1544,
      "step": 1783
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.680379105357194e-05,
      "loss": 0.133,
      "step": 1784
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6798974971171946e-05,
      "loss": 0.1583,
      "step": 1785
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6794155954340098e-05,
      "loss": 0.1566,
      "step": 1786
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6789334005156278e-05,
      "loss": 0.1632,
      "step": 1787
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6784509125701634e-05,
      "loss": 0.1432,
      "step": 1788
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6779681318058577e-05,
      "loss": 0.1862,
      "step": 1789
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6774850584310795e-05,
      "loss": 0.1378,
      "step": 1790
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.677001692654322e-05,
      "loss": 0.1465,
      "step": 1791
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.6765180346842057e-05,
      "loss": 0.1549,
      "step": 1792
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.676034084729477e-05,
      "loss": 0.1638,
      "step": 1793
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6755498429990082e-05,
      "loss": 0.1456,
      "step": 1794
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6750653097017978e-05,
      "loss": 0.1648,
      "step": 1795
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6745804850469694e-05,
      "loss": 0.1998,
      "step": 1796
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.674095369243774e-05,
      "loss": 0.186,
      "step": 1797
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6736099625015862e-05,
      "loss": 0.1688,
      "step": 1798
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6731242650299074e-05,
      "loss": 0.1532,
      "step": 1799
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6726382770383645e-05,
      "loss": 0.1759,
      "step": 1800
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6721519987367094e-05,
      "loss": 0.1664,
      "step": 1801
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.671665430334819e-05,
      "loss": 0.1594,
      "step": 1802
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.671178572042696e-05,
      "loss": 0.1432,
      "step": 1803
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6706914240704682e-05,
      "loss": 0.1513,
      "step": 1804
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6702039866283886e-05,
      "loss": 0.1322,
      "step": 1805
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6697162599268342e-05,
      "loss": 0.134,
      "step": 1806
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.669228244176307e-05,
      "loss": 0.1634,
      "step": 1807
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6687399395874356e-05,
      "loss": 0.1226,
      "step": 1808
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.668251346370971e-05,
      "loss": 0.1575,
      "step": 1809
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6677624647377893e-05,
      "loss": 0.1427,
      "step": 1810
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.667273294898892e-05,
      "loss": 0.1367,
      "step": 1811
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.666783837065404e-05,
      "loss": 0.1555,
      "step": 1812
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6662940914485747e-05,
      "loss": 0.1796,
      "step": 1813
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.665804058259779e-05,
      "loss": 0.1825,
      "step": 1814
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6653137377105136e-05,
      "loss": 0.1725,
      "step": 1815
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.664823130012401e-05,
      "loss": 0.1512,
      "step": 1816
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6643322353771866e-05,
      "loss": 0.1476,
      "step": 1817
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.663841054016741e-05,
      "loss": 0.186,
      "step": 1818
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6633495861430568e-05,
      "loss": 0.1602,
      "step": 1819
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.6628578319682515e-05,
      "loss": 0.1555,
      "step": 1820
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6623657917045655e-05,
      "loss": 0.1611,
      "step": 1821
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6618734655643637e-05,
      "loss": 0.1983,
      "step": 1822
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6613808537601327e-05,
      "loss": 0.1631,
      "step": 1823
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.660887956504484e-05,
      "loss": 0.1338,
      "step": 1824
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6603947740101513e-05,
      "loss": 0.1431,
      "step": 1825
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.659901306489992e-05,
      "loss": 0.1563,
      "step": 1826
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6594075541569857e-05,
      "loss": 0.1507,
      "step": 1827
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6589135172242363e-05,
      "loss": 0.1256,
      "step": 1828
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6584191959049695e-05,
      "loss": 0.1476,
      "step": 1829
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6579245904125335e-05,
      "loss": 0.1627,
      "step": 1830
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6574297009604e-05,
      "loss": 0.1692,
      "step": 1831
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6569345277621625e-05,
      "loss": 0.1569,
      "step": 1832
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.656439071031538e-05,
      "loss": 0.1641,
      "step": 1833
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.655943330982365e-05,
      "loss": 0.1474,
      "step": 1834
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6554473078286036e-05,
      "loss": 0.1744,
      "step": 1835
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.654951001784338e-05,
      "loss": 0.1653,
      "step": 1836
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6544544130637733e-05,
      "loss": 0.1544,
      "step": 1837
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.653957541881237e-05,
      "loss": 0.1425,
      "step": 1838
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6534603884511774e-05,
      "loss": 0.1676,
      "step": 1839
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.652962952988166e-05,
      "loss": 0.141,
      "step": 1840
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6524652357068958e-05,
      "loss": 0.1686,
      "step": 1841
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6519672368221808e-05,
      "loss": 0.1794,
      "step": 1842
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6514689565489566e-05,
      "loss": 0.1591,
      "step": 1843
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6509703951022813e-05,
      "loss": 0.1367,
      "step": 1844
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6504715526973334e-05,
      "loss": 0.1431,
      "step": 1845
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.6499724295494125e-05,
      "loss": 0.1322,
      "step": 1846
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.64947302587394e-05,
      "loss": 0.1741,
      "step": 1847
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6489733418864578e-05,
      "loss": 0.158,
      "step": 1848
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6484733778026293e-05,
      "loss": 0.1457,
      "step": 1849
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.647973133838239e-05,
      "loss": 0.1658,
      "step": 1850
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.647472610209191e-05,
      "loss": 0.1741,
      "step": 1851
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6469718071315114e-05,
      "loss": 0.1482,
      "step": 1852
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6464707248213465e-05,
      "loss": 0.1922,
      "step": 1853
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6459693634949624e-05,
      "loss": 0.1451,
      "step": 1854
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6454677233687466e-05,
      "loss": 0.1634,
      "step": 1855
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6449658046592065e-05,
      "loss": 0.1633,
      "step": 1856
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.64446360758297e-05,
      "loss": 0.1415,
      "step": 1857
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6439611323567844e-05,
      "loss": 0.1829,
      "step": 1858
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.643458379197518e-05,
      "loss": 0.1639,
      "step": 1859
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6429553483221584e-05,
      "loss": 0.1533,
      "step": 1860
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6424520399478143e-05,
      "loss": 0.1481,
      "step": 1861
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.641948454291711e-05,
      "loss": 0.184,
      "step": 1862
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.641444591571198e-05,
      "loss": 0.1634,
      "step": 1863
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6409404520037406e-05,
      "loss": 0.1296,
      "step": 1864
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6404360358069256e-05,
      "loss": 0.1636,
      "step": 1865
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6399313431984583e-05,
      "loss": 0.1466,
      "step": 1866
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6394263743961637e-05,
      "loss": 0.156,
      "step": 1867
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.638921129617986e-05,
      "loss": 0.1738,
      "step": 1868
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6384156090819883e-05,
      "loss": 0.1703,
      "step": 1869
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.637909813006353e-05,
      "loss": 0.1341,
      "step": 1870
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.6374037416093806e-05,
      "loss": 0.1463,
      "step": 1871
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.636897395109492e-05,
      "loss": 0.1687,
      "step": 1872
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.636390773725226e-05,
      "loss": 0.1909,
      "step": 1873
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.635883877675239e-05,
      "loss": 0.1544,
      "step": 1874
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6353767071783077e-05,
      "loss": 0.1656,
      "step": 1875
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6348692624533258e-05,
      "loss": 0.1563,
      "step": 1876
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6343615437193066e-05,
      "loss": 0.1473,
      "step": 1877
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6338535511953807e-05,
      "loss": 0.1829,
      "step": 1878
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6333452851007975e-05,
      "loss": 0.1862,
      "step": 1879
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6328367456549245e-05,
      "loss": 0.1544,
      "step": 1880
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.632327933077246e-05,
      "loss": 0.1616,
      "step": 1881
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6318188475873657e-05,
      "loss": 0.1454,
      "step": 1882
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6313094894050047e-05,
      "loss": 0.167,
      "step": 1883
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.630799858750001e-05,
      "loss": 0.1743,
      "step": 1884
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.630289955842311e-05,
      "loss": 0.1361,
      "step": 1885
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.629779780902008e-05,
      "loss": 0.1619,
      "step": 1886
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.629269334149284e-05,
      "loss": 0.15,
      "step": 1887
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6287586158044467e-05,
      "loss": 0.1377,
      "step": 1888
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6282476260879215e-05,
      "loss": 0.1822,
      "step": 1889
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.627736365220252e-05,
      "loss": 0.1678,
      "step": 1890
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.627224833422097e-05,
      "loss": 0.177,
      "step": 1891
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6267130309142337e-05,
      "loss": 0.1679,
      "step": 1892
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6262009579175555e-05,
      "loss": 0.1646,
      "step": 1893
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.625688614653073e-05,
      "loss": 0.1454,
      "step": 1894
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6251760013419123e-05,
      "loss": 0.1686,
      "step": 1895
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.6246631182053177e-05,
      "loss": 0.1527,
      "step": 1896
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.624149965464649e-05,
      "loss": 0.1836,
      "step": 1897
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.623636543341382e-05,
      "loss": 0.1579,
      "step": 1898
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.62312285205711e-05,
      "loss": 0.1508,
      "step": 1899
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6226088918335406e-05,
      "loss": 0.1192,
      "step": 1900
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6220946628924997e-05,
      "loss": 0.1611,
      "step": 1901
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6215801654559276e-05,
      "loss": 0.1582,
      "step": 1902
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.621065399745881e-05,
      "loss": 0.1298,
      "step": 1903
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6205503659845328e-05,
      "loss": 0.1418,
      "step": 1904
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6200350643941702e-05,
      "loss": 0.1889,
      "step": 1905
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.619519495197198e-05,
      "loss": 0.1872,
      "step": 1906
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6190036586161342e-05,
      "loss": 0.1794,
      "step": 1907
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6184875548736142e-05,
      "loss": 0.1543,
      "step": 1908
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6179711841923885e-05,
      "loss": 0.1556,
      "step": 1909
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.617454546795321e-05,
      "loss": 0.1375,
      "step": 1910
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6169376429053927e-05,
      "loss": 0.1622,
      "step": 1911
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.616420472745699e-05,
      "loss": 0.1465,
      "step": 1912
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6159030365394493e-05,
      "loss": 0.158,
      "step": 1913
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6153853345099694e-05,
      "loss": 0.1503,
      "step": 1914
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6148673668806993e-05,
      "loss": 0.1676,
      "step": 1915
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6143491338751922e-05,
      "loss": 0.1524,
      "step": 1916
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6138306357171182e-05,
      "loss": 0.1963,
      "step": 1917
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6133118726302603e-05,
      "loss": 0.1531,
      "step": 1918
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.612792844838516e-05,
      "loss": 0.1346,
      "step": 1919
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6122735525658974e-05,
      "loss": 0.1559,
      "step": 1920
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6117539960365305e-05,
      "loss": 0.1902,
      "step": 1921
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6112341754746553e-05,
      "loss": 0.1398,
      "step": 1922
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6107140911046266e-05,
      "loss": 0.1584,
      "step": 1923
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.6101937431509118e-05,
      "loss": 0.161,
      "step": 1924
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.609673131838093e-05,
      "loss": 0.1296,
      "step": 1925
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.609152257390865e-05,
      "loss": 0.1797,
      "step": 1926
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6086311200340374e-05,
      "loss": 0.1562,
      "step": 1927
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6081097199925325e-05,
      "loss": 0.2047,
      "step": 1928
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6075880574913857e-05,
      "loss": 0.1743,
      "step": 1929
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6070661327557465e-05,
      "loss": 0.1568,
      "step": 1930
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.606543946010877e-05,
      "loss": 0.1622,
      "step": 1931
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6060214974821532e-05,
      "loss": 0.1418,
      "step": 1932
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6054987873950623e-05,
      "loss": 0.2009,
      "step": 1933
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6049758159752062e-05,
      "loss": 0.154,
      "step": 1934
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6044525834482992e-05,
      "loss": 0.169,
      "step": 1935
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6039290900401674e-05,
      "loss": 0.1935,
      "step": 1936
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6034053359767504e-05,
      "loss": 0.195,
      "step": 1937
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6028813214841e-05,
      "loss": 0.1733,
      "step": 1938
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6023570467883803e-05,
      "loss": 0.2171,
      "step": 1939
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.601832512115868e-05,
      "loss": 0.1479,
      "step": 1940
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6013077176929517e-05,
      "loss": 0.1504,
      "step": 1941
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6007826637461323e-05,
      "loss": 0.1793,
      "step": 1942
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.6002573505020228e-05,
      "loss": 0.1886,
      "step": 1943
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.5997317781873474e-05,
      "loss": 0.1544,
      "step": 1944
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.599205947028943e-05,
      "loss": 0.1529,
      "step": 1945
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.5986798572537584e-05,
      "loss": 0.1351,
      "step": 1946
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.5981535090888524e-05,
      "loss": 0.1698,
      "step": 1947
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.5976269027613975e-05,
      "loss": 0.1447,
      "step": 1948
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.597100038498676e-05,
      "loss": 0.1542,
      "step": 1949
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.5965729165280822e-05,
      "loss": 0.1522,
      "step": 1950
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.596045537077121e-05,
      "loss": 0.1557,
      "step": 1951
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.59551790037341e-05,
      "loss": 0.161,
      "step": 1952
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.5949900066446758e-05,
      "loss": 0.1659,
      "step": 1953
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5944618561187573e-05,
      "loss": 0.1632,
      "step": 1954
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5939334490236037e-05,
      "loss": 0.1319,
      "step": 1955
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.593404785587275e-05,
      "loss": 0.1506,
      "step": 1956
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.592875866037942e-05,
      "loss": 0.1921,
      "step": 1957
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.592346690603886e-05,
      "loss": 0.1445,
      "step": 1958
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5918172595134988e-05,
      "loss": 0.1386,
      "step": 1959
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5912875729952817e-05,
      "loss": 0.1858,
      "step": 1960
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.590757631277848e-05,
      "loss": 0.1629,
      "step": 1961
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.590227434589919e-05,
      "loss": 0.1544,
      "step": 1962
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.589696983160328e-05,
      "loss": 0.1682,
      "step": 1963
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5891662772180173e-05,
      "loss": 0.1658,
      "step": 1964
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5886353169920385e-05,
      "loss": 0.1467,
      "step": 1965
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.588104102711554e-05,
      "loss": 0.1705,
      "step": 1966
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.587572634605836e-05,
      "loss": 0.161,
      "step": 1967
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5870409129042645e-05,
      "loss": 0.154,
      "step": 1968
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.586508937836331e-05,
      "loss": 0.174,
      "step": 1969
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5859767096316354e-05,
      "loss": 0.1568,
      "step": 1970
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5854442285198867e-05,
      "loss": 0.1522,
      "step": 1971
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5849114947309033e-05,
      "loss": 0.1597,
      "step": 1972
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5843785084946127e-05,
      "loss": 0.1502,
      "step": 1973
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5838452700410517e-05,
      "loss": 0.1651,
      "step": 1974
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.583311779600365e-05,
      "loss": 0.1591,
      "step": 1975
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5827780374028065e-05,
      "loss": 0.1595,
      "step": 1976
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.58224404367874e-05,
      "loss": 0.1869,
      "step": 1977
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.5817097986586355e-05,
      "loss": 0.1357,
      "step": 1978
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.581175302573073e-05,
      "loss": 0.1842,
      "step": 1979
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.580640555652741e-05,
      "loss": 0.1781,
      "step": 1980
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5801055581284353e-05,
      "loss": 0.141,
      "step": 1981
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5795703102310613e-05,
      "loss": 0.1412,
      "step": 1982
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.57903481219163e-05,
      "loss": 0.1423,
      "step": 1983
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5784990642412635e-05,
      "loss": 0.1855,
      "step": 1984
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5779630666111893e-05,
      "loss": 0.1464,
      "step": 1985
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5774268195327432e-05,
      "loss": 0.1727,
      "step": 1986
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5768903232373696e-05,
      "loss": 0.1528,
      "step": 1987
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.57635357795662e-05,
      "loss": 0.1527,
      "step": 1988
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5758165839221523e-05,
      "loss": 0.1332,
      "step": 1989
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5752793413657336e-05,
      "loss": 0.1712,
      "step": 1990
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5747418505192375e-05,
      "loss": 0.1643,
      "step": 1991
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5742041116146433e-05,
      "loss": 0.1609,
      "step": 1992
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.57366612488404e-05,
      "loss": 0.1358,
      "step": 1993
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5731278905596215e-05,
      "loss": 0.1699,
      "step": 1994
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5725894088736894e-05,
      "loss": 0.175,
      "step": 1995
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5720506800586522e-05,
      "loss": 0.1813,
      "step": 1996
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5715117043470244e-05,
      "loss": 0.1405,
      "step": 1997
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5709724819714277e-05,
      "loss": 0.1783,
      "step": 1998
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5704330131645898e-05,
      "loss": 0.1482,
      "step": 1999
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5698932981593456e-05,
      "loss": 0.1529,
      "step": 2000
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5693533371886348e-05,
      "loss": 0.1375,
      "step": 2001
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5688131304855047e-05,
      "loss": 0.1671,
      "step": 2002
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5682726782831078e-05,
      "loss": 0.1589,
      "step": 2003
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5677319808147032e-05,
      "loss": 0.1665,
      "step": 2004
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.567191038313655e-05,
      "loss": 0.1545,
      "step": 2005
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5666498510134335e-05,
      "loss": 0.1882,
      "step": 2006
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.566108419147615e-05,
      "loss": 0.1728,
      "step": 2007
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5655667429498803e-05,
      "loss": 0.1431,
      "step": 2008
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5650248226540177e-05,
      "loss": 0.1577,
      "step": 2009
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5644826584939185e-05,
      "loss": 0.1656,
      "step": 2010
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.563940250703581e-05,
      "loss": 0.1561,
      "step": 2011
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.563397599517107e-05,
      "loss": 0.1464,
      "step": 2012
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.562854705168705e-05,
      "loss": 0.1755,
      "step": 2013
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.562311567892688e-05,
      "loss": 0.1685,
      "step": 2014
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5617681879234727e-05,
      "loss": 0.1701,
      "step": 2015
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.561224565495582e-05,
      "loss": 0.1588,
      "step": 2016
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5606807008436435e-05,
      "loss": 0.1573,
      "step": 2017
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.560136594202387e-05,
      "loss": 0.1852,
      "step": 2018
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.55959224580665e-05,
      "loss": 0.1642,
      "step": 2019
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.559047655891372e-05,
      "loss": 0.1836,
      "step": 2020
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5585028246915987e-05,
      "loss": 0.1841,
      "step": 2021
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.557957752442477e-05,
      "loss": 0.17,
      "step": 2022
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.557412439379261e-05,
      "loss": 0.1673,
      "step": 2023
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.556866885737307e-05,
      "loss": 0.1731,
      "step": 2024
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.556321091752075e-05,
      "loss": 0.1657,
      "step": 2025
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5557750576591297e-05,
      "loss": 0.1693,
      "step": 2026
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.555228783694139e-05,
      "loss": 0.1293,
      "step": 2027
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5546822700928744e-05,
      "loss": 0.1853,
      "step": 2028
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5541355170912103e-05,
      "loss": 0.1752,
      "step": 2029
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.553588524925125e-05,
      "loss": 0.1523,
      "step": 2030
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.5530412938306997e-05,
      "loss": 0.168,
      "step": 2031
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.552493824044119e-05,
      "loss": 0.1458,
      "step": 2032
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5519461158016706e-05,
      "loss": 0.1593,
      "step": 2033
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.551398169339745e-05,
      "loss": 0.1457,
      "step": 2034
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.550849984894835e-05,
      "loss": 0.1403,
      "step": 2035
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5503015627035365e-05,
      "loss": 0.1641,
      "step": 2036
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5497529030025486e-05,
      "loss": 0.1444,
      "step": 2037
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.549204006028672e-05,
      "loss": 0.153,
      "step": 2038
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5486548720188103e-05,
      "loss": 0.1905,
      "step": 2039
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5481055012099686e-05,
      "loss": 0.1928,
      "step": 2040
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.547555893839256e-05,
      "loss": 0.1601,
      "step": 2041
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.547006050143882e-05,
      "loss": 0.1464,
      "step": 2042
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5464559703611586e-05,
      "loss": 0.1376,
      "step": 2043
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5459056547284998e-05,
      "loss": 0.1562,
      "step": 2044
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5453551034834216e-05,
      "loss": 0.1399,
      "step": 2045
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5448043168635413e-05,
      "loss": 0.1641,
      "step": 2046
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.544253295106578e-05,
      "loss": 0.1631,
      "step": 2047
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.543702038450352e-05,
      "loss": 0.1724,
      "step": 2048
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5431505471327858e-05,
      "loss": 0.1403,
      "step": 2049
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5425988213919016e-05,
      "loss": 0.1576,
      "step": 2050
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5420468614658245e-05,
      "loss": 0.1946,
      "step": 2051
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.54149466759278e-05,
      "loss": 0.1786,
      "step": 2052
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5409422400110943e-05,
      "loss": 0.1515,
      "step": 2053
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5403895789591943e-05,
      "loss": 0.1678,
      "step": 2054
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5398366846756085e-05,
      "loss": 0.1719,
      "step": 2055
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5392835573989656e-05,
      "loss": 0.1695,
      "step": 2056
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.538730197367995e-05,
      "loss": 0.1688,
      "step": 2057
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.538176604821526e-05,
      "loss": 0.141,
      "step": 2058
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.5376227799984886e-05,
      "loss": 0.15,
      "step": 2059
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5370687231379135e-05,
      "loss": 0.138,
      "step": 2060
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5365144344789315e-05,
      "loss": 0.1635,
      "step": 2061
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5359599142607722e-05,
      "loss": 0.1494,
      "step": 2062
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5354051627227666e-05,
      "loss": 0.1727,
      "step": 2063
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.534850180104345e-05,
      "loss": 0.1963,
      "step": 2064
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5342949666450374e-05,
      "loss": 0.1854,
      "step": 2065
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5337395225844733e-05,
      "loss": 0.1538,
      "step": 2066
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.533183848162382e-05,
      "loss": 0.1627,
      "step": 2067
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5326279436185923e-05,
      "loss": 0.1589,
      "step": 2068
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.532071809193032e-05,
      "loss": 0.1353,
      "step": 2069
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5315154451257276e-05,
      "loss": 0.1835,
      "step": 2070
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5309588516568068e-05,
      "loss": 0.1679,
      "step": 2071
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.530402029026494e-05,
      "loss": 0.1633,
      "step": 2072
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5298449774751126e-05,
      "loss": 0.1696,
      "step": 2073
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.529287697243087e-05,
      "loss": 0.1466,
      "step": 2074
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5287301885709384e-05,
      "loss": 0.1403,
      "step": 2075
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5281724516992863e-05,
      "loss": 0.1446,
      "step": 2076
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5276144868688505e-05,
      "loss": 0.1617,
      "step": 2077
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.527056294320448e-05,
      "loss": 0.1773,
      "step": 2078
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5264978742949935e-05,
      "loss": 0.1833,
      "step": 2079
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5259392270335017e-05,
      "loss": 0.1486,
      "step": 2080
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5253803527770831e-05,
      "loss": 0.1504,
      "step": 2081
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.524821251766948e-05,
      "loss": 0.1657,
      "step": 2082
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5242619242444044e-05,
      "loss": 0.1533,
      "step": 2083
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5237023704508564e-05,
      "loss": 0.1833,
      "step": 2084
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.5231425906278079e-05,
      "loss": 0.1619,
      "step": 2085
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.522582585016859e-05,
      "loss": 0.1775,
      "step": 2086
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5220223538597079e-05,
      "loss": 0.1597,
      "step": 2087
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5214618973981496e-05,
      "loss": 0.1786,
      "step": 2088
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5209012158740768e-05,
      "loss": 0.1927,
      "step": 2089
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5203403095294794e-05,
      "loss": 0.1493,
      "step": 2090
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5197791786064433e-05,
      "loss": 0.1718,
      "step": 2091
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.519217823347153e-05,
      "loss": 0.1521,
      "step": 2092
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5186562439938888e-05,
      "loss": 0.1615,
      "step": 2093
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5180944407890275e-05,
      "loss": 0.162,
      "step": 2094
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5175324139750432e-05,
      "loss": 0.1471,
      "step": 2095
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5169701637945057e-05,
      "loss": 0.148,
      "step": 2096
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5164076904900826e-05,
      "loss": 0.167,
      "step": 2097
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5158449943045361e-05,
      "loss": 0.1513,
      "step": 2098
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5152820754807258e-05,
      "loss": 0.1501,
      "step": 2099
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5147189342616072e-05,
      "loss": 0.1669,
      "step": 2100
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5141555708902308e-05,
      "loss": 0.1764,
      "step": 2101
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5135919856097445e-05,
      "loss": 0.1548,
      "step": 2102
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.513028178663391e-05,
      "loss": 0.1729,
      "step": 2103
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.512464150294509e-05,
      "loss": 0.1633,
      "step": 2104
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.511899900746532e-05,
      "loss": 0.1833,
      "step": 2105
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5113354302629904e-05,
      "loss": 0.1458,
      "step": 2106
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.510770739087509e-05,
      "loss": 0.1531,
      "step": 2107
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.510205827463808e-05,
      "loss": 0.1779,
      "step": 2108
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5096406956357026e-05,
      "loss": 0.1521,
      "step": 2109
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.5090753438471034e-05,
      "loss": 0.1512,
      "step": 2110
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.508509772342016e-05,
      "loss": 0.1451,
      "step": 2111
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.50794398136454e-05,
      "loss": 0.1417,
      "step": 2112
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5073779711588705e-05,
      "loss": 0.1487,
      "step": 2113
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5068117419692976e-05,
      "loss": 0.1614,
      "step": 2114
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5062452940402044e-05,
      "loss": 0.142,
      "step": 2115
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5056786276160703e-05,
      "loss": 0.1803,
      "step": 2116
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5051117429414673e-05,
      "loss": 0.1755,
      "step": 2117
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5045446402610626e-05,
      "loss": 0.1389,
      "step": 2118
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5039773198196174e-05,
      "loss": 0.1677,
      "step": 2119
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5034097818619863e-05,
      "loss": 0.1435,
      "step": 2120
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5028420266331187e-05,
      "loss": 0.1509,
      "step": 2121
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5022740543780572e-05,
      "loss": 0.1861,
      "step": 2122
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.501705865341938e-05,
      "loss": 0.1548,
      "step": 2123
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5011374597699911e-05,
      "loss": 0.1377,
      "step": 2124
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5005688379075394e-05,
      "loss": 0.1586,
      "step": 2125
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 0.1301,
      "step": 2126
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4994309462928833e-05,
      "loss": 0.1436,
      "step": 2127
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4988616770317915e-05,
      "loss": 0.146,
      "step": 2128
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4982921924624214e-05,
      "loss": 0.1548,
      "step": 2129
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4977224928305623e-05,
      "loss": 0.1761,
      "step": 2130
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4971525783820951e-05,
      "loss": 0.1708,
      "step": 2131
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.496582449362995e-05,
      "loss": 0.1492,
      "step": 2132
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4960121060193294e-05,
      "loss": 0.1595,
      "step": 2133
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4954415485972573e-05,
      "loss": 0.1353,
      "step": 2134
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4948707773430313e-05,
      "loss": 0.1604,
      "step": 2135
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4942997925029961e-05,
      "loss": 0.1524,
      "step": 2136
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4937285943235877e-05,
      "loss": 0.163,
      "step": 2137
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.493157183051335e-05,
      "loss": 0.1409,
      "step": 2138
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.4925855589328586e-05,
      "loss": 0.1731,
      "step": 2139
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.492013722214871e-05,
      "loss": 0.135,
      "step": 2140
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.491441673144176e-05,
      "loss": 0.1696,
      "step": 2141
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4908694119676703e-05,
      "loss": 0.1586,
      "step": 2142
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.490296938932341e-05,
      "loss": 0.1527,
      "step": 2143
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4897242542852667e-05,
      "loss": 0.155,
      "step": 2144
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4891513582736181e-05,
      "loss": 0.159,
      "step": 2145
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4885782511446562e-05,
      "loss": 0.1424,
      "step": 2146
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4880049331457339e-05,
      "loss": 0.1137,
      "step": 2147
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4874314045242941e-05,
      "loss": 0.1656,
      "step": 2148
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4868576655278723e-05,
      "loss": 0.1354,
      "step": 2149
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4862837164040928e-05,
      "loss": 0.1607,
      "step": 2150
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4857095574006722e-05,
      "loss": 0.1838,
      "step": 2151
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4851351887654168e-05,
      "loss": 0.1699,
      "step": 2152
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4845606107462239e-05,
      "loss": 0.1581,
      "step": 2153
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4839858235910802e-05,
      "loss": 0.159,
      "step": 2154
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4834108275480642e-05,
      "loss": 0.1689,
      "step": 2155
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.482835622865343e-05,
      "loss": 0.1443,
      "step": 2156
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4822602097911747e-05,
      "loss": 0.1716,
      "step": 2157
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4816845885739071e-05,
      "loss": 0.1624,
      "step": 2158
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4811087594619783e-05,
      "loss": 0.1768,
      "step": 2159
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4805327227039147e-05,
      "loss": 0.1428,
      "step": 2160
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.479956478548334e-05,
      "loss": 0.1515,
      "step": 2161
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4793800272439427e-05,
      "loss": 0.1452,
      "step": 2162
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4788033690395363e-05,
      "loss": 0.16,
      "step": 2163
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4782265041839999e-05,
      "loss": 0.1699,
      "step": 2164
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.4776494329263084e-05,
      "loss": 0.1686,
      "step": 2165
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4770721555155251e-05,
      "loss": 0.1528,
      "step": 2166
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4764946722008024e-05,
      "loss": 0.1641,
      "step": 2167
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4759169832313818e-05,
      "loss": 0.139,
      "step": 2168
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4753390888565928e-05,
      "loss": 0.1289,
      "step": 2169
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4747609893258546e-05,
      "loss": 0.156,
      "step": 2170
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4741826848886746e-05,
      "loss": 0.1486,
      "step": 2171
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4736041757946477e-05,
      "loss": 0.1543,
      "step": 2172
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.473025462293459e-05,
      "loss": 0.1671,
      "step": 2173
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4724465446348801e-05,
      "loss": 0.142,
      "step": 2174
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4718674230687718e-05,
      "loss": 0.1584,
      "step": 2175
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4712880978450818e-05,
      "loss": 0.1461,
      "step": 2176
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4707085692138469e-05,
      "loss": 0.1866,
      "step": 2177
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4701288374251912e-05,
      "loss": 0.1236,
      "step": 2178
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.469548902729326e-05,
      "loss": 0.1819,
      "step": 2179
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.468968765376551e-05,
      "loss": 0.167,
      "step": 2180
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4683884256172531e-05,
      "loss": 0.1838,
      "step": 2181
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4678078837019064e-05,
      "loss": 0.1568,
      "step": 2182
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4672271398810717e-05,
      "loss": 0.1402,
      "step": 2183
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4666461944053984e-05,
      "loss": 0.1343,
      "step": 2184
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4660650475256217e-05,
      "loss": 0.1411,
      "step": 2185
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.465483699492564e-05,
      "loss": 0.1442,
      "step": 2186
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4649021505571344e-05,
      "loss": 0.2063,
      "step": 2187
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4643204009703298e-05,
      "loss": 0.1858,
      "step": 2188
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.463738450983232e-05,
      "loss": 0.17,
      "step": 2189
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4631563008470102e-05,
      "loss": 0.1464,
      "step": 2190
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4625739508129205e-05,
      "loss": 0.1774,
      "step": 2191
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.4619914011323042e-05,
      "loss": 0.1516,
      "step": 2192
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4614086520565891e-05,
      "loss": 0.1324,
      "step": 2193
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4608257038372894e-05,
      "loss": 0.1484,
      "step": 2194
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4602425567260055e-05,
      "loss": 0.1928,
      "step": 2195
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4596592109744225e-05,
      "loss": 0.1704,
      "step": 2196
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4590756668343123e-05,
      "loss": 0.1737,
      "step": 2197
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4584919245575323e-05,
      "loss": 0.1607,
      "step": 2198
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4579079843960244e-05,
      "loss": 0.1606,
      "step": 2199
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4573238466018173e-05,
      "loss": 0.1814,
      "step": 2200
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4567395114270242e-05,
      "loss": 0.1541,
      "step": 2201
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4561549791238438e-05,
      "loss": 0.142,
      "step": 2202
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4555702499445597e-05,
      "loss": 0.1563,
      "step": 2203
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4549853241415404e-05,
      "loss": 0.1476,
      "step": 2204
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4544002019672395e-05,
      "loss": 0.1441,
      "step": 2205
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.453814883674195e-05,
      "loss": 0.1476,
      "step": 2206
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4532293695150301e-05,
      "loss": 0.1606,
      "step": 2207
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.452643659742452e-05,
      "loss": 0.1687,
      "step": 2208
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4520577546092529e-05,
      "loss": 0.1615,
      "step": 2209
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4514716543683083e-05,
      "loss": 0.1473,
      "step": 2210
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4508853592725792e-05,
      "loss": 0.1377,
      "step": 2211
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.45029886957511e-05,
      "loss": 0.1665,
      "step": 2212
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4497121855290283e-05,
      "loss": 0.1599,
      "step": 2213
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4491253073875477e-05,
      "loss": 0.1872,
      "step": 2214
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4485382354039637e-05,
      "loss": 0.1609,
      "step": 2215
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4479509698316556e-05,
      "loss": 0.1479,
      "step": 2216
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4473635109240878e-05,
      "loss": 0.1528,
      "step": 2217
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4467758589348065e-05,
      "loss": 0.1574,
      "step": 2218
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4461880141174414e-05,
      "loss": 0.1441,
      "step": 2219
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4455999767257068e-05,
      "loss": 0.1914,
      "step": 2220
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.445011747013398e-05,
      "loss": 0.1495,
      "step": 2221
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4444233252343951e-05,
      "loss": 0.1518,
      "step": 2222
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4438347116426605e-05,
      "loss": 0.1776,
      "step": 2223
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4432459064922393e-05,
      "loss": 0.1753,
      "step": 2224
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4426569100372591e-05,
      "loss": 0.166,
      "step": 2225
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4420677225319301e-05,
      "loss": 0.179,
      "step": 2226
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4414783442305458e-05,
      "loss": 0.1564,
      "step": 2227
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.440888775387481e-05,
      "loss": 0.144,
      "step": 2228
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4402990162571926e-05,
      "loss": 0.1675,
      "step": 2229
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4397090670942208e-05,
      "loss": 0.1469,
      "step": 2230
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4391189281531871e-05,
      "loss": 0.14,
      "step": 2231
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.438528599688795e-05,
      "loss": 0.1522,
      "step": 2232
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4379380819558295e-05,
      "loss": 0.1674,
      "step": 2233
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4373473752091575e-05,
      "loss": 0.1531,
      "step": 2234
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.436756479703728e-05,
      "loss": 0.1623,
      "step": 2235
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4361653956945708e-05,
      "loss": 0.1468,
      "step": 2236
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4355741234367968e-05,
      "loss": 0.1426,
      "step": 2237
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4349826631855993e-05,
      "loss": 0.1355,
      "step": 2238
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4343910151962512e-05,
      "loss": 0.1401,
      "step": 2239
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4337991797241082e-05,
      "loss": 0.1458,
      "step": 2240
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4332071570246058e-05,
      "loss": 0.1526,
      "step": 2241
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4326149473532598e-05,
      "loss": 0.1667,
      "step": 2242
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4320225509656682e-05,
      "loss": 0.1446,
      "step": 2243
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4314299681175085e-05,
      "loss": 0.1765,
      "step": 2244
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.4308371990645387e-05,
      "loss": 0.1395,
      "step": 2245
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4302442440625974e-05,
      "loss": 0.1819,
      "step": 2246
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4296511033676042e-05,
      "loss": 0.1578,
      "step": 2247
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4290577772355575e-05,
      "loss": 0.1507,
      "step": 2248
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4284642659225365e-05,
      "loss": 0.173,
      "step": 2249
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4278705696847002e-05,
      "loss": 0.1492,
      "step": 2250
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4272766887782873e-05,
      "loss": 0.1317,
      "step": 2251
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4266826234596163e-05,
      "loss": 0.154,
      "step": 2252
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4260883739850854e-05,
      "loss": 0.1442,
      "step": 2253
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4254939406111723e-05,
      "loss": 0.1729,
      "step": 2254
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4248993235944335e-05,
      "loss": 0.1625,
      "step": 2255
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4243045231915058e-05,
      "loss": 0.1734,
      "step": 2256
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.423709539659104e-05,
      "loss": 0.1445,
      "step": 2257
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.423114373254023e-05,
      "loss": 0.1378,
      "step": 2258
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4225190242331356e-05,
      "loss": 0.1626,
      "step": 2259
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.421923492853394e-05,
      "loss": 0.1443,
      "step": 2260
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4213277793718293e-05,
      "loss": 0.1635,
      "step": 2261
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.420731884045551e-05,
      "loss": 0.1152,
      "step": 2262
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.420135807131747e-05,
      "loss": 0.149,
      "step": 2263
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4195395488876833e-05,
      "loss": 0.1263,
      "step": 2264
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4189431095707045e-05,
      "loss": 0.1471,
      "step": 2265
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4183464894382335e-05,
      "loss": 0.1637,
      "step": 2266
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4177496887477708e-05,
      "loss": 0.1875,
      "step": 2267
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4171527077568953e-05,
      "loss": 0.1467,
      "step": 2268
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4165555467232637e-05,
      "loss": 0.1827,
      "step": 2269
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4159582059046097e-05,
      "loss": 0.1269,
      "step": 2270
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.4153606855587454e-05,
      "loss": 0.1765,
      "step": 2271
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4147629859435601e-05,
      "loss": 0.139,
      "step": 2272
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4141651073170204e-05,
      "loss": 0.1897,
      "step": 2273
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4135670499371698e-05,
      "loss": 0.1663,
      "step": 2274
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4129688140621298e-05,
      "loss": 0.1721,
      "step": 2275
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4123703999500987e-05,
      "loss": 0.1492,
      "step": 2276
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.411771807859351e-05,
      "loss": 0.1551,
      "step": 2277
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4111730380482387e-05,
      "loss": 0.162,
      "step": 2278
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4105740907751907e-05,
      "loss": 0.147,
      "step": 2279
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4099749662987118e-05,
      "loss": 0.1387,
      "step": 2280
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4093756648773836e-05,
      "loss": 0.1621,
      "step": 2281
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.408776186769864e-05,
      "loss": 0.1636,
      "step": 2282
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4081765322348877e-05,
      "loss": 0.1514,
      "step": 2283
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4075767015312644e-05,
      "loss": 0.122,
      "step": 2284
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4069766949178814e-05,
      "loss": 0.1622,
      "step": 2285
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4063765126537005e-05,
      "loss": 0.1451,
      "step": 2286
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4057761549977598e-05,
      "loss": 0.1476,
      "step": 2287
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4051756222091735e-05,
      "loss": 0.1502,
      "step": 2288
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4045749145471304e-05,
      "loss": 0.1411,
      "step": 2289
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4039740322708961e-05,
      "loss": 0.16,
      "step": 2290
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4033729756398106e-05,
      "loss": 0.1433,
      "step": 2291
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4027717449132897e-05,
      "loss": 0.1914,
      "step": 2292
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4021703403508237e-05,
      "loss": 0.1394,
      "step": 2293
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4015687622119785e-05,
      "loss": 0.18,
      "step": 2294
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4009670107563947e-05,
      "loss": 0.165,
      "step": 2295
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.4003650862437874e-05,
      "loss": 0.1814,
      "step": 2296
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.3997629889339472e-05,
      "loss": 0.1573,
      "step": 2297
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.3991607190867385e-05,
      "loss": 0.1459,
      "step": 2298
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3985582769621006e-05,
      "loss": 0.1625,
      "step": 2299
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.397955662820047e-05,
      "loss": 0.1774,
      "step": 2300
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3973528769206655e-05,
      "loss": 0.1584,
      "step": 2301
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3967499195241176e-05,
      "loss": 0.1677,
      "step": 2302
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3961467908906397e-05,
      "loss": 0.1839,
      "step": 2303
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3955434912805413e-05,
      "loss": 0.1362,
      "step": 2304
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3949400209542064e-05,
      "loss": 0.1673,
      "step": 2305
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3943363801720918e-05,
      "loss": 0.1429,
      "step": 2306
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3937325691947284e-05,
      "loss": 0.1508,
      "step": 2307
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3931285882827208e-05,
      "loss": 0.1542,
      "step": 2308
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3925244376967467e-05,
      "loss": 0.1441,
      "step": 2309
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3919201176975566e-05,
      "loss": 0.1471,
      "step": 2310
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3913156285459743e-05,
      "loss": 0.1459,
      "step": 2311
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3907109705028979e-05,
      "loss": 0.1508,
      "step": 2312
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3901061438292961e-05,
      "loss": 0.1402,
      "step": 2313
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3895011487862123e-05,
      "loss": 0.1284,
      "step": 2314
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3888959856347618e-05,
      "loss": 0.1545,
      "step": 2315
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3882906546361322e-05,
      "loss": 0.1641,
      "step": 2316
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3876851560515837e-05,
      "loss": 0.1354,
      "step": 2317
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3870794901424493e-05,
      "loss": 0.1783,
      "step": 2318
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.386473657170134e-05,
      "loss": 0.1629,
      "step": 2319
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3858676573961147e-05,
      "loss": 0.1603,
      "step": 2320
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3852614910819404e-05,
      "loss": 0.1404,
      "step": 2321
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3846551584892318e-05,
      "loss": 0.1533,
      "step": 2322
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3840486598796816e-05,
      "loss": 0.1796,
      "step": 2323
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3834419955150545e-05,
      "loss": 0.1574,
      "step": 2324
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.3828351656571855e-05,
      "loss": 0.1624,
      "step": 2325
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3822281705679826e-05,
      "loss": 0.1699,
      "step": 2326
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3816210105094236e-05,
      "loss": 0.182,
      "step": 2327
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3810136857435591e-05,
      "loss": 0.1589,
      "step": 2328
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.38040619653251e-05,
      "loss": 0.1568,
      "step": 2329
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3797985431384672e-05,
      "loss": 0.1481,
      "step": 2330
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.379190725823694e-05,
      "loss": 0.1453,
      "step": 2331
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3785827448505243e-05,
      "loss": 0.1743,
      "step": 2332
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3779746004813607e-05,
      "loss": 0.1556,
      "step": 2333
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3773662929786795e-05,
      "loss": 0.1642,
      "step": 2334
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3767578226050247e-05,
      "loss": 0.1667,
      "step": 2335
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3761491896230119e-05,
      "loss": 0.1569,
      "step": 2336
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3755403942953264e-05,
      "loss": 0.1471,
      "step": 2337
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3749314368847242e-05,
      "loss": 0.1598,
      "step": 2338
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3743223176540301e-05,
      "loss": 0.1594,
      "step": 2339
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3737130368661398e-05,
      "loss": 0.1546,
      "step": 2340
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3731035947840188e-05,
      "loss": 0.1671,
      "step": 2341
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3724939916707013e-05,
      "loss": 0.1375,
      "step": 2342
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3718842277892912e-05,
      "loss": 0.1689,
      "step": 2343
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3712743034029628e-05,
      "loss": 0.1624,
      "step": 2344
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3706642187749586e-05,
      "loss": 0.1729,
      "step": 2345
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3700539741685903e-05,
      "loss": 0.1204,
      "step": 2346
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3694435698472397e-05,
      "loss": 0.1688,
      "step": 2347
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.368833006074356e-05,
      "loss": 0.1356,
      "step": 2348
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3682222831134585e-05,
      "loss": 0.1469,
      "step": 2349
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3676114012281345e-05,
      "loss": 0.163,
      "step": 2350
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.3670003606820403e-05,
      "loss": 0.1345,
      "step": 2351
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3663891617389005e-05,
      "loss": 0.13,
      "step": 2352
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3657778046625076e-05,
      "loss": 0.1478,
      "step": 2353
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3651662897167234e-05,
      "loss": 0.1487,
      "step": 2354
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3645546171654768e-05,
      "loss": 0.1411,
      "step": 2355
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3639427872727654e-05,
      "loss": 0.1475,
      "step": 2356
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3633308003026544e-05,
      "loss": 0.2076,
      "step": 2357
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3627186565192767e-05,
      "loss": 0.1644,
      "step": 2358
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3621063561868331e-05,
      "loss": 0.1679,
      "step": 2359
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3614938995695921e-05,
      "loss": 0.146,
      "step": 2360
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3608812869318894e-05,
      "loss": 0.1588,
      "step": 2361
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3602685185381282e-05,
      "loss": 0.1779,
      "step": 2362
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3596555946527783e-05,
      "loss": 0.1722,
      "step": 2363
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3590425155403777e-05,
      "loss": 0.1606,
      "step": 2364
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3584292814655304e-05,
      "loss": 0.1862,
      "step": 2365
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3578158926929081e-05,
      "loss": 0.1717,
      "step": 2366
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3572023494872487e-05,
      "loss": 0.1531,
      "step": 2367
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3565886521133572e-05,
      "loss": 0.1837,
      "step": 2368
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.355974800836104e-05,
      "loss": 0.174,
      "step": 2369
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3553607959204276e-05,
      "loss": 0.1754,
      "step": 2370
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3547466376313319e-05,
      "loss": 0.1706,
      "step": 2371
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3541323262338868e-05,
      "loss": 0.1507,
      "step": 2372
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.353517861993229e-05,
      "loss": 0.1591,
      "step": 2373
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3529032451745608e-05,
      "loss": 0.1604,
      "step": 2374
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3522884760431494e-05,
      "loss": 0.136,
      "step": 2375
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3516735548643296e-05,
      "loss": 0.1436,
      "step": 2376
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.3510584819035004e-05,
      "loss": 0.1334,
      "step": 2377
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.350443257426127e-05,
      "loss": 0.1911,
      "step": 2378
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3498278816977395e-05,
      "loss": 0.1516,
      "step": 2379
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3492123549839339e-05,
      "loss": 0.1619,
      "step": 2380
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3485966775503706e-05,
      "loss": 0.1807,
      "step": 2381
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3479808496627758e-05,
      "loss": 0.1831,
      "step": 2382
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3473648715869403e-05,
      "loss": 0.1548,
      "step": 2383
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3467487435887192e-05,
      "loss": 0.1909,
      "step": 2384
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3461324659340333e-05,
      "loss": 0.1578,
      "step": 2385
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3455160388888678e-05,
      "loss": 0.1564,
      "step": 2386
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3448994627192715e-05,
      "loss": 0.1482,
      "step": 2387
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3442827376913581e-05,
      "loss": 0.1643,
      "step": 2388
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3436658640713063e-05,
      "loss": 0.1846,
      "step": 2389
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3430488421253576e-05,
      "loss": 0.1299,
      "step": 2390
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3424316721198183e-05,
      "loss": 0.1804,
      "step": 2391
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3418143543210585e-05,
      "loss": 0.2098,
      "step": 2392
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3411968889955121e-05,
      "loss": 0.1519,
      "step": 2393
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3405792764096768e-05,
      "loss": 0.2004,
      "step": 2394
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.339961516830113e-05,
      "loss": 0.1625,
      "step": 2395
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3393436105234462e-05,
      "loss": 0.1819,
      "step": 2396
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3387255577563632e-05,
      "loss": 0.1707,
      "step": 2397
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3381073587956154e-05,
      "loss": 0.1412,
      "step": 2398
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3374890139080173e-05,
      "loss": 0.1359,
      "step": 2399
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.336870523360446e-05,
      "loss": 0.1597,
      "step": 2400
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3362518874198414e-05,
      "loss": 0.134,
      "step": 2401
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3356331063532058e-05,
      "loss": 0.1579,
      "step": 2402
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.335014180427605e-05,
      "loss": 0.1539,
      "step": 2403
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.3343951099101671e-05,
      "loss": 0.1416,
      "step": 2404
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.333775895068082e-05,
      "loss": 0.1526,
      "step": 2405
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3331565361686023e-05,
      "loss": 0.1589,
      "step": 2406
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3325370334790432e-05,
      "loss": 0.1439,
      "step": 2407
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3319173872667812e-05,
      "loss": 0.1362,
      "step": 2408
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3312975977992553e-05,
      "loss": 0.1319,
      "step": 2409
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.330677665343966e-05,
      "loss": 0.1626,
      "step": 2410
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3300575901684754e-05,
      "loss": 0.1835,
      "step": 2411
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3294373725404077e-05,
      "loss": 0.1941,
      "step": 2412
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3288170127274485e-05,
      "loss": 0.161,
      "step": 2413
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3281965109973439e-05,
      "loss": 0.2129,
      "step": 2414
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3275758676179022e-05,
      "loss": 0.1465,
      "step": 2415
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.326955082856993e-05,
      "loss": 0.161,
      "step": 2416
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3263341569825463e-05,
      "loss": 0.1357,
      "step": 2417
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3257130902625528e-05,
      "loss": 0.1529,
      "step": 2418
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3250918829650647e-05,
      "loss": 0.1442,
      "step": 2419
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3244705353581946e-05,
      "loss": 0.187,
      "step": 2420
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.323849047710115e-05,
      "loss": 0.1611,
      "step": 2421
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3232274202890602e-05,
      "loss": 0.1422,
      "step": 2422
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3226056533633239e-05,
      "loss": 0.1617,
      "step": 2423
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3219837472012598e-05,
      "loss": 0.1533,
      "step": 2424
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3213617020712825e-05,
      "loss": 0.157,
      "step": 2425
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.320739518241866e-05,
      "loss": 0.1435,
      "step": 2426
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3201171959815442e-05,
      "loss": 0.1842,
      "step": 2427
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.319494735558911e-05,
      "loss": 0.1734,
      "step": 2428
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3188721372426198e-05,
      "loss": 0.162,
      "step": 2429
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3182494013013836e-05,
      "loss": 0.1469,
      "step": 2430
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.3176265280039743e-05,
      "loss": 0.1969,
      "step": 2431
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.317003517619224e-05,
      "loss": 0.1599,
      "step": 2432
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3163803704160231e-05,
      "loss": 0.1666,
      "step": 2433
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3157570866633216e-05,
      "loss": 0.1556,
      "step": 2434
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3151336666301279e-05,
      "loss": 0.1616,
      "step": 2435
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.31451011058551e-05,
      "loss": 0.1511,
      "step": 2436
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3138864187985937e-05,
      "loss": 0.1264,
      "step": 2437
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3132625915385647e-05,
      "loss": 0.1443,
      "step": 2438
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3126386290746655e-05,
      "loss": 0.1682,
      "step": 2439
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3120145316761979e-05,
      "loss": 0.1439,
      "step": 2440
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.311390299612522e-05,
      "loss": 0.1745,
      "step": 2441
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.310765933153056e-05,
      "loss": 0.1707,
      "step": 2442
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3101414325672755e-05,
      "loss": 0.1507,
      "step": 2443
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3095167981247143e-05,
      "loss": 0.154,
      "step": 2444
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.308892030094965e-05,
      "loss": 0.1432,
      "step": 2445
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.308267128747676e-05,
      "loss": 0.1559,
      "step": 2446
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3076420943525547e-05,
      "loss": 0.1541,
      "step": 2447
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3070169271793653e-05,
      "loss": 0.1638,
      "step": 2448
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3063916274979296e-05,
      "loss": 0.1449,
      "step": 2449
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3057661955781258e-05,
      "loss": 0.1459,
      "step": 2450
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3051406316898905e-05,
      "loss": 0.1819,
      "step": 2451
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3045149361032161e-05,
      "loss": 0.159,
      "step": 2452
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3038891090881524e-05,
      "loss": 0.1533,
      "step": 2453
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3032631509148062e-05,
      "loss": 0.1386,
      "step": 2454
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3026370618533401e-05,
      "loss": 0.1589,
      "step": 2455
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3020108421739735e-05,
      "loss": 0.137,
      "step": 2456
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.3013844921469824e-05,
      "loss": 0.1584,
      "step": 2457
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3007580120426993e-05,
      "loss": 0.1276,
      "step": 2458
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3001314021315122e-05,
      "loss": 0.1291,
      "step": 2459
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.299504662683865e-05,
      "loss": 0.1427,
      "step": 2460
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2988777939702587e-05,
      "loss": 0.1605,
      "step": 2461
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.298250796261249e-05,
      "loss": 0.1335,
      "step": 2462
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2976236698274473e-05,
      "loss": 0.1638,
      "step": 2463
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2969964149395212e-05,
      "loss": 0.1895,
      "step": 2464
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.296369031868193e-05,
      "loss": 0.1409,
      "step": 2465
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2957415208842412e-05,
      "loss": 0.1634,
      "step": 2466
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2951138822584985e-05,
      "loss": 0.1558,
      "step": 2467
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2944861162618536e-05,
      "loss": 0.1731,
      "step": 2468
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2938582231652499e-05,
      "loss": 0.1431,
      "step": 2469
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2932302032396849e-05,
      "loss": 0.1575,
      "step": 2470
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2926020567562118e-05,
      "loss": 0.1482,
      "step": 2471
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2919737839859382e-05,
      "loss": 0.1414,
      "step": 2472
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.291345385200026e-05,
      "loss": 0.1365,
      "step": 2473
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2907168606696916e-05,
      "loss": 0.166,
      "step": 2474
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2900882106662058e-05,
      "loss": 0.1199,
      "step": 2475
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.289459435460893e-05,
      "loss": 0.143,
      "step": 2476
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2888305353251325e-05,
      "loss": 0.1369,
      "step": 2477
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2882015105303569e-05,
      "loss": 0.1561,
      "step": 2478
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2875723613480525e-05,
      "loss": 0.1459,
      "step": 2479
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2869430880497599e-05,
      "loss": 0.1851,
      "step": 2480
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2863136909070729e-05,
      "loss": 0.1641,
      "step": 2481
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2856841701916389e-05,
      "loss": 0.1458,
      "step": 2482
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.285054526175158e-05,
      "loss": 0.1618,
      "step": 2483
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2844247591293846e-05,
      "loss": 0.1629,
      "step": 2484
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2837948693261254e-05,
      "loss": 0.1522,
      "step": 2485
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2831648570372397e-05,
      "loss": 0.1472,
      "step": 2486
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2825347225346411e-05,
      "loss": 0.1536,
      "step": 2487
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2819044660902953e-05,
      "loss": 0.1729,
      "step": 2488
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2812740879762195e-05,
      "loss": 0.1252,
      "step": 2489
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2806435884644853e-05,
      "loss": 0.1575,
      "step": 2490
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.280012967827215e-05,
      "loss": 0.1928,
      "step": 2491
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2793822263365843e-05,
      "loss": 0.128,
      "step": 2492
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2787513642648208e-05,
      "loss": 0.1497,
      "step": 2493
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2781203818842038e-05,
      "loss": 0.1473,
      "step": 2494
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.277489279467065e-05,
      "loss": 0.1658,
      "step": 2495
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2768580572857874e-05,
      "loss": 0.1441,
      "step": 2496
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2762267156128063e-05,
      "loss": 0.1527,
      "step": 2497
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2755952547206079e-05,
      "loss": 0.1508,
      "step": 2498
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.27496367488173e-05,
      "loss": 0.1388,
      "step": 2499
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2743319763687627e-05,
      "loss": 0.1503,
      "step": 2500
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2737001594543459e-05,
      "loss": 0.1461,
      "step": 2501
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2730682244111715e-05,
      "loss": 0.1654,
      "step": 2502
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2724361715119815e-05,
      "loss": 0.1525,
      "step": 2503
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2718040010295706e-05,
      "loss": 0.1589,
      "step": 2504
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.271171713236782e-05,
      "loss": 0.1474,
      "step": 2505
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2705393084065114e-05,
      "loss": 0.1534,
      "step": 2506
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2699067868117037e-05,
      "loss": 0.1761,
      "step": 2507
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2692741487253546e-05,
      "loss": 0.1836,
      "step": 2508
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.26864139442051e-05,
      "loss": 0.1345,
      "step": 2509
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2680085241702669e-05,
      "loss": 0.1828,
      "step": 2510
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2673755382477713e-05,
      "loss": 0.1445,
      "step": 2511
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2667424369262186e-05,
      "loss": 0.1553,
      "step": 2512
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2661092204788558e-05,
      "loss": 0.1466,
      "step": 2513
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2654758891789782e-05,
      "loss": 0.1284,
      "step": 2514
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2648424432999309e-05,
      "loss": 0.1618,
      "step": 2515
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2642088831151086e-05,
      "loss": 0.1633,
      "step": 2516
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2635752088979556e-05,
      "loss": 0.1555,
      "step": 2517
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.262941420921965e-05,
      "loss": 0.1579,
      "step": 2518
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2623075194606793e-05,
      "loss": 0.1627,
      "step": 2519
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2616735047876897e-05,
      "loss": 0.1539,
      "step": 2520
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2610393771766365e-05,
      "loss": 0.1813,
      "step": 2521
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2604051369012083e-05,
      "loss": 0.1571,
      "step": 2522
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2597707842351432e-05,
      "loss": 0.1546,
      "step": 2523
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2591363194522268e-05,
      "loss": 0.1661,
      "step": 2524
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2585017428262938e-05,
      "loss": 0.1712,
      "step": 2525
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.257867054631227e-05,
      "loss": 0.1909,
      "step": 2526
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2572322551409572e-05,
      "loss": 0.147,
      "step": 2527
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2565973446294632e-05,
      "loss": 0.1831,
      "step": 2528
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2559623233707722e-05,
      "loss": 0.1459,
      "step": 2529
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2553271916389587e-05,
      "loss": 0.1498,
      "step": 2530
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2546919497081445e-05,
      "loss": 0.1566,
      "step": 2531
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2540565978525e-05,
      "loss": 0.1908,
      "step": 2532
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2534211363462425e-05,
      "loss": 0.1639,
      "step": 2533
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2527855654636367e-05,
      "loss": 0.1585,
      "step": 2534
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2521498854789943e-05,
      "loss": 0.1839,
      "step": 2535
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2515140966666742e-05,
      "loss": 0.1799,
      "step": 2536
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2508781993010823e-05,
      "loss": 0.1277,
      "step": 2537
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2502421936566713e-05,
      "loss": 0.1547,
      "step": 2538
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2496060800079413e-05,
      "loss": 0.1528,
      "step": 2539
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2489698586294376e-05,
      "loss": 0.1654,
      "step": 2540
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2483335297957533e-05,
      "loss": 0.1348,
      "step": 2541
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2476970937815274e-05,
      "loss": 0.1674,
      "step": 2542
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2470605508614451e-05,
      "loss": 0.1708,
      "step": 2543
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2464239013102373e-05,
      "loss": 0.169,
      "step": 2544
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2457871454026823e-05,
      "loss": 0.1434,
      "step": 2545
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2451502834136025e-05,
      "loss": 0.1569,
      "step": 2546
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2445133156178675e-05,
      "loss": 0.1333,
      "step": 2547
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2438762422903921e-05,
      "loss": 0.1689,
      "step": 2548
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2432390637061365e-05,
      "loss": 0.1385,
      "step": 2549
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2426017801401062e-05,
      "loss": 0.1602,
      "step": 2550
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2419643918673525e-05,
      "loss": 0.119,
      "step": 2551
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2413268991629719e-05,
      "loss": 0.1681,
      "step": 2552
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2406893023021052e-05,
      "loss": 0.1401,
      "step": 2553
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2400516015599389e-05,
      "loss": 0.1428,
      "step": 2554
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2394137972117042e-05,
      "loss": 0.1623,
      "step": 2555
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.238775889532677e-05,
      "loss": 0.15,
      "step": 2556
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.238137878798177e-05,
      "loss": 0.1468,
      "step": 2557
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.23749976528357e-05,
      "loss": 0.1538,
      "step": 2558
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2368615492642651e-05,
      "loss": 0.1767,
      "step": 2559
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2362232310157149e-05,
      "loss": 0.1256,
      "step": 2560
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.235584810813418e-05,
      "loss": 0.153,
      "step": 2561
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2349462889329158e-05,
      "loss": 0.1767,
      "step": 2562
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2343076656497939e-05,
      "loss": 0.1474,
      "step": 2563
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.2336689412396811e-05,
      "loss": 0.15,
      "step": 2564
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2330301159782507e-05,
      "loss": 0.1671,
      "step": 2565
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.232391190141219e-05,
      "loss": 0.1545,
      "step": 2566
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.231752164004346e-05,
      "loss": 0.1533,
      "step": 2567
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2311130378434346e-05,
      "loss": 0.1686,
      "step": 2568
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2304738119343313e-05,
      "loss": 0.1786,
      "step": 2569
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2298344865529254e-05,
      "loss": 0.1595,
      "step": 2570
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2291950619751493e-05,
      "loss": 0.1771,
      "step": 2571
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.228555538476978e-05,
      "loss": 0.177,
      "step": 2572
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2279159163344294e-05,
      "loss": 0.1608,
      "step": 2573
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2272761958235634e-05,
      "loss": 0.1504,
      "step": 2574
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2266363772204834e-05,
      "loss": 0.1671,
      "step": 2575
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2259964608013346e-05,
      "loss": 0.1696,
      "step": 2576
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2253564468423039e-05,
      "loss": 0.1354,
      "step": 2577
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2247163356196212e-05,
      "loss": 0.1408,
      "step": 2578
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.224076127409558e-05,
      "loss": 0.1457,
      "step": 2579
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.223435822488427e-05,
      "loss": 0.1787,
      "step": 2580
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2227954211325843e-05,
      "loss": 0.1546,
      "step": 2581
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2221549236184255e-05,
      "loss": 0.1643,
      "step": 2582
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2215143302223895e-05,
      "loss": 0.1518,
      "step": 2583
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2208736412209557e-05,
      "loss": 0.147,
      "step": 2584
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2202328568906447e-05,
      "loss": 0.1539,
      "step": 2585
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.219591977508019e-05,
      "loss": 0.1351,
      "step": 2586
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.218951003349681e-05,
      "loss": 0.1638,
      "step": 2587
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.218309934692275e-05,
      "loss": 0.1375,
      "step": 2588
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2176687718124854e-05,
      "loss": 0.1363,
      "step": 2589
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.2170275149870374e-05,
      "loss": 0.1542,
      "step": 2590
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2163861644926978e-05,
      "loss": 0.165,
      "step": 2591
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.215744720606272e-05,
      "loss": 0.1647,
      "step": 2592
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2151031836046072e-05,
      "loss": 0.1535,
      "step": 2593
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2144615537645898e-05,
      "loss": 0.1743,
      "step": 2594
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2138198313631473e-05,
      "loss": 0.1238,
      "step": 2595
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2131780166772458e-05,
      "loss": 0.1561,
      "step": 2596
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2125361099838924e-05,
      "loss": 0.1485,
      "step": 2597
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2118941115601334e-05,
      "loss": 0.1592,
      "step": 2598
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.211252021683055e-05,
      "loss": 0.1278,
      "step": 2599
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2106098406297822e-05,
      "loss": 0.1671,
      "step": 2600
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2099675686774804e-05,
      "loss": 0.1559,
      "step": 2601
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2093252061033532e-05,
      "loss": 0.1372,
      "step": 2602
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2086827531846434e-05,
      "loss": 0.1626,
      "step": 2603
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2080402101986337e-05,
      "loss": 0.1752,
      "step": 2604
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2073975774226452e-05,
      "loss": 0.1506,
      "step": 2605
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2067548551340373e-05,
      "loss": 0.1748,
      "step": 2606
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2061120436102087e-05,
      "loss": 0.1483,
      "step": 2607
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2054691431285962e-05,
      "loss": 0.1474,
      "step": 2608
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2048261539666745e-05,
      "loss": 0.1618,
      "step": 2609
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2041830764019584e-05,
      "loss": 0.1567,
      "step": 2610
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2035399107119987e-05,
      "loss": 0.137,
      "step": 2611
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2028966571743851e-05,
      "loss": 0.149,
      "step": 2612
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.202253316066746e-05,
      "loss": 0.1141,
      "step": 2613
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2016098876667465e-05,
      "loss": 0.1379,
      "step": 2614
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.20096637225209e-05,
      "loss": 0.1529,
      "step": 2615
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2003227701005163e-05,
      "loss": 0.1467,
      "step": 2616
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.1996790814898048e-05,
      "loss": 0.1705,
      "step": 2617
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.19903530669777e-05,
      "loss": 0.166,
      "step": 2618
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.198391446002265e-05,
      "loss": 0.1626,
      "step": 2619
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1977474996811793e-05,
      "loss": 0.1524,
      "step": 2620
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1971034680124401e-05,
      "loss": 0.1656,
      "step": 2621
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1964593512740103e-05,
      "loss": 0.1394,
      "step": 2622
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1958151497438904e-05,
      "loss": 0.1853,
      "step": 2623
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1951708637001173e-05,
      "loss": 0.178,
      "step": 2624
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1945264934207646e-05,
      "loss": 0.1666,
      "step": 2625
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1938820391839408e-05,
      "loss": 0.1409,
      "step": 2626
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1932375012677933e-05,
      "loss": 0.1412,
      "step": 2627
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1925928799505033e-05,
      "loss": 0.165,
      "step": 2628
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1919481755102889e-05,
      "loss": 0.1751,
      "step": 2629
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1913033882254042e-05,
      "loss": 0.1327,
      "step": 2630
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1906585183741387e-05,
      "loss": 0.1418,
      "step": 2631
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1900135662348173e-05,
      "loss": 0.142,
      "step": 2632
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1893685320858017e-05,
      "loss": 0.1417,
      "step": 2633
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1887234162054872e-05,
      "loss": 0.1403,
      "step": 2634
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.188078218872306e-05,
      "loss": 0.1397,
      "step": 2635
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1874329403647237e-05,
      "loss": 0.1801,
      "step": 2636
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1867875809612435e-05,
      "loss": 0.159,
      "step": 2637
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1861421409404006e-05,
      "loss": 0.1596,
      "step": 2638
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.185496620580767e-05,
      "loss": 0.1612,
      "step": 2639
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1848510201609487e-05,
      "loss": 0.1463,
      "step": 2640
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.184205339959586e-05,
      "loss": 0.1786,
      "step": 2641
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1835595802553542e-05,
      "loss": 0.1777,
      "step": 2642
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1829137413269624e-05,
      "loss": 0.1502,
      "step": 2643
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1822678234531549e-05,
      "loss": 0.1588,
      "step": 2644
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1816218269127082e-05,
      "loss": 0.1489,
      "step": 2645
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1809757519844344e-05,
      "loss": 0.1696,
      "step": 2646
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1803295989471792e-05,
      "loss": 0.1713,
      "step": 2647
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1796833680798212e-05,
      "loss": 0.1418,
      "step": 2648
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1790370596612733e-05,
      "loss": 0.1463,
      "step": 2649
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1783906739704819e-05,
      "loss": 0.1348,
      "step": 2650
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1777442112864259e-05,
      "loss": 0.1629,
      "step": 2651
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1770976718881185e-05,
      "loss": 0.141,
      "step": 2652
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1764510560546057e-05,
      "loss": 0.1532,
      "step": 2653
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1758043640649659e-05,
      "loss": 0.1449,
      "step": 2654
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.175157596198311e-05,
      "loss": 0.1633,
      "step": 2655
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1745107527337855e-05,
      "loss": 0.1376,
      "step": 2656
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1738638339505668e-05,
      "loss": 0.1561,
      "step": 2657
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1732168401278638e-05,
      "loss": 0.17,
      "step": 2658
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1725697715449191e-05,
      "loss": 0.1236,
      "step": 2659
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.171922628481007e-05,
      "loss": 0.1431,
      "step": 2660
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1712754112154331e-05,
      "loss": 0.1555,
      "step": 2661
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1706281200275368e-05,
      "loss": 0.1912,
      "step": 2662
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1699807551966877e-05,
      "loss": 0.1415,
      "step": 2663
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1693333170022884e-05,
      "loss": 0.1391,
      "step": 2664
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1686858057237725e-05,
      "loss": 0.1338,
      "step": 2665
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1680382216406052e-05,
      "loss": 0.1582,
      "step": 2666
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1673905650322834e-05,
      "loss": 0.144,
      "step": 2667
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.166742836178335e-05,
      "loss": 0.1645,
      "step": 2668
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1660950353583196e-05,
      "loss": 0.1467,
      "step": 2669
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.165447162851827e-05,
      "loss": 0.1602,
      "step": 2670
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1647992189384787e-05,
      "loss": 0.1999,
      "step": 2671
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1641512038979265e-05,
      "loss": 0.1482,
      "step": 2672
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1635031180098537e-05,
      "loss": 0.153,
      "step": 2673
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.162854961553973e-05,
      "loss": 0.1705,
      "step": 2674
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1622067348100283e-05,
      "loss": 0.1449,
      "step": 2675
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1615584380577941e-05,
      "loss": 0.175,
      "step": 2676
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1609100715770747e-05,
      "loss": 0.1767,
      "step": 2677
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1602616356477036e-05,
      "loss": 0.1574,
      "step": 2678
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1596131305495464e-05,
      "loss": 0.1362,
      "step": 2679
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1589645565624968e-05,
      "loss": 0.1563,
      "step": 2680
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1583159139664785e-05,
      "loss": 0.1465,
      "step": 2681
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.157667203041446e-05,
      "loss": 0.1963,
      "step": 2682
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1570184240673817e-05,
      "loss": 0.178,
      "step": 2683
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1563695773242978e-05,
      "loss": 0.1444,
      "step": 2684
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1557206630922364e-05,
      "loss": 0.1513,
      "step": 2685
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1550716816512686e-05,
      "loss": 0.1413,
      "step": 2686
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.154422633281494e-05,
      "loss": 0.1365,
      "step": 2687
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1537735182630409e-05,
      "loss": 0.1653,
      "step": 2688
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1531243368760676e-05,
      "loss": 0.1329,
      "step": 2689
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1524750894007599e-05,
      "loss": 0.137,
      "step": 2690
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1518257761173319e-05,
      "loss": 0.1486,
      "step": 2691
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1511763973060273e-05,
      "loss": 0.1671,
      "step": 2692
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1505269532471175e-05,
      "loss": 0.1555,
      "step": 2693
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1498774442209018e-05,
      "loss": 0.1408,
      "step": 2694
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1492278705077076e-05,
      "loss": 0.1509,
      "step": 2695
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1485782323878907e-05,
      "loss": 0.1405,
      "step": 2696
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.1479285301418339e-05,
      "loss": 0.1695,
      "step": 2697
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1472787640499488e-05,
      "loss": 0.1314,
      "step": 2698
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1466289343926733e-05,
      "loss": 0.1454,
      "step": 2699
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1459790414504735e-05,
      "loss": 0.1509,
      "step": 2700
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1453290855038425e-05,
      "loss": 0.1175,
      "step": 2701
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1446790668333012e-05,
      "loss": 0.1591,
      "step": 2702
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.144028985719397e-05,
      "loss": 0.134,
      "step": 2703
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1433788424427038e-05,
      "loss": 0.1613,
      "step": 2704
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1427286372838235e-05,
      "loss": 0.1178,
      "step": 2705
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1420783705233843e-05,
      "loss": 0.1617,
      "step": 2706
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1414280424420394e-05,
      "loss": 0.1551,
      "step": 2707
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1407776533204715e-05,
      "loss": 0.1337,
      "step": 2708
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1401272034393873e-05,
      "loss": 0.1556,
      "step": 2709
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1394766930795202e-05,
      "loss": 0.1772,
      "step": 2710
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1388261225216304e-05,
      "loss": 0.1417,
      "step": 2711
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1381754920465033e-05,
      "loss": 0.1438,
      "step": 2712
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1375248019349502e-05,
      "loss": 0.1408,
      "step": 2713
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.136874052467809e-05,
      "loss": 0.137,
      "step": 2714
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1362232439259425e-05,
      "loss": 0.1593,
      "step": 2715
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1355723765902388e-05,
      "loss": 0.1251,
      "step": 2716
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.134921450741612e-05,
      "loss": 0.1181,
      "step": 2717
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.134270466661001e-05,
      "loss": 0.1512,
      "step": 2718
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.13361942462937e-05,
      "loss": 0.1874,
      "step": 2719
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1329683249277076e-05,
      "loss": 0.1681,
      "step": 2720
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1323171678370291e-05,
      "loss": 0.1767,
      "step": 2721
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1316659536383723e-05,
      "loss": 0.1344,
      "step": 2722
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.1310146826128014e-05,
      "loss": 0.1601,
      "step": 2723
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1303633550414039e-05,
      "loss": 0.1366,
      "step": 2724
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1297119712052929e-05,
      "loss": 0.1786,
      "step": 2725
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1290605313856045e-05,
      "loss": 0.1576,
      "step": 2726
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1284090358634994e-05,
      "loss": 0.1702,
      "step": 2727
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1277574849201632e-05,
      "loss": 0.1378,
      "step": 2728
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1271058788368041e-05,
      "loss": 0.1362,
      "step": 2729
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.126454217894655e-05,
      "loss": 0.1622,
      "step": 2730
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1258025023749728e-05,
      "loss": 0.1478,
      "step": 2731
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1251507325590366e-05,
      "loss": 0.1665,
      "step": 2732
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1244989087281497e-05,
      "loss": 0.1523,
      "step": 2733
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1238470311636392e-05,
      "loss": 0.1472,
      "step": 2734
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1231951001468545e-05,
      "loss": 0.1711,
      "step": 2735
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1225431159591684e-05,
      "loss": 0.1539,
      "step": 2736
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1218910788819772e-05,
      "loss": 0.1647,
      "step": 2737
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1212389891966993e-05,
      "loss": 0.1406,
      "step": 2738
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.120586847184776e-05,
      "loss": 0.1375,
      "step": 2739
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1199346531276714e-05,
      "loss": 0.1656,
      "step": 2740
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1192824073068717e-05,
      "loss": 0.1473,
      "step": 2741
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.118630110003886e-05,
      "loss": 0.1333,
      "step": 2742
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1179777615002445e-05,
      "loss": 0.1437,
      "step": 2743
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1173253620775009e-05,
      "loss": 0.1384,
      "step": 2744
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1166729120172303e-05,
      "loss": 0.1289,
      "step": 2745
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1160204116010291e-05,
      "loss": 0.1372,
      "step": 2746
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1153678611105167e-05,
      "loss": 0.1527,
      "step": 2747
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1147152608273329e-05,
      "loss": 0.1083,
      "step": 2748
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1140626110331389e-05,
      "loss": 0.1443,
      "step": 2749
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.1134099120096187e-05,
      "loss": 0.1562,
      "step": 2750
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.112757164038476e-05,
      "loss": 0.1587,
      "step": 2751
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1121043674014367e-05,
      "loss": 0.1622,
      "step": 2752
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1114515223802469e-05,
      "loss": 0.1688,
      "step": 2753
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1107986292566746e-05,
      "loss": 0.1406,
      "step": 2754
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1101456883125072e-05,
      "loss": 0.137,
      "step": 2755
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1094926998295536e-05,
      "loss": 0.1503,
      "step": 2756
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1088396640896435e-05,
      "loss": 0.1386,
      "step": 2757
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1081865813746259e-05,
      "loss": 0.1343,
      "step": 2758
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1075334519663714e-05,
      "loss": 0.1528,
      "step": 2759
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.10688027614677e-05,
      "loss": 0.169,
      "step": 2760
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1062270541977316e-05,
      "loss": 0.1373,
      "step": 2761
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1055737864011861e-05,
      "loss": 0.137,
      "step": 2762
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1049204730390839e-05,
      "loss": 0.1666,
      "step": 2763
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1042671143933941e-05,
      "loss": 0.1235,
      "step": 2764
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1036137107461056e-05,
      "loss": 0.1737,
      "step": 2765
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1029602623792268e-05,
      "loss": 0.1486,
      "step": 2766
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1023067695747863e-05,
      "loss": 0.1405,
      "step": 2767
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1016532326148301e-05,
      "loss": 0.1654,
      "step": 2768
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1009996517814245e-05,
      "loss": 0.122,
      "step": 2769
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.1003460273566546e-05,
      "loss": 0.1615,
      "step": 2770
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.0996923596226244e-05,
      "loss": 0.1628,
      "step": 2771
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.0990386488614553e-05,
      "loss": 0.1554,
      "step": 2772
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.0983848953552891e-05,
      "loss": 0.1603,
      "step": 2773
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.0977310993862856e-05,
      "loss": 0.1562,
      "step": 2774
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.097077261236622e-05,
      "loss": 0.1441,
      "step": 2775
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.0964233811884942e-05,
      "loss": 0.1483,
      "step": 2776
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0957694595241169e-05,
      "loss": 0.1533,
      "step": 2777
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0951154965257214e-05,
      "loss": 0.144,
      "step": 2778
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.094461492475558e-05,
      "loss": 0.1404,
      "step": 2779
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0938074476558939e-05,
      "loss": 0.1381,
      "step": 2780
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0931533623490152e-05,
      "loss": 0.1532,
      "step": 2781
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0924992368372236e-05,
      "loss": 0.1474,
      "step": 2782
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0918450714028397e-05,
      "loss": 0.1423,
      "step": 2783
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0911908663282007e-05,
      "loss": 0.1402,
      "step": 2784
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0905366218956603e-05,
      "loss": 0.1702,
      "step": 2785
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0898823383875909e-05,
      "loss": 0.1488,
      "step": 2786
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0892280160863799e-05,
      "loss": 0.157,
      "step": 2787
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0885736552744325e-05,
      "loss": 0.1214,
      "step": 2788
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0879192562341703e-05,
      "loss": 0.1664,
      "step": 2789
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0872648192480314e-05,
      "loss": 0.155,
      "step": 2790
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0866103445984701e-05,
      "loss": 0.1278,
      "step": 2791
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0859558325679574e-05,
      "loss": 0.1668,
      "step": 2792
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0853012834389798e-05,
      "loss": 0.1282,
      "step": 2793
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0846466974940403e-05,
      "loss": 0.1485,
      "step": 2794
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0839920750156577e-05,
      "loss": 0.1574,
      "step": 2795
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0833374162863664e-05,
      "loss": 0.1783,
      "step": 2796
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0826827215887168e-05,
      "loss": 0.1757,
      "step": 2797
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0820279912052741e-05,
      "loss": 0.1483,
      "step": 2798
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.08137322541862e-05,
      "loss": 0.1287,
      "step": 2799
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0807184245113505e-05,
      "loss": 0.1514,
      "step": 2800
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.080063588766077e-05,
      "loss": 0.1615,
      "step": 2801
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0794087184654258e-05,
      "loss": 0.13,
      "step": 2802
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.0787538138920395e-05,
      "loss": 0.1277,
      "step": 2803
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0780988753285732e-05,
      "loss": 0.1514,
      "step": 2804
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0774439030576984e-05,
      "loss": 0.1562,
      "step": 2805
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0767888973621006e-05,
      "loss": 0.1646,
      "step": 2806
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0761338585244795e-05,
      "loss": 0.1446,
      "step": 2807
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0754787868275489e-05,
      "loss": 0.1274,
      "step": 2808
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.074823682554038e-05,
      "loss": 0.1611,
      "step": 2809
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.074168545986689e-05,
      "loss": 0.1306,
      "step": 2810
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.073513377408258e-05,
      "loss": 0.1672,
      "step": 2811
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0728581771015154e-05,
      "loss": 0.1722,
      "step": 2812
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0722029453492452e-05,
      "loss": 0.1388,
      "step": 2813
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0715476824342448e-05,
      "loss": 0.164,
      "step": 2814
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0708923886393252e-05,
      "loss": 0.1285,
      "step": 2815
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0702370642473104e-05,
      "loss": 0.1433,
      "step": 2816
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0695817095410377e-05,
      "loss": 0.159,
      "step": 2817
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0689263248033581e-05,
      "loss": 0.1263,
      "step": 2818
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0682709103171352e-05,
      "loss": 0.1589,
      "step": 2819
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0676154663652449e-05,
      "loss": 0.1503,
      "step": 2820
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0669599932305762e-05,
      "loss": 0.1408,
      "step": 2821
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0663044911960311e-05,
      "loss": 0.1652,
      "step": 2822
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.065648960544524e-05,
      "loss": 0.1417,
      "step": 2823
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0649934015589805e-05,
      "loss": 0.1563,
      "step": 2824
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0643378145223398e-05,
      "loss": 0.1661,
      "step": 2825
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0636821997175528e-05,
      "loss": 0.1422,
      "step": 2826
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0630265574275822e-05,
      "loss": 0.131,
      "step": 2827
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0623708879354025e-05,
      "loss": 0.1709,
      "step": 2828
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.0617151915240004e-05,
      "loss": 0.1814,
      "step": 2829
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0610594684763737e-05,
      "loss": 0.1575,
      "step": 2830
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0604037190755321e-05,
      "loss": 0.1586,
      "step": 2831
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0597479436044959e-05,
      "loss": 0.1323,
      "step": 2832
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0590921423462983e-05,
      "loss": 0.1318,
      "step": 2833
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0584363155839815e-05,
      "loss": 0.1423,
      "step": 2834
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0577804636006006e-05,
      "loss": 0.1372,
      "step": 2835
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0571245866792204e-05,
      "loss": 0.1289,
      "step": 2836
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0564686851029165e-05,
      "loss": 0.131,
      "step": 2837
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.055812759154776e-05,
      "loss": 0.1598,
      "step": 2838
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0551568091178956e-05,
      "loss": 0.1268,
      "step": 2839
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0545008352753834e-05,
      "loss": 0.1409,
      "step": 2840
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0538448379103563e-05,
      "loss": 0.1548,
      "step": 2841
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0531888173059428e-05,
      "loss": 0.135,
      "step": 2842
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0525327737452808e-05,
      "loss": 0.1423,
      "step": 2843
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.051876707511518e-05,
      "loss": 0.1635,
      "step": 2844
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0512206188878117e-05,
      "loss": 0.1417,
      "step": 2845
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.05056450815733e-05,
      "loss": 0.1739,
      "step": 2846
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0499083756032487e-05,
      "loss": 0.1606,
      "step": 2847
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.049252221508755e-05,
      "loss": 0.1454,
      "step": 2848
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0485960461570443e-05,
      "loss": 0.1464,
      "step": 2849
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.047939849831321e-05,
      "loss": 0.1399,
      "step": 2850
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0472836328147988e-05,
      "loss": 0.1356,
      "step": 2851
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0466273953907009e-05,
      "loss": 0.1445,
      "step": 2852
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0459711378422582e-05,
      "loss": 0.1273,
      "step": 2853
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0453148604527112e-05,
      "loss": 0.1472,
      "step": 2854
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0446585635053091e-05,
      "loss": 0.1316,
      "step": 2855
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.0440022472833085e-05,
      "loss": 0.1832,
      "step": 2856
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.043345912069975e-05,
      "loss": 0.1378,
      "step": 2857
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0426895581485826e-05,
      "loss": 0.1602,
      "step": 2858
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.042033185802413e-05,
      "loss": 0.1271,
      "step": 2859
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0413767953147558e-05,
      "loss": 0.1334,
      "step": 2860
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0407203869689084e-05,
      "loss": 0.1683,
      "step": 2861
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0400639610481768e-05,
      "loss": 0.1746,
      "step": 2862
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0394075178358734e-05,
      "loss": 0.1507,
      "step": 2863
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0387510576153185e-05,
      "loss": 0.1335,
      "step": 2864
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.03809458066984e-05,
      "loss": 0.1443,
      "step": 2865
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0374380872827725e-05,
      "loss": 0.1643,
      "step": 2866
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0367815777374581e-05,
      "loss": 0.1454,
      "step": 2867
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.036125052317246e-05,
      "loss": 0.156,
      "step": 2868
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0354685113054917e-05,
      "loss": 0.1603,
      "step": 2869
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0348119549855577e-05,
      "loss": 0.1444,
      "step": 2870
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0341553836408138e-05,
      "loss": 0.1649,
      "step": 2871
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.033498797554635e-05,
      "loss": 0.1668,
      "step": 2872
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0328421970104033e-05,
      "loss": 0.1881,
      "step": 2873
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.032185582291507e-05,
      "loss": 0.1555,
      "step": 2874
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.031528953681341e-05,
      "loss": 0.1452,
      "step": 2875
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0308723114633048e-05,
      "loss": 0.1354,
      "step": 2876
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0302156559208049e-05,
      "loss": 0.1722,
      "step": 2877
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.029558987337254e-05,
      "loss": 0.1525,
      "step": 2878
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0289023059960687e-05,
      "loss": 0.1455,
      "step": 2879
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0282456121806728e-05,
      "loss": 0.1448,
      "step": 2880
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0275889061744944e-05,
      "loss": 0.155,
      "step": 2881
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0269321882609675e-05,
      "loss": 0.1426,
      "step": 2882
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.0262754587235307e-05,
      "loss": 0.1289,
      "step": 2883
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.025618717845628e-05,
      "loss": 0.1541,
      "step": 2884
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0249619659107088e-05,
      "loss": 0.1518,
      "step": 2885
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0243052032022257e-05,
      "loss": 0.1678,
      "step": 2886
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0236484300036377e-05,
      "loss": 0.1396,
      "step": 2887
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0229916465984074e-05,
      "loss": 0.1523,
      "step": 2888
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0223348532700015e-05,
      "loss": 0.1835,
      "step": 2889
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0216780503018919e-05,
      "loss": 0.1322,
      "step": 2890
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0210212379775544e-05,
      "loss": 0.148,
      "step": 2891
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.020364416580468e-05,
      "loss": 0.1721,
      "step": 2892
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0197075863941166e-05,
      "loss": 0.16,
      "step": 2893
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0190507477019878e-05,
      "loss": 0.1496,
      "step": 2894
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0183939007875723e-05,
      "loss": 0.1604,
      "step": 2895
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0177370459343648e-05,
      "loss": 0.1641,
      "step": 2896
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.017080183425863e-05,
      "loss": 0.1579,
      "step": 2897
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0164233135455686e-05,
      "loss": 0.1737,
      "step": 2898
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0157664365769858e-05,
      "loss": 0.1744,
      "step": 2899
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0151095528036226e-05,
      "loss": 0.1465,
      "step": 2900
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.014452662508989e-05,
      "loss": 0.178,
      "step": 2901
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0137957659765985e-05,
      "loss": 0.1335,
      "step": 2902
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0131388634899671e-05,
      "loss": 0.1625,
      "step": 2903
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0124819553326133e-05,
      "loss": 0.1709,
      "step": 2904
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.011825041788058e-05,
      "loss": 0.1862,
      "step": 2905
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0111681231398246e-05,
      "loss": 0.1673,
      "step": 2906
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0105111996714387e-05,
      "loss": 0.1624,
      "step": 2907
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0098542716664282e-05,
      "loss": 0.1489,
      "step": 2908
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.0091973394083221e-05,
      "loss": 0.15,
      "step": 2909
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0085404031806524e-05,
      "loss": 0.1554,
      "step": 2910
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0078834632669517e-05,
      "loss": 0.1591,
      "step": 2911
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0072265199507543e-05,
      "loss": 0.1736,
      "step": 2912
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0065695735155979e-05,
      "loss": 0.1754,
      "step": 2913
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0059126242450187e-05,
      "loss": 0.1244,
      "step": 2914
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0052556724225558e-05,
      "loss": 0.1452,
      "step": 2915
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0045987183317494e-05,
      "loss": 0.1698,
      "step": 2916
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.00394176225614e-05,
      "loss": 0.1575,
      "step": 2917
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.003284804479269e-05,
      "loss": 0.1702,
      "step": 2918
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0026278452846796e-05,
      "loss": 0.1447,
      "step": 2919
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0019708849559142e-05,
      "loss": 0.1485,
      "step": 2920
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0013139237765166e-05,
      "loss": 0.1617,
      "step": 2921
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.0006569620300303e-05,
      "loss": 0.1432,
      "step": 2922
    },
    {
      "epoch": 1.1,
      "learning_rate": 1e-05,
      "loss": 0.1462,
      "step": 2923
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.993430379699698e-06,
      "loss": 0.1417,
      "step": 2924
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.986860762234837e-06,
      "loss": 0.1496,
      "step": 2925
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.980291150440861e-06,
      "loss": 0.1814,
      "step": 2926
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.973721547153209e-06,
      "loss": 0.1616,
      "step": 2927
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.96715195520731e-06,
      "loss": 0.171,
      "step": 2928
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.960582377438605e-06,
      "loss": 0.1417,
      "step": 2929
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.954012816682511e-06,
      "loss": 0.1339,
      "step": 2930
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.947443275774442e-06,
      "loss": 0.1347,
      "step": 2931
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.940873757549815e-06,
      "loss": 0.1561,
      "step": 2932
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.934304264844023e-06,
      "loss": 0.1376,
      "step": 2933
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.927734800492457e-06,
      "loss": 0.1428,
      "step": 2934
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.921165367330488e-06,
      "loss": 0.1621,
      "step": 2935
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.914595968193481e-06,
      "loss": 0.1504,
      "step": 2936
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.908026605916782e-06,
      "loss": 0.1513,
      "step": 2937
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.90145728333572e-06,
      "loss": 0.1452,
      "step": 2938
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.894888003285615e-06,
      "loss": 0.1674,
      "step": 2939
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.888318768601758e-06,
      "loss": 0.1714,
      "step": 2940
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.881749582119423e-06,
      "loss": 0.1283,
      "step": 2941
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.875180446673872e-06,
      "loss": 0.1399,
      "step": 2942
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.868611365100334e-06,
      "loss": 0.1534,
      "step": 2943
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.862042340234019e-06,
      "loss": 0.1345,
      "step": 2944
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.855473374910115e-06,
      "loss": 0.138,
      "step": 2945
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.848904471963776e-06,
      "loss": 0.1293,
      "step": 2946
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.842335634230142e-06,
      "loss": 0.1555,
      "step": 2947
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.835766864544315e-06,
      "loss": 0.1741,
      "step": 2948
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.829198165741371e-06,
      "loss": 0.149,
      "step": 2949
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.822629540656357e-06,
      "loss": 0.1644,
      "step": 2950
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.816060992124278e-06,
      "loss": 0.1439,
      "step": 2951
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.809492522980123e-06,
      "loss": 0.1459,
      "step": 2952
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.802924136058836e-06,
      "loss": 0.1691,
      "step": 2953
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.796355834195323e-06,
      "loss": 0.1312,
      "step": 2954
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.78978762022446e-06,
      "loss": 0.1321,
      "step": 2955
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.783219496981083e-06,
      "loss": 0.1637,
      "step": 2956
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.776651467299986e-06,
      "loss": 0.168,
      "step": 2957
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.77008353401593e-06,
      "loss": 0.1661,
      "step": 2958
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.763515699963628e-06,
      "loss": 0.1396,
      "step": 2959
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.756947967977744e-06,
      "loss": 0.1335,
      "step": 2960
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.750380340892913e-06,
      "loss": 0.1414,
      "step": 2961
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.74381282154372e-06,
      "loss": 0.1288,
      "step": 2962
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.737245412764695e-06,
      "loss": 0.1524,
      "step": 2963
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.730678117390328e-06,
      "loss": 0.1066,
      "step": 2964
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.724110938255058e-06,
      "loss": 0.1357,
      "step": 2965
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.717543878193276e-06,
      "loss": 0.1207,
      "step": 2966
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.710976940039315e-06,
      "loss": 0.1246,
      "step": 2967
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.704410126627464e-06,
      "loss": 0.135,
      "step": 2968
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.697843440791953e-06,
      "loss": 0.1311,
      "step": 2969
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.691276885366956e-06,
      "loss": 0.1582,
      "step": 2970
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.684710463186596e-06,
      "loss": 0.1441,
      "step": 2971
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.678144177084933e-06,
      "loss": 0.1177,
      "step": 2972
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.67157802989597e-06,
      "loss": 0.1445,
      "step": 2973
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.665012024453656e-06,
      "loss": 0.1388,
      "step": 2974
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.658446163591864e-06,
      "loss": 0.1208,
      "step": 2975
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.651880450144423e-06,
      "loss": 0.158,
      "step": 2976
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.645314886945084e-06,
      "loss": 0.1511,
      "step": 2977
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.638749476827542e-06,
      "loss": 0.1622,
      "step": 2978
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.63218422262542e-06,
      "loss": 0.1445,
      "step": 2979
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.625619127172278e-06,
      "loss": 0.1486,
      "step": 2980
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.619054193301602e-06,
      "loss": 0.1326,
      "step": 2981
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.612489423846816e-06,
      "loss": 0.132,
      "step": 2982
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.605924821641269e-06,
      "loss": 0.1625,
      "step": 2983
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.599360389518234e-06,
      "loss": 0.1497,
      "step": 2984
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.592796130310918e-06,
      "loss": 0.1226,
      "step": 2985
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.586232046852447e-06,
      "loss": 0.1382,
      "step": 2986
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.579668141975875e-06,
      "loss": 0.1715,
      "step": 2987
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.573104418514179e-06,
      "loss": 0.1711,
      "step": 2988
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.566540879300254e-06,
      "loss": 0.151,
      "step": 2989
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.559977527166918e-06,
      "loss": 0.1696,
      "step": 2990
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.55341436494691e-06,
      "loss": 0.1404,
      "step": 2991
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.54685139547289e-06,
      "loss": 0.1462,
      "step": 2992
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.540288621577422e-06,
      "loss": 0.1485,
      "step": 2993
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.533726046092995e-06,
      "loss": 0.1493,
      "step": 2994
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.527163671852014e-06,
      "loss": 0.1239,
      "step": 2995
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.520601501686794e-06,
      "loss": 0.1507,
      "step": 2996
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.51403953842956e-06,
      "loss": 0.1797,
      "step": 2997
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.507477784912451e-06,
      "loss": 0.1668,
      "step": 2998
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.500916243967515e-06,
      "loss": 0.1562,
      "step": 2999
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.494354918426704e-06,
      "loss": 0.1293,
      "step": 3000
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.487793811121888e-06,
      "loss": 0.1737,
      "step": 3001
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.481232924884827e-06,
      "loss": 0.119,
      "step": 3002
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.474672262547197e-06,
      "loss": 0.1592,
      "step": 3003
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.468111826940572e-06,
      "loss": 0.1133,
      "step": 3004
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.461551620896438e-06,
      "loss": 0.1319,
      "step": 3005
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.454991647246168e-06,
      "loss": 0.1213,
      "step": 3006
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.448431908821044e-06,
      "loss": 0.1528,
      "step": 3007
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.441872408452242e-06,
      "loss": 0.1699,
      "step": 3008
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.435313148970839e-06,
      "loss": 0.1241,
      "step": 3009
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.428754133207799e-06,
      "loss": 0.153,
      "step": 3010
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.422195363993997e-06,
      "loss": 0.13,
      "step": 3011
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.415636844160188e-06,
      "loss": 0.1335,
      "step": 3012
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.40907857653702e-06,
      "loss": 0.151,
      "step": 3013
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.402520563955043e-06,
      "loss": 0.1584,
      "step": 3014
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.395962809244685e-06,
      "loss": 0.174,
      "step": 3015
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.389405315236266e-06,
      "loss": 0.1648,
      "step": 3016
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.382848084760001e-06,
      "loss": 0.1619,
      "step": 3017
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.37629112064598e-06,
      "loss": 0.1761,
      "step": 3018
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.369734425724181e-06,
      "loss": 0.1567,
      "step": 3019
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.363178002824474e-06,
      "loss": 0.1587,
      "step": 3020
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.356621854776604e-06,
      "loss": 0.1419,
      "step": 3021
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.350065984410198e-06,
      "loss": 0.1615,
      "step": 3022
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.343510394554764e-06,
      "loss": 0.1616,
      "step": 3023
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.33695508803969e-06,
      "loss": 0.1564,
      "step": 3024
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.330400067694241e-06,
      "loss": 0.1501,
      "step": 3025
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.323845336347554e-06,
      "loss": 0.1353,
      "step": 3026
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.317290896828651e-06,
      "loss": 0.1644,
      "step": 3027
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.310736751966422e-06,
      "loss": 0.169,
      "step": 3028
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.304182904589624e-06,
      "loss": 0.1365,
      "step": 3029
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.297629357526901e-06,
      "loss": 0.1539,
      "step": 3030
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.291076113606753e-06,
      "loss": 0.1264,
      "step": 3031
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.284523175657556e-06,
      "loss": 0.1316,
      "step": 3032
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.277970546507551e-06,
      "loss": 0.1565,
      "step": 3033
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.271418228984846e-06,
      "loss": 0.1294,
      "step": 3034
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.264866225917421e-06,
      "loss": 0.1626,
      "step": 3035
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.258314540133111e-06,
      "loss": 0.1685,
      "step": 3036
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.251763174459621e-06,
      "loss": 0.1374,
      "step": 3037
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.245212131724513e-06,
      "loss": 0.1607,
      "step": 3038
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.238661414755209e-06,
      "loss": 0.1529,
      "step": 3039
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.232111026378997e-06,
      "loss": 0.139,
      "step": 3040
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.225560969423019e-06,
      "loss": 0.2029,
      "step": 3041
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.21901124671427e-06,
      "loss": 0.1743,
      "step": 3042
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.212461861079608e-06,
      "loss": 0.138,
      "step": 3043
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.205912815345743e-06,
      "loss": 0.1473,
      "step": 3044
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.199364112339236e-06,
      "loss": 0.1652,
      "step": 3045
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.192815754886502e-06,
      "loss": 0.1482,
      "step": 3046
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.186267745813806e-06,
      "loss": 0.1396,
      "step": 3047
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.179720087947259e-06,
      "loss": 0.1554,
      "step": 3048
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.173172784112833e-06,
      "loss": 0.1631,
      "step": 3049
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.166625837136338e-06,
      "loss": 0.1509,
      "step": 3050
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.160079249843425e-06,
      "loss": 0.1281,
      "step": 3051
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.153533025059598e-06,
      "loss": 0.1344,
      "step": 3052
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.146987165610205e-06,
      "loss": 0.1432,
      "step": 3053
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.14044167432043e-06,
      "loss": 0.1877,
      "step": 3054
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.1338965540153e-06,
      "loss": 0.1379,
      "step": 3055
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.12735180751969e-06,
      "loss": 0.1546,
      "step": 3056
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.120807437658302e-06,
      "loss": 0.1658,
      "step": 3057
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.114263447255677e-06,
      "loss": 0.1308,
      "step": 3058
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.107719839136206e-06,
      "loss": 0.1585,
      "step": 3059
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.101176616124096e-06,
      "loss": 0.1827,
      "step": 3060
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.094633781043398e-06,
      "loss": 0.1664,
      "step": 3061
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.088091336717998e-06,
      "loss": 0.1369,
      "step": 3062
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.081549285971603e-06,
      "loss": 0.1498,
      "step": 3063
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.075007631627764e-06,
      "loss": 0.1553,
      "step": 3064
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.068466376509848e-06,
      "loss": 0.1665,
      "step": 3065
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.06192552344106e-06,
      "loss": 0.1577,
      "step": 3066
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.055385075244424e-06,
      "loss": 0.1202,
      "step": 3067
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.048845034742788e-06,
      "loss": 0.1281,
      "step": 3068
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.042305404758836e-06,
      "loss": 0.1431,
      "step": 3069
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.03576618811506e-06,
      "loss": 0.1639,
      "step": 3070
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.029227387633783e-06,
      "loss": 0.1332,
      "step": 3071
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.022689006137148e-06,
      "loss": 0.1414,
      "step": 3072
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.01615104644711e-06,
      "loss": 0.1264,
      "step": 3073
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.00961351138545e-06,
      "loss": 0.1562,
      "step": 3074
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.003076403773763e-06,
      "loss": 0.1512,
      "step": 3075
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.996539726433458e-06,
      "loss": 0.1447,
      "step": 3076
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.990003482185757e-06,
      "loss": 0.1709,
      "step": 3077
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.9834676738517e-06,
      "loss": 0.1369,
      "step": 3078
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.976932304252139e-06,
      "loss": 0.1792,
      "step": 3079
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.970397376207732e-06,
      "loss": 0.1319,
      "step": 3080
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.963862892538949e-06,
      "loss": 0.1413,
      "step": 3081
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.957328856066062e-06,
      "loss": 0.1601,
      "step": 3082
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.950795269609165e-06,
      "loss": 0.1643,
      "step": 3083
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.94426213598814e-06,
      "loss": 0.1455,
      "step": 3084
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.937729458022685e-06,
      "loss": 0.1303,
      "step": 3085
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.931197238532301e-06,
      "loss": 0.1333,
      "step": 3086
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.924665480336288e-06,
      "loss": 0.1669,
      "step": 3087
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.918134186253743e-06,
      "loss": 0.1622,
      "step": 3088
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.91160335910357e-06,
      "loss": 0.1694,
      "step": 3089
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.905073001704466e-06,
      "loss": 0.1527,
      "step": 3090
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.898543116874933e-06,
      "loss": 0.1621,
      "step": 3091
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.892013707433255e-06,
      "loss": 0.1605,
      "step": 3092
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.885484776197531e-06,
      "loss": 0.1751,
      "step": 3093
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.878956325985636e-06,
      "loss": 0.1629,
      "step": 3094
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.87242835961524e-06,
      "loss": 0.1432,
      "step": 3095
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.865900879903817e-06,
      "loss": 0.1725,
      "step": 3096
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.859373889668615e-06,
      "loss": 0.1502,
      "step": 3097
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.852847391726676e-06,
      "loss": 0.1516,
      "step": 3098
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.846321388894837e-06,
      "loss": 0.1247,
      "step": 3099
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.83979588398971e-06,
      "loss": 0.1503,
      "step": 3100
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.8332708798277e-06,
      "loss": 0.1506,
      "step": 3101
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.826746379224994e-06,
      "loss": 0.1694,
      "step": 3102
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.820222384997561e-06,
      "loss": 0.1776,
      "step": 3103
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.813698899961147e-06,
      "loss": 0.1702,
      "step": 3104
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.807175926931288e-06,
      "loss": 0.1487,
      "step": 3105
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.800653468723293e-06,
      "loss": 0.1531,
      "step": 3106
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.794131528152242e-06,
      "loss": 0.1216,
      "step": 3107
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.787610108033007e-06,
      "loss": 0.1396,
      "step": 3108
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.78108921118023e-06,
      "loss": 0.1421,
      "step": 3109
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.774568840408318e-06,
      "loss": 0.1521,
      "step": 3110
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.768048998531457e-06,
      "loss": 0.1473,
      "step": 3111
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.761529688363612e-06,
      "loss": 0.1225,
      "step": 3112
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.755010912718507e-06,
      "loss": 0.1409,
      "step": 3113
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.748492674409637e-06,
      "loss": 0.1425,
      "step": 3114
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.741974976250276e-06,
      "loss": 0.1292,
      "step": 3115
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.735457821053451e-06,
      "loss": 0.1405,
      "step": 3116
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.728941211631962e-06,
      "loss": 0.1792,
      "step": 3117
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.722425150798373e-06,
      "loss": 0.127,
      "step": 3118
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.715909641365011e-06,
      "loss": 0.1578,
      "step": 3119
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.709394686143962e-06,
      "loss": 0.1505,
      "step": 3120
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.702880287947076e-06,
      "loss": 0.1637,
      "step": 3121
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.69636644958596e-06,
      "loss": 0.1582,
      "step": 3122
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.689853173871987e-06,
      "loss": 0.1756,
      "step": 3123
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.683340463616277e-06,
      "loss": 0.1228,
      "step": 3124
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.67682832162971e-06,
      "loss": 0.153,
      "step": 3125
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.670316750722925e-06,
      "loss": 0.1416,
      "step": 3126
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.663805753706305e-06,
      "loss": 0.1522,
      "step": 3127
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.657295333389995e-06,
      "loss": 0.1551,
      "step": 3128
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.650785492583885e-06,
      "loss": 0.1372,
      "step": 3129
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.644276234097613e-06,
      "loss": 0.1575,
      "step": 3130
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.637767560740577e-06,
      "loss": 0.1797,
      "step": 3131
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.631259475321913e-06,
      "loss": 0.1437,
      "step": 3132
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.624751980650499e-06,
      "loss": 0.1373,
      "step": 3133
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.618245079534972e-06,
      "loss": 0.1558,
      "step": 3134
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.611738774783701e-06,
      "loss": 0.1502,
      "step": 3135
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.6052330692048e-06,
      "loss": 0.137,
      "step": 3136
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.598727965606128e-06,
      "loss": 0.1585,
      "step": 3137
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.592223466795287e-06,
      "loss": 0.1322,
      "step": 3138
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.585719575579606e-06,
      "loss": 0.1495,
      "step": 3139
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.579216294766162e-06,
      "loss": 0.1736,
      "step": 3140
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.572713627161767e-06,
      "loss": 0.1655,
      "step": 3141
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.566211575572965e-06,
      "loss": 0.1301,
      "step": 3142
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.559710142806034e-06,
      "loss": 0.1607,
      "step": 3143
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.55320933166699e-06,
      "loss": 0.1576,
      "step": 3144
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.546709144961578e-06,
      "loss": 0.142,
      "step": 3145
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.54020958549527e-06,
      "loss": 0.1359,
      "step": 3146
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.533710656073272e-06,
      "loss": 0.1355,
      "step": 3147
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.527212359500518e-06,
      "loss": 0.1569,
      "step": 3148
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.520714698581663e-06,
      "loss": 0.163,
      "step": 3149
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.514217676121098e-06,
      "loss": 0.1729,
      "step": 3150
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.507721294922926e-06,
      "loss": 0.1199,
      "step": 3151
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.501225557790983e-06,
      "loss": 0.1365,
      "step": 3152
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.494730467528825e-06,
      "loss": 0.147,
      "step": 3153
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.488236026939727e-06,
      "loss": 0.1778,
      "step": 3154
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.481742238826683e-06,
      "loss": 0.1321,
      "step": 3155
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.475249105992404e-06,
      "loss": 0.1455,
      "step": 3156
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.468756631239326e-06,
      "loss": 0.1291,
      "step": 3157
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.462264817369593e-06,
      "loss": 0.1495,
      "step": 3158
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.455773667185063e-06,
      "loss": 0.1679,
      "step": 3159
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.449283183487317e-06,
      "loss": 0.1433,
      "step": 3160
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.442793369077638e-06,
      "loss": 0.1747,
      "step": 3161
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.436304226757025e-06,
      "loss": 0.1362,
      "step": 3162
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.42981575932619e-06,
      "loss": 0.1339,
      "step": 3163
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.423327969585547e-06,
      "loss": 0.1533,
      "step": 3164
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.416840860335217e-06,
      "loss": 0.1534,
      "step": 3165
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.410354434375034e-06,
      "loss": 0.1329,
      "step": 3166
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.403868694504538e-06,
      "loss": 0.1845,
      "step": 3167
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.397383643522965e-06,
      "loss": 0.1411,
      "step": 3168
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.390899284229256e-06,
      "loss": 0.1629,
      "step": 3169
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.38441561942206e-06,
      "loss": 0.15,
      "step": 3170
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.377932651899718e-06,
      "loss": 0.1361,
      "step": 3171
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.371450384460272e-06,
      "loss": 0.1165,
      "step": 3172
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.364968819901467e-06,
      "loss": 0.1434,
      "step": 3173
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.358487961020737e-06,
      "loss": 0.1422,
      "step": 3174
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.352007810615216e-06,
      "loss": 0.1455,
      "step": 3175
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.345528371481734e-06,
      "loss": 0.1458,
      "step": 3176
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.339049646416809e-06,
      "loss": 0.1873,
      "step": 3177
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.332571638216652e-06,
      "loss": 0.1352,
      "step": 3178
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.32609434967717e-06,
      "loss": 0.1696,
      "step": 3179
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.319617783593948e-06,
      "loss": 0.1354,
      "step": 3180
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.313141942762277e-06,
      "loss": 0.1517,
      "step": 3181
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.306666829977119e-06,
      "loss": 0.1644,
      "step": 3182
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.300192448033123e-06,
      "loss": 0.1553,
      "step": 3183
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.293718799724635e-06,
      "loss": 0.1377,
      "step": 3184
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.28724588784567e-06,
      "loss": 0.1428,
      "step": 3185
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.280773715189934e-06,
      "loss": 0.1563,
      "step": 3186
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.27430228455081e-06,
      "loss": 0.1542,
      "step": 3187
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.267831598721365e-06,
      "loss": 0.1386,
      "step": 3188
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.261361660494336e-06,
      "loss": 0.1238,
      "step": 3189
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.254892472662148e-06,
      "loss": 0.1602,
      "step": 3190
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.248424038016895e-06,
      "loss": 0.13,
      "step": 3191
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.241956359350344e-06,
      "loss": 0.1268,
      "step": 3192
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.23548943945395e-06,
      "loss": 0.1743,
      "step": 3193
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.229023281118817e-06,
      "loss": 0.1362,
      "step": 3194
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.222557887135743e-06,
      "loss": 0.1534,
      "step": 3195
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.216093260295183e-06,
      "loss": 0.1577,
      "step": 3196
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.209629403387269e-06,
      "loss": 0.1346,
      "step": 3197
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.203166319201791e-06,
      "loss": 0.1511,
      "step": 3198
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.19670401052821e-06,
      "loss": 0.1767,
      "step": 3199
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.190242480155657e-06,
      "loss": 0.1454,
      "step": 3200
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.183781730872922e-06,
      "loss": 0.141,
      "step": 3201
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.177321765468456e-06,
      "loss": 0.1597,
      "step": 3202
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.170862586730377e-06,
      "loss": 0.1783,
      "step": 3203
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.164404197446463e-06,
      "loss": 0.1344,
      "step": 3204
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.157946600404144e-06,
      "loss": 0.1358,
      "step": 3205
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.151489798390518e-06,
      "loss": 0.1619,
      "step": 3206
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.145033794192334e-06,
      "loss": 0.1721,
      "step": 3207
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.138578590595998e-06,
      "loss": 0.1343,
      "step": 3208
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.132124190387572e-06,
      "loss": 0.1551,
      "step": 3209
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.125670596352761e-06,
      "loss": 0.1648,
      "step": 3210
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.119217811276946e-06,
      "loss": 0.1259,
      "step": 3211
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.11276583794513e-06,
      "loss": 0.1238,
      "step": 3212
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.106314679141987e-06,
      "loss": 0.122,
      "step": 3213
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.099864337651829e-06,
      "loss": 0.1595,
      "step": 3214
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.093414816258617e-06,
      "loss": 0.1391,
      "step": 3215
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.086966117745961e-06,
      "loss": 0.1511,
      "step": 3216
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.080518244897114e-06,
      "loss": 0.1551,
      "step": 3217
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.074071200494969e-06,
      "loss": 0.1553,
      "step": 3218
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.06762498732207e-06,
      "loss": 0.134,
      "step": 3219
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.061179608160594e-06,
      "loss": 0.1411,
      "step": 3220
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.054735065792361e-06,
      "loss": 0.1304,
      "step": 3221
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.048291362998832e-06,
      "loss": 0.1463,
      "step": 3222
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.041848502561102e-06,
      "loss": 0.1347,
      "step": 3223
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.0354064872599e-06,
      "loss": 0.1201,
      "step": 3224
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.0289653198756e-06,
      "loss": 0.1607,
      "step": 3225
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.022525003188208e-06,
      "loss": 0.1379,
      "step": 3226
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.016085539977354e-06,
      "loss": 0.1504,
      "step": 3227
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.009646933022302e-06,
      "loss": 0.1288,
      "step": 3228
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.003209185101957e-06,
      "loss": 0.134,
      "step": 3229
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.996772298994839e-06,
      "loss": 0.1741,
      "step": 3230
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.990336277479105e-06,
      "loss": 0.1473,
      "step": 3231
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.983901123332538e-06,
      "loss": 0.1349,
      "step": 3232
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.977466839332543e-06,
      "loss": 0.1501,
      "step": 3233
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.97103342825615e-06,
      "loss": 0.1454,
      "step": 3234
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.964600892880018e-06,
      "loss": 0.136,
      "step": 3235
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.958169235980423e-06,
      "loss": 0.1622,
      "step": 3236
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.951738460333256e-06,
      "loss": 0.147,
      "step": 3237
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.945308568714043e-06,
      "loss": 0.1427,
      "step": 3238
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.938879563897914e-06,
      "loss": 0.175,
      "step": 3239
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.932451448659629e-06,
      "loss": 0.1475,
      "step": 3240
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.926024225773548e-06,
      "loss": 0.1336,
      "step": 3241
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.919597898013663e-06,
      "loss": 0.1285,
      "step": 3242
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.91317246815357e-06,
      "loss": 0.127,
      "step": 3243
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.906747938966473e-06,
      "loss": 0.1656,
      "step": 3244
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.9003243132252e-06,
      "loss": 0.1463,
      "step": 3245
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.89390159370218e-06,
      "loss": 0.1306,
      "step": 3246
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.887479783169453e-06,
      "loss": 0.1462,
      "step": 3247
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.881058884398667e-06,
      "loss": 0.1251,
      "step": 3248
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.87463890016108e-06,
      "loss": 0.1434,
      "step": 3249
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.868219833227546e-06,
      "loss": 0.1867,
      "step": 3250
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.861801686368532e-06,
      "loss": 0.1671,
      "step": 3251
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.855384462354105e-06,
      "loss": 0.1412,
      "step": 3252
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.848968163953931e-06,
      "loss": 0.1531,
      "step": 3253
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.842552793937279e-06,
      "loss": 0.1352,
      "step": 3254
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.836138355073023e-06,
      "loss": 0.1344,
      "step": 3255
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.829724850129626e-06,
      "loss": 0.1746,
      "step": 3256
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.823312281875148e-06,
      "loss": 0.1355,
      "step": 3257
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.816900653077254e-06,
      "loss": 0.1588,
      "step": 3258
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.810489966503195e-06,
      "loss": 0.1322,
      "step": 3259
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.804080224919813e-06,
      "loss": 0.1373,
      "step": 3260
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.797671431093555e-06,
      "loss": 0.1544,
      "step": 3261
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.791263587790446e-06,
      "loss": 0.1606,
      "step": 3262
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.784856697776106e-06,
      "loss": 0.1416,
      "step": 3263
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.778450763815746e-06,
      "loss": 0.1441,
      "step": 3264
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.772045788674162e-06,
      "loss": 0.157,
      "step": 3265
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.765641775115731e-06,
      "loss": 0.1329,
      "step": 3266
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.759238725904424e-06,
      "loss": 0.1426,
      "step": 3267
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.752836643803787e-06,
      "loss": 0.1587,
      "step": 3268
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.746435531576961e-06,
      "loss": 0.1713,
      "step": 3269
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.740035391986654e-06,
      "loss": 0.1695,
      "step": 3270
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.733636227795168e-06,
      "loss": 0.138,
      "step": 3271
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.727238041764367e-06,
      "loss": 0.1351,
      "step": 3272
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.72084083665571e-06,
      "loss": 0.171,
      "step": 3273
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.714444615230223e-06,
      "loss": 0.1579,
      "step": 3274
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.70804938024851e-06,
      "loss": 0.1297,
      "step": 3275
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.701655134470747e-06,
      "loss": 0.1679,
      "step": 3276
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.69526188065669e-06,
      "loss": 0.172,
      "step": 3277
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.688869621565657e-06,
      "loss": 0.1509,
      "step": 3278
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.682478359956544e-06,
      "loss": 0.1736,
      "step": 3279
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.676088098587813e-06,
      "loss": 0.1448,
      "step": 3280
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.669698840217498e-06,
      "loss": 0.1553,
      "step": 3281
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.663310587603192e-06,
      "loss": 0.1384,
      "step": 3282
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.656923343502064e-06,
      "loss": 0.1448,
      "step": 3283
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.650537110670842e-06,
      "loss": 0.1595,
      "step": 3284
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.644151891865822e-06,
      "loss": 0.1474,
      "step": 3285
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.637767689842853e-06,
      "loss": 0.12,
      "step": 3286
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.631384507357354e-06,
      "loss": 0.1443,
      "step": 3287
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.625002347164303e-06,
      "loss": 0.152,
      "step": 3288
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.618621212018234e-06,
      "loss": 0.1641,
      "step": 3289
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.612241104673235e-06,
      "loss": 0.1493,
      "step": 3290
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.605862027882962e-06,
      "loss": 0.1239,
      "step": 3291
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.599483984400615e-06,
      "loss": 0.1623,
      "step": 3292
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.593106976978951e-06,
      "loss": 0.1173,
      "step": 3293
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.586731008370286e-06,
      "loss": 0.1342,
      "step": 3294
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.580356081326477e-06,
      "loss": 0.1344,
      "step": 3295
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.573982198598941e-06,
      "loss": 0.12,
      "step": 3296
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.567609362938641e-06,
      "loss": 0.1419,
      "step": 3297
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.5612375770960814e-06,
      "loss": 0.1521,
      "step": 3298
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.554866843821328e-06,
      "loss": 0.1745,
      "step": 3299
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.5484971658639774e-06,
      "loss": 0.1282,
      "step": 3300
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.5421285459731806e-06,
      "loss": 0.1654,
      "step": 3301
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.535760986897629e-06,
      "loss": 0.1312,
      "step": 3302
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.529394491385552e-06,
      "loss": 0.1237,
      "step": 3303
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.523029062184728e-06,
      "loss": 0.1637,
      "step": 3304
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.516664702042469e-06,
      "loss": 0.1421,
      "step": 3305
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.510301413705625e-06,
      "loss": 0.152,
      "step": 3306
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.503939199920592e-06,
      "loss": 0.1645,
      "step": 3307
    },
    {
      "epoch": 1.24,
      "learning_rate": 7.497578063433289e-06,
      "loss": 0.1582,
      "step": 3308
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.49121800698918e-06,
      "loss": 0.1358,
      "step": 3309
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.4848590333332635e-06,
      "loss": 0.1273,
      "step": 3310
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.478501145210064e-06,
      "loss": 0.1553,
      "step": 3311
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.472144345363636e-06,
      "loss": 0.1541,
      "step": 3312
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.465788636537574e-06,
      "loss": 0.1405,
      "step": 3313
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.4594340214750006e-06,
      "loss": 0.1428,
      "step": 3314
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.453080502918557e-06,
      "loss": 0.144,
      "step": 3315
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.446728083610417e-06,
      "loss": 0.1292,
      "step": 3316
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.44037676629228e-06,
      "loss": 0.1522,
      "step": 3317
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.434026553705369e-06,
      "loss": 0.1361,
      "step": 3318
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.427677448590429e-06,
      "loss": 0.1429,
      "step": 3319
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.421329453687732e-06,
      "loss": 0.1661,
      "step": 3320
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.414982571737065e-06,
      "loss": 0.158,
      "step": 3321
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.408636805477734e-06,
      "loss": 0.1425,
      "step": 3322
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.402292157648572e-06,
      "loss": 0.1418,
      "step": 3323
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.395948630987921e-06,
      "loss": 0.1754,
      "step": 3324
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.38960622823364e-06,
      "loss": 0.1496,
      "step": 3325
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.383264952123108e-06,
      "loss": 0.1467,
      "step": 3326
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.3769248053932085e-06,
      "loss": 0.138,
      "step": 3327
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.370585790780352e-06,
      "loss": 0.1612,
      "step": 3328
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.364247911020444e-06,
      "loss": 0.1374,
      "step": 3329
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.357911168848916e-06,
      "loss": 0.1326,
      "step": 3330
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.351575567000695e-06,
      "loss": 0.1662,
      "step": 3331
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.345241108210222e-06,
      "loss": 0.1478,
      "step": 3332
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.338907795211445e-06,
      "loss": 0.1687,
      "step": 3333
    },
    {
      "epoch": 1.25,
      "learning_rate": 7.332575630737817e-06,
      "loss": 0.1411,
      "step": 3334
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.326244617522291e-06,
      "loss": 0.1501,
      "step": 3335
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.319914758297333e-06,
      "loss": 0.1635,
      "step": 3336
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.313586055794901e-06,
      "loss": 0.1547,
      "step": 3337
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.307258512746458e-06,
      "loss": 0.141,
      "step": 3338
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.300932131882967e-06,
      "loss": 0.1484,
      "step": 3339
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.294606915934891e-06,
      "loss": 0.1683,
      "step": 3340
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.288282867632181e-06,
      "loss": 0.1439,
      "step": 3341
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.281959989704295e-06,
      "loss": 0.1491,
      "step": 3342
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.275638284880184e-06,
      "loss": 0.1499,
      "step": 3343
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.2693177558882895e-06,
      "loss": 0.1421,
      "step": 3344
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.262998405456544e-06,
      "loss": 0.1266,
      "step": 3345
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.256680236312377e-06,
      "loss": 0.1338,
      "step": 3346
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.2503632511827015e-06,
      "loss": 0.1441,
      "step": 3347
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.244047452793925e-06,
      "loss": 0.1547,
      "step": 3348
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.237732843871941e-06,
      "loss": 0.198,
      "step": 3349
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.231419427142129e-06,
      "loss": 0.1515,
      "step": 3350
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.225107205329352e-06,
      "loss": 0.1376,
      "step": 3351
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.218796181157965e-06,
      "loss": 0.1777,
      "step": 3352
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.212486357351796e-06,
      "loss": 0.1496,
      "step": 3353
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.206177736634159e-06,
      "loss": 0.1486,
      "step": 3354
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.199870321727853e-06,
      "loss": 0.1628,
      "step": 3355
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.193564115355149e-06,
      "loss": 0.1456,
      "step": 3356
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.187259120237806e-06,
      "loss": 0.1328,
      "step": 3357
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.180955339097049e-06,
      "loss": 0.126,
      "step": 3358
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.174652774653588e-06,
      "loss": 0.1436,
      "step": 3359
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.1683514296276046e-06,
      "loss": 0.134,
      "step": 3360
    },
    {
      "epoch": 1.26,
      "learning_rate": 7.1620513067387506e-06,
      "loss": 0.1366,
      "step": 3361
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.155752408706158e-06,
      "loss": 0.1536,
      "step": 3362
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.149454738248423e-06,
      "loss": 0.1693,
      "step": 3363
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.143158298083615e-06,
      "loss": 0.1463,
      "step": 3364
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.136863090929274e-06,
      "loss": 0.1518,
      "step": 3365
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.130569119502404e-06,
      "loss": 0.1396,
      "step": 3366
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.124276386519478e-06,
      "loss": 0.1278,
      "step": 3367
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.117984894696437e-06,
      "loss": 0.1433,
      "step": 3368
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.111694646748682e-06,
      "loss": 0.1688,
      "step": 3369
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.105405645391075e-06,
      "loss": 0.1282,
      "step": 3370
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.099117893337943e-06,
      "loss": 0.1164,
      "step": 3371
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.092831393303085e-06,
      "loss": 0.1496,
      "step": 3372
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.0865461479997424e-06,
      "loss": 0.1411,
      "step": 3373
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.08026216014062e-06,
      "loss": 0.1684,
      "step": 3374
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.073979432437884e-06,
      "loss": 0.145,
      "step": 3375
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.067697967603155e-06,
      "loss": 0.1688,
      "step": 3376
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.0614177683475045e-06,
      "loss": 0.144,
      "step": 3377
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.055138837381465e-06,
      "loss": 0.1639,
      "step": 3378
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.048861177415017e-06,
      "loss": 0.1525,
      "step": 3379
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.042584791157591e-06,
      "loss": 0.1224,
      "step": 3380
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.036309681318072e-06,
      "loss": 0.151,
      "step": 3381
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.030035850604792e-06,
      "loss": 0.1349,
      "step": 3382
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.0237633017255305e-06,
      "loss": 0.1568,
      "step": 3383
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.017492037387514e-06,
      "loss": 0.1606,
      "step": 3384
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.011222060297417e-06,
      "loss": 0.1523,
      "step": 3385
    },
    {
      "epoch": 1.27,
      "learning_rate": 7.00495337316135e-06,
      "loss": 0.1371,
      "step": 3386
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.998685978684881e-06,
      "loss": 0.1425,
      "step": 3387
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.9924198795730084e-06,
      "loss": 0.1279,
      "step": 3388
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.9861550785301766e-06,
      "loss": 0.1609,
      "step": 3389
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.979891578260268e-06,
      "loss": 0.1363,
      "step": 3390
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.973629381466603e-06,
      "loss": 0.1391,
      "step": 3391
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.96736849085194e-06,
      "loss": 0.1585,
      "step": 3392
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.961108909118478e-06,
      "loss": 0.1372,
      "step": 3393
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.95485063896784e-06,
      "loss": 0.1272,
      "step": 3394
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.948593683101099e-06,
      "loss": 0.1488,
      "step": 3395
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.942338044218744e-06,
      "loss": 0.1696,
      "step": 3396
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.9360837250207105e-06,
      "loss": 0.132,
      "step": 3397
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.929830728206351e-06,
      "loss": 0.1609,
      "step": 3398
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.9235790564744565e-06,
      "loss": 0.1508,
      "step": 3399
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.917328712523242e-06,
      "loss": 0.1393,
      "step": 3400
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.911079699050352e-06,
      "loss": 0.1914,
      "step": 3401
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.9048320187528575e-06,
      "loss": 0.1592,
      "step": 3402
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.898585674327249e-06,
      "loss": 0.1451,
      "step": 3403
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.892340668469442e-06,
      "loss": 0.1331,
      "step": 3404
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.886097003874781e-06,
      "loss": 0.1765,
      "step": 3405
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.879854683238023e-06,
      "loss": 0.149,
      "step": 3406
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.873613709253347e-06,
      "loss": 0.1641,
      "step": 3407
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.867374084614356e-06,
      "loss": 0.1578,
      "step": 3408
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.8611358120140634e-06,
      "loss": 0.1623,
      "step": 3409
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.854898894144903e-06,
      "loss": 0.1358,
      "step": 3410
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.848663333698724e-06,
      "loss": 0.1549,
      "step": 3411
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.84242913336679e-06,
      "loss": 0.1708,
      "step": 3412
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.8361962958397735e-06,
      "loss": 0.1878,
      "step": 3413
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.829964823807764e-06,
      "loss": 0.1451,
      "step": 3414
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.823734719960257e-06,
      "loss": 0.1698,
      "step": 3415
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.817505986986166e-06,
      "loss": 0.1207,
      "step": 3416
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.811278627573802e-06,
      "loss": 0.1411,
      "step": 3417
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.80505264441089e-06,
      "loss": 0.147,
      "step": 3418
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.798828040184559e-06,
      "loss": 0.1123,
      "step": 3419
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.792604817581341e-06,
      "loss": 0.1637,
      "step": 3420
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.786382979287176e-06,
      "loss": 0.145,
      "step": 3421
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.780162527987404e-06,
      "loss": 0.1484,
      "step": 3422
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.773943466366764e-06,
      "loss": 0.1919,
      "step": 3423
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.767725797109401e-06,
      "loss": 0.1379,
      "step": 3424
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.761509522898854e-06,
      "loss": 0.1572,
      "step": 3425
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.7552946464180605e-06,
      "loss": 0.1537,
      "step": 3426
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.749081170349358e-06,
      "loss": 0.1492,
      "step": 3427
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.742869097374478e-06,
      "loss": 0.1443,
      "step": 3428
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.736658430174542e-06,
      "loss": 0.1547,
      "step": 3429
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.7304491714300705e-06,
      "loss": 0.1436,
      "step": 3430
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.724241323820979e-06,
      "loss": 0.148,
      "step": 3431
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.7180348900265655e-06,
      "loss": 0.128,
      "step": 3432
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.711829872725519e-06,
      "loss": 0.1578,
      "step": 3433
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.705626274595926e-06,
      "loss": 0.1224,
      "step": 3434
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.69942409831525e-06,
      "loss": 0.1317,
      "step": 3435
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.693223346560344e-06,
      "loss": 0.1476,
      "step": 3436
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.68702402200745e-06,
      "loss": 0.1577,
      "step": 3437
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.680826127332192e-06,
      "loss": 0.1612,
      "step": 3438
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.67462966520957e-06,
      "loss": 0.1453,
      "step": 3439
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.66843463831398e-06,
      "loss": 0.1789,
      "step": 3440
    },
    {
      "epoch": 1.29,
      "learning_rate": 6.6622410493191845e-06,
      "loss": 0.1209,
      "step": 3441
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.656048900898333e-06,
      "loss": 0.1315,
      "step": 3442
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.649858195723954e-06,
      "loss": 0.1505,
      "step": 3443
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.643668936467944e-06,
      "loss": 0.1631,
      "step": 3444
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.63748112580159e-06,
      "loss": 0.1378,
      "step": 3445
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.631294766395541e-06,
      "loss": 0.1588,
      "step": 3446
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.625109860919827e-06,
      "loss": 0.1657,
      "step": 3447
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.618926412043847e-06,
      "loss": 0.1452,
      "step": 3448
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.612744422436371e-06,
      "loss": 0.1441,
      "step": 3449
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.606563894765543e-06,
      "loss": 0.1416,
      "step": 3450
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.6003848316988726e-06,
      "loss": 0.1414,
      "step": 3451
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.594207235903236e-06,
      "loss": 0.1276,
      "step": 3452
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.588031110044882e-06,
      "loss": 0.1469,
      "step": 3453
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.581856456789419e-06,
      "loss": 0.166,
      "step": 3454
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.57568327880182e-06,
      "loss": 0.1529,
      "step": 3455
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.569511578746428e-06,
      "loss": 0.1478,
      "step": 3456
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.563341359286942e-06,
      "loss": 0.1313,
      "step": 3457
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.557172623086421e-06,
      "loss": 0.1573,
      "step": 3458
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.551005372807287e-06,
      "loss": 0.1248,
      "step": 3459
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.544839611111324e-06,
      "loss": 0.1228,
      "step": 3460
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.538675340659667e-06,
      "loss": 0.1725,
      "step": 3461
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.532512564112809e-06,
      "loss": 0.149,
      "step": 3462
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.526351284130601e-06,
      "loss": 0.1629,
      "step": 3463
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.520191503372244e-06,
      "loss": 0.1337,
      "step": 3464
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.5140332244962944e-06,
      "loss": 0.1489,
      "step": 3465
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.507876450160664e-06,
      "loss": 0.1563,
      "step": 3466
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.5017211830226075e-06,
      "loss": 0.143,
      "step": 3467
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.495567425738734e-06,
      "loss": 0.1271,
      "step": 3468
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.489415180965e-06,
      "loss": 0.134,
      "step": 3469
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.483264451356709e-06,
      "loss": 0.1909,
      "step": 3470
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.477115239568509e-06,
      "loss": 0.1394,
      "step": 3471
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.470967548254399e-06,
      "loss": 0.1617,
      "step": 3472
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.464821380067714e-06,
      "loss": 0.1539,
      "step": 3473
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.4586767376611315e-06,
      "loss": 0.1573,
      "step": 3474
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.452533623686681e-06,
      "loss": 0.124,
      "step": 3475
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.446392040795726e-06,
      "loss": 0.1462,
      "step": 3476
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.4402519916389635e-06,
      "loss": 0.1351,
      "step": 3477
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.4341134788664325e-06,
      "loss": 0.1345,
      "step": 3478
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.427976505127516e-06,
      "loss": 0.1209,
      "step": 3479
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.421841073070922e-06,
      "loss": 0.1365,
      "step": 3480
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.415707185344697e-06,
      "loss": 0.1566,
      "step": 3481
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.409574844596227e-06,
      "loss": 0.1595,
      "step": 3482
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.403444053472221e-06,
      "loss": 0.1636,
      "step": 3483
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.397314814618722e-06,
      "loss": 0.1363,
      "step": 3484
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.391187130681109e-06,
      "loss": 0.1434,
      "step": 3485
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.385061004304083e-06,
      "loss": 0.1629,
      "step": 3486
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.378936438131671e-06,
      "loss": 0.1484,
      "step": 3487
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.3728134348072356e-06,
      "loss": 0.1246,
      "step": 3488
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.366691996973457e-06,
      "loss": 0.1526,
      "step": 3489
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.360572127272349e-06,
      "loss": 0.1438,
      "step": 3490
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.354453828345235e-06,
      "loss": 0.1242,
      "step": 3491
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.348337102832768e-06,
      "loss": 0.1462,
      "step": 3492
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.342221953374925e-06,
      "loss": 0.1302,
      "step": 3493
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.336108382610999e-06,
      "loss": 0.1365,
      "step": 3494
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.329996393179598e-06,
      "loss": 0.1411,
      "step": 3495
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.323885987718657e-06,
      "loss": 0.1207,
      "step": 3496
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.317777168865418e-06,
      "loss": 0.148,
      "step": 3497
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.311669939256443e-06,
      "loss": 0.1492,
      "step": 3498
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.3055643015276094e-06,
      "loss": 0.1432,
      "step": 3499
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.2994602583140985e-06,
      "loss": 0.1288,
      "step": 3500
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.293357812250419e-06,
      "loss": 0.1439,
      "step": 3501
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.287256965970377e-06,
      "loss": 0.1602,
      "step": 3502
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.281157722107089e-06,
      "loss": 0.1725,
      "step": 3503
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.275060083292991e-06,
      "loss": 0.1446,
      "step": 3504
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.268964052159814e-06,
      "loss": 0.1695,
      "step": 3505
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.262869631338602e-06,
      "loss": 0.1333,
      "step": 3506
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.256776823459703e-06,
      "loss": 0.1491,
      "step": 3507
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.250685631152762e-06,
      "loss": 0.1618,
      "step": 3508
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.2445960570467375e-06,
      "loss": 0.155,
      "step": 3509
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.238508103769885e-06,
      "loss": 0.1554,
      "step": 3510
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.232421773949755e-06,
      "loss": 0.1316,
      "step": 3511
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.226337070213208e-06,
      "loss": 0.1409,
      "step": 3512
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.220253995186395e-06,
      "loss": 0.1372,
      "step": 3513
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.214172551494764e-06,
      "loss": 0.1329,
      "step": 3514
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.208092741763063e-06,
      "loss": 0.1472,
      "step": 3515
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.2020145686153334e-06,
      "loss": 0.1219,
      "step": 3516
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.195938034674906e-06,
      "loss": 0.1395,
      "step": 3517
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.189863142564407e-06,
      "loss": 0.1487,
      "step": 3518
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.183789894905764e-06,
      "loss": 0.1683,
      "step": 3519
    },
    {
      "epoch": 1.32,
      "learning_rate": 6.177718294320178e-06,
      "loss": 0.1542,
      "step": 3520
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.171648343428147e-06,
      "loss": 0.1584,
      "step": 3521
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.16558004484946e-06,
      "loss": 0.1462,
      "step": 3522
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.159513401203188e-06,
      "loss": 0.1232,
      "step": 3523
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.153448415107684e-06,
      "loss": 0.1762,
      "step": 3524
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.1473850891805995e-06,
      "loss": 0.1377,
      "step": 3525
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.141323426038857e-06,
      "loss": 0.1444,
      "step": 3526
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.1352634282986615e-06,
      "loss": 0.1227,
      "step": 3527
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.12920509857551e-06,
      "loss": 0.1563,
      "step": 3528
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.123148439484168e-06,
      "loss": 0.1289,
      "step": 3529
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.117093453638685e-06,
      "loss": 0.1418,
      "step": 3530
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.111040143652388e-06,
      "loss": 0.1395,
      "step": 3531
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.104988512137877e-06,
      "loss": 0.1531,
      "step": 3532
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.098938561707039e-06,
      "loss": 0.1319,
      "step": 3533
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.092890294971024e-06,
      "loss": 0.126,
      "step": 3534
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.086843714540256e-06,
      "loss": 0.1306,
      "step": 3535
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.0807988230244385e-06,
      "loss": 0.1237,
      "step": 3536
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.074755623032536e-06,
      "loss": 0.126,
      "step": 3537
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.068714117172793e-06,
      "loss": 0.1738,
      "step": 3538
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.0626743080527185e-06,
      "loss": 0.1574,
      "step": 3539
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.056636198279085e-06,
      "loss": 0.1691,
      "step": 3540
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.05059979045794e-06,
      "loss": 0.1444,
      "step": 3541
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.0445650871945895e-06,
      "loss": 0.1597,
      "step": 3542
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.038532091093605e-06,
      "loss": 0.1407,
      "step": 3543
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.032500804758827e-06,
      "loss": 0.1375,
      "step": 3544
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.026471230793351e-06,
      "loss": 0.1457,
      "step": 3545
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.020443371799534e-06,
      "loss": 0.1597,
      "step": 3546
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.0144172303789935e-06,
      "loss": 0.1243,
      "step": 3547
    },
    {
      "epoch": 1.34,
      "learning_rate": 6.008392809132616e-06,
      "loss": 0.163,
      "step": 3548
    },
    {
      "epoch": 1.34,
      "learning_rate": 6.00237011066053e-06,
      "loss": 0.1327,
      "step": 3549
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.996349137562128e-06,
      "loss": 0.128,
      "step": 3550
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.990329892436057e-06,
      "loss": 0.1348,
      "step": 3551
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.98431237788022e-06,
      "loss": 0.1516,
      "step": 3552
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.978296596491765e-06,
      "loss": 0.1449,
      "step": 3553
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.972282550867106e-06,
      "loss": 0.1476,
      "step": 3554
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.966270243601896e-06,
      "loss": 0.1589,
      "step": 3555
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.96025967729104e-06,
      "loss": 0.1407,
      "step": 3556
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.954250854528698e-06,
      "loss": 0.1752,
      "step": 3557
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.948243777908272e-06,
      "loss": 0.1158,
      "step": 3558
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.942238450022405e-06,
      "loss": 0.1506,
      "step": 3559
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.936234873462999e-06,
      "loss": 0.1493,
      "step": 3560
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.930233050821191e-06,
      "loss": 0.1526,
      "step": 3561
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.924232984687355e-06,
      "loss": 0.1381,
      "step": 3562
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.918234677651124e-06,
      "loss": 0.1722,
      "step": 3563
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.912238132301361e-06,
      "loss": 0.1461,
      "step": 3564
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.906243351226168e-06,
      "loss": 0.1345,
      "step": 3565
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.900250337012885e-06,
      "loss": 0.1618,
      "step": 3566
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.894259092248096e-06,
      "loss": 0.1662,
      "step": 3567
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.888269619517615e-06,
      "loss": 0.1559,
      "step": 3568
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.8822819214064915e-06,
      "loss": 0.1584,
      "step": 3569
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.876296000499017e-06,
      "loss": 0.1475,
      "step": 3570
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.870311859378705e-06,
      "loss": 0.181,
      "step": 3571
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.864329500628305e-06,
      "loss": 0.1676,
      "step": 3572
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.858348926829802e-06,
      "loss": 0.1521,
      "step": 3573
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.852370140564403e-06,
      "loss": 0.1499,
      "step": 3574
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.8463931444125476e-06,
      "loss": 0.1747,
      "step": 3575
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.840417940953902e-06,
      "loss": 0.1544,
      "step": 3576
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.834444532767365e-06,
      "loss": 0.1275,
      "step": 3577
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.828472922431047e-06,
      "loss": 0.1511,
      "step": 3578
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.8225031125222915e-06,
      "loss": 0.155,
      "step": 3579
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.816535105617669e-06,
      "loss": 0.1412,
      "step": 3580
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.810568904292958e-06,
      "loss": 0.1543,
      "step": 3581
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.80460451112317e-06,
      "loss": 0.1419,
      "step": 3582
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.798641928682536e-06,
      "loss": 0.1665,
      "step": 3583
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.7926811595444935e-06,
      "loss": 0.1375,
      "step": 3584
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.786722206281707e-06,
      "loss": 0.1479,
      "step": 3585
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.7807650714660644e-06,
      "loss": 0.1396,
      "step": 3586
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.774809757668649e-06,
      "loss": 0.1529,
      "step": 3587
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.768856267459774e-06,
      "loss": 0.1643,
      "step": 3588
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.7629046034089655e-06,
      "loss": 0.138,
      "step": 3589
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.756954768084947e-06,
      "loss": 0.1577,
      "step": 3590
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.751006764055664e-06,
      "loss": 0.1542,
      "step": 3591
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.74506059388828e-06,
      "loss": 0.1303,
      "step": 3592
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.739116260149147e-06,
      "loss": 0.1615,
      "step": 3593
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.7331737654038365e-06,
      "loss": 0.157,
      "step": 3594
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.72723311221713e-06,
      "loss": 0.1449,
      "step": 3595
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.721294303153e-06,
      "loss": 0.196,
      "step": 3596
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.715357340774636e-06,
      "loss": 0.1231,
      "step": 3597
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.709422227644428e-06,
      "loss": 0.147,
      "step": 3598
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.70348896632396e-06,
      "loss": 0.1411,
      "step": 3599
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.697557559374025e-06,
      "loss": 0.1534,
      "step": 3600
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.691628009354618e-06,
      "loss": 0.16,
      "step": 3601
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.6857003188249204e-06,
      "loss": 0.133,
      "step": 3602
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.679774490343321e-06,
      "loss": 0.1437,
      "step": 3603
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.673850526467406e-06,
      "loss": 0.1667,
      "step": 3604
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.667928429753948e-06,
      "loss": 0.1541,
      "step": 3605
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.6620082027589196e-06,
      "loss": 0.156,
      "step": 3606
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.656089848037489e-06,
      "loss": 0.1221,
      "step": 3607
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.65017336814401e-06,
      "loss": 0.1521,
      "step": 3608
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.644258765632035e-06,
      "loss": 0.1386,
      "step": 3609
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.638346043054297e-06,
      "loss": 0.1425,
      "step": 3610
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.632435202962721e-06,
      "loss": 0.126,
      "step": 3611
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.626526247908428e-06,
      "loss": 0.157,
      "step": 3612
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.620619180441709e-06,
      "loss": 0.149,
      "step": 3613
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.614714003112051e-06,
      "loss": 0.144,
      "step": 3614
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.608810718468132e-06,
      "loss": 0.1638,
      "step": 3615
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.602909329057794e-06,
      "loss": 0.1483,
      "step": 3616
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.597009837428076e-06,
      "loss": 0.1465,
      "step": 3617
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.591112246125198e-06,
      "loss": 0.1385,
      "step": 3618
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.585216557694546e-06,
      "loss": 0.132,
      "step": 3619
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.579322774680697e-06,
      "loss": 0.1532,
      "step": 3620
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.573430899627412e-06,
      "loss": 0.1436,
      "step": 3621
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.567540935077609e-06,
      "loss": 0.1286,
      "step": 3622
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.561652883573394e-06,
      "loss": 0.1434,
      "step": 3623
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.555766747656051e-06,
      "loss": 0.1323,
      "step": 3624
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.549882529866022e-06,
      "loss": 0.1532,
      "step": 3625
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.544000232742934e-06,
      "loss": 0.1398,
      "step": 3626
    },
    {
      "epoch": 1.36,
      "learning_rate": 5.538119858825588e-06,
      "loss": 0.145,
      "step": 3627
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.532241410651937e-06,
      "loss": 0.1645,
      "step": 3628
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.5263648907591216e-06,
      "loss": 0.1481,
      "step": 3629
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.520490301683446e-06,
      "loss": 0.1456,
      "step": 3630
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.514617645960368e-06,
      "loss": 0.1204,
      "step": 3631
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.5087469261245245e-06,
      "loss": 0.1482,
      "step": 3632
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.50287814470972e-06,
      "loss": 0.1669,
      "step": 3633
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.497011304248907e-06,
      "loss": 0.159,
      "step": 3634
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.49114640727421e-06,
      "loss": 0.1436,
      "step": 3635
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.485283456316919e-06,
      "loss": 0.1436,
      "step": 3636
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.479422453907473e-06,
      "loss": 0.1308,
      "step": 3637
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.473563402575484e-06,
      "loss": 0.1376,
      "step": 3638
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.467706304849701e-06,
      "loss": 0.1385,
      "step": 3639
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.461851163258052e-06,
      "loss": 0.1694,
      "step": 3640
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.45599798032761e-06,
      "loss": 0.1592,
      "step": 3641
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.450146758584599e-06,
      "loss": 0.1286,
      "step": 3642
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.4442975005544055e-06,
      "loss": 0.1443,
      "step": 3643
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.438450208761565e-06,
      "loss": 0.1547,
      "step": 3644
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.4326048857297605e-06,
      "loss": 0.1723,
      "step": 3645
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.426761533981829e-06,
      "loss": 0.1423,
      "step": 3646
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.420920156039761e-06,
      "loss": 0.1488,
      "step": 3647
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.415080754424684e-06,
      "loss": 0.1474,
      "step": 3648
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.4092433316568795e-06,
      "loss": 0.1342,
      "step": 3649
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.403407890255776e-06,
      "loss": 0.129,
      "step": 3650
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.397574432739947e-06,
      "loss": 0.1871,
      "step": 3651
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.391742961627109e-06,
      "loss": 0.1621,
      "step": 3652
    },
    {
      "epoch": 1.37,
      "learning_rate": 5.385913479434112e-06,
      "loss": 0.1482,
      "step": 3653
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.3800859886769605e-06,
      "loss": 0.1319,
      "step": 3654
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.374260491870799e-06,
      "loss": 0.1448,
      "step": 3655
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.3684369915299e-06,
      "loss": 0.1524,
      "step": 3656
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.362615490167682e-06,
      "loss": 0.1388,
      "step": 3657
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.356795990296708e-06,
      "loss": 0.1738,
      "step": 3658
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.350978494428658e-06,
      "loss": 0.1504,
      "step": 3659
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.345163005074363e-06,
      "loss": 0.1556,
      "step": 3660
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.3393495247437885e-06,
      "loss": 0.1715,
      "step": 3661
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.333538055946019e-06,
      "loss": 0.1446,
      "step": 3662
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.3277286011892845e-06,
      "loss": 0.1683,
      "step": 3663
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.32192116298094e-06,
      "loss": 0.1277,
      "step": 3664
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.3161157438274706e-06,
      "loss": 0.1546,
      "step": 3665
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.310312346234489e-06,
      "loss": 0.1521,
      "step": 3666
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.3045109727067425e-06,
      "loss": 0.1524,
      "step": 3667
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.298711625748092e-06,
      "loss": 0.127,
      "step": 3668
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.292914307861532e-06,
      "loss": 0.1696,
      "step": 3669
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.2871190215491865e-06,
      "loss": 0.134,
      "step": 3670
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.281325769312286e-06,
      "loss": 0.1096,
      "step": 3671
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.275534553651201e-06,
      "loss": 0.1508,
      "step": 3672
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.269745377065414e-06,
      "loss": 0.1526,
      "step": 3673
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.263958242053524e-06,
      "loss": 0.1283,
      "step": 3674
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.258173151113257e-06,
      "loss": 0.1473,
      "step": 3675
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.252390106741458e-06,
      "loss": 0.1465,
      "step": 3676
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.246609111434076e-06,
      "loss": 0.1546,
      "step": 3677
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.240830167686186e-06,
      "loss": 0.1203,
      "step": 3678
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.235053277991976e-06,
      "loss": 0.1545,
      "step": 3679
    },
    {
      "epoch": 1.38,
      "learning_rate": 5.229278444844748e-06,
      "loss": 0.1529,
      "step": 3680
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.223505670736917e-06,
      "loss": 0.146,
      "step": 3681
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.217734958160002e-06,
      "loss": 0.159,
      "step": 3682
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.21196630960464e-06,
      "loss": 0.1509,
      "step": 3683
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.206199727560579e-06,
      "loss": 0.1212,
      "step": 3684
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.200435214516663e-06,
      "loss": 0.1445,
      "step": 3685
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.194672772960856e-06,
      "loss": 0.1282,
      "step": 3686
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.188912405380224e-06,
      "loss": 0.1614,
      "step": 3687
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.183154114260932e-06,
      "loss": 0.1537,
      "step": 3688
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.177397902088256e-06,
      "loss": 0.1652,
      "step": 3689
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.171643771346577e-06,
      "loss": 0.1488,
      "step": 3690
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.165891724519364e-06,
      "loss": 0.1384,
      "step": 3691
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.1601417640892004e-06,
      "loss": 0.1672,
      "step": 3692
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.154393892537768e-06,
      "loss": 0.1515,
      "step": 3693
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.148648112345831e-06,
      "loss": 0.1426,
      "step": 3694
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.142904425993277e-06,
      "loss": 0.1507,
      "step": 3695
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.137162835959073e-06,
      "loss": 0.1371,
      "step": 3696
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.13142334472128e-06,
      "loss": 0.1499,
      "step": 3697
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.125685954757058e-06,
      "loss": 0.15,
      "step": 3698
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.1199506685426655e-06,
      "loss": 0.1493,
      "step": 3699
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.114217488553441e-06,
      "loss": 0.1319,
      "step": 3700
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.108486417263821e-06,
      "loss": 0.1287,
      "step": 3701
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.102757457147336e-06,
      "loss": 0.1768,
      "step": 3702
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.097030610676593e-06,
      "loss": 0.1209,
      "step": 3703
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.091305880323298e-06,
      "loss": 0.1641,
      "step": 3704
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.085583268558243e-06,
      "loss": 0.1492,
      "step": 3705
    },
    {
      "epoch": 1.39,
      "learning_rate": 5.0798627778512955e-06,
      "loss": 0.1236,
      "step": 3706
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.074144410671418e-06,
      "loss": 0.1439,
      "step": 3707
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.068428169486652e-06,
      "loss": 0.1429,
      "step": 3708
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.062714056764123e-06,
      "loss": 0.1412,
      "step": 3709
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.057002074970042e-06,
      "loss": 0.1405,
      "step": 3710
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.051292226569687e-06,
      "loss": 0.1453,
      "step": 3711
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.045584514027429e-06,
      "loss": 0.143,
      "step": 3712
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.039878939806711e-06,
      "loss": 0.1585,
      "step": 3713
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.0341755063700525e-06,
      "loss": 0.1495,
      "step": 3714
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.028474216179051e-06,
      "loss": 0.118,
      "step": 3715
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.0227750716943835e-06,
      "loss": 0.1239,
      "step": 3716
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.017078075375788e-06,
      "loss": 0.1575,
      "step": 3717
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.011383229682085e-06,
      "loss": 0.1526,
      "step": 3718
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.005690537071173e-06,
      "loss": 0.1426,
      "step": 3719
    },
    {
      "epoch": 1.4,
      "learning_rate": 5.000000000000003e-06,
      "loss": 0.1436,
      "step": 3720
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.994311620924609e-06,
      "loss": 0.147,
      "step": 3721
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.988625402300097e-06,
      "loss": 0.1316,
      "step": 3722
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.982941346580622e-06,
      "loss": 0.1374,
      "step": 3723
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.977259456219427e-06,
      "loss": 0.1362,
      "step": 3724
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.971579733668815e-06,
      "loss": 0.1524,
      "step": 3725
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.965902181380137e-06,
      "loss": 0.1651,
      "step": 3726
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.960226801803827e-06,
      "loss": 0.1466,
      "step": 3727
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.954553597389378e-06,
      "loss": 0.1547,
      "step": 3728
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.9488825705853305e-06,
      "loss": 0.1536,
      "step": 3729
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.9432137238393e-06,
      "loss": 0.1427,
      "step": 3730
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.937547059597959e-06,
      "loss": 0.1607,
      "step": 3731
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.931882580307028e-06,
      "loss": 0.1413,
      "step": 3732
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.926220288411295e-06,
      "loss": 0.1317,
      "step": 3733
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.9205601863546046e-06,
      "loss": 0.147,
      "step": 3734
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.914902276579845e-06,
      "loss": 0.1682,
      "step": 3735
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.909246561528967e-06,
      "loss": 0.1627,
      "step": 3736
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.903593043642979e-06,
      "loss": 0.1386,
      "step": 3737
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.89794172536192e-06,
      "loss": 0.1449,
      "step": 3738
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.892292609124912e-06,
      "loss": 0.1708,
      "step": 3739
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.8866456973700975e-06,
      "loss": 0.1355,
      "step": 3740
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.88100099253468e-06,
      "loss": 0.1737,
      "step": 3741
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.875358497054915e-06,
      "loss": 0.1645,
      "step": 3742
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.869718213366093e-06,
      "loss": 0.1438,
      "step": 3743
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.864080143902555e-06,
      "loss": 0.1324,
      "step": 3744
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.858444291097694e-06,
      "loss": 0.1207,
      "step": 3745
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.852810657383931e-06,
      "loss": 0.1415,
      "step": 3746
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.8471792451927414e-06,
      "loss": 0.1382,
      "step": 3747
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.841550056954641e-06,
      "loss": 0.144,
      "step": 3748
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.8359230950991776e-06,
      "loss": 0.1371,
      "step": 3749
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.8302983620549436e-06,
      "loss": 0.1374,
      "step": 3750
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.824675860249575e-06,
      "loss": 0.1639,
      "step": 3751
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.819055592109726e-06,
      "loss": 0.1234,
      "step": 3752
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.813437560061116e-06,
      "loss": 0.1484,
      "step": 3753
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.807821766528471e-06,
      "loss": 0.1992,
      "step": 3754
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.802208213935568e-06,
      "loss": 0.1511,
      "step": 3755
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.796596904705212e-06,
      "loss": 0.1173,
      "step": 3756
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.790987841259235e-06,
      "loss": 0.1409,
      "step": 3757
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.7853810260185055e-06,
      "loss": 0.1258,
      "step": 3758
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.779776461402926e-06,
      "loss": 0.15,
      "step": 3759
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.774174149831413e-06,
      "loss": 0.1579,
      "step": 3760
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.768574093721923e-06,
      "loss": 0.1205,
      "step": 3761
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.76297629549144e-06,
      "loss": 0.14,
      "step": 3762
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.757380757555963e-06,
      "loss": 0.1663,
      "step": 3763
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.751787482330521e-06,
      "loss": 0.1427,
      "step": 3764
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.746196472229171e-06,
      "loss": 0.155,
      "step": 3765
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.74060772966499e-06,
      "loss": 0.1488,
      "step": 3766
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.735021257050063e-06,
      "loss": 0.1508,
      "step": 3767
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.729437056795523e-06,
      "loss": 0.1329,
      "step": 3768
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.7238551313114944e-06,
      "loss": 0.1441,
      "step": 3769
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.718275483007136e-06,
      "loss": 0.1292,
      "step": 3770
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.712698114290622e-06,
      "loss": 0.1553,
      "step": 3771
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.707123027569132e-06,
      "loss": 0.157,
      "step": 3772
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.701550225248873e-06,
      "loss": 0.1698,
      "step": 3773
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.695979709735067e-06,
      "loss": 0.1561,
      "step": 3774
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.6904114834319346e-06,
      "loss": 0.1487,
      "step": 3775
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.684845548742723e-06,
      "loss": 0.1296,
      "step": 3776
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.679281908069686e-06,
      "loss": 0.1651,
      "step": 3777
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.67372056381408e-06,
      "loss": 0.1339,
      "step": 3778
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.668161518376182e-06,
      "loss": 0.1409,
      "step": 3779
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.662604774155273e-06,
      "loss": 0.1557,
      "step": 3780
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.657050333549631e-06,
      "loss": 0.163,
      "step": 3781
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.651498198956554e-06,
      "loss": 0.1575,
      "step": 3782
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.645948372772338e-06,
      "loss": 0.1547,
      "step": 3783
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.640400857392281e-06,
      "loss": 0.15,
      "step": 3784
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.634855655210692e-06,
      "loss": 0.1474,
      "step": 3785
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.629312768620867e-06,
      "loss": 0.1674,
      "step": 3786
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.623772200015117e-06,
      "loss": 0.1509,
      "step": 3787
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.6182339517847465e-06,
      "loss": 0.1318,
      "step": 3788
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.612698026320055e-06,
      "loss": 0.1204,
      "step": 3789
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.607164426010346e-06,
      "loss": 0.1519,
      "step": 3790
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.601633153243919e-06,
      "loss": 0.1358,
      "step": 3791
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.59610421040806e-06,
      "loss": 0.1409,
      "step": 3792
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.590577599889061e-06,
      "loss": 0.144,
      "step": 3793
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.585053324072205e-06,
      "loss": 0.1174,
      "step": 3794
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.579531385341758e-06,
      "loss": 0.1528,
      "step": 3795
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.574011786080983e-06,
      "loss": 0.1341,
      "step": 3796
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.568494528672146e-06,
      "loss": 0.1498,
      "step": 3797
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.56297961549648e-06,
      "loss": 0.145,
      "step": 3798
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.55746704893422e-06,
      "loss": 0.1557,
      "step": 3799
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.551956831364589e-06,
      "loss": 0.1613,
      "step": 3800
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.546448965165785e-06,
      "loss": 0.1497,
      "step": 3801
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.540943452715002e-06,
      "loss": 0.1755,
      "step": 3802
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.535440296388418e-06,
      "loss": 0.138,
      "step": 3803
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.529939498561183e-06,
      "loss": 0.1727,
      "step": 3804
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.524441061607442e-06,
      "loss": 0.1415,
      "step": 3805
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.518944987900318e-06,
      "loss": 0.1471,
      "step": 3806
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.513451279811903e-06,
      "loss": 0.1611,
      "step": 3807
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.507959939713283e-06,
      "loss": 0.1293,
      "step": 3808
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.50247096997452e-06,
      "loss": 0.1384,
      "step": 3809
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.4969843729646375e-06,
      "loss": 0.1583,
      "step": 3810
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.491500151051654e-06,
      "loss": 0.154,
      "step": 3811
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.486018306602552e-06,
      "loss": 0.1487,
      "step": 3812
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.480538841983293e-06,
      "loss": 0.1625,
      "step": 3813
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.475061759558812e-06,
      "loss": 0.1523,
      "step": 3814
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.469587061693006e-06,
      "loss": 0.1371,
      "step": 3815
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.464114750748752e-06,
      "loss": 0.1561,
      "step": 3816
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.458644829087902e-06,
      "loss": 0.1428,
      "step": 3817
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.45317729907126e-06,
      "loss": 0.145,
      "step": 3818
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.44771216305861e-06,
      "loss": 0.1615,
      "step": 3819
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.442249423408707e-06,
      "loss": 0.1747,
      "step": 3820
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.436789082479253e-06,
      "loss": 0.1381,
      "step": 3821
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.431331142626936e-06,
      "loss": 0.168,
      "step": 3822
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.425875606207396e-06,
      "loss": 0.1355,
      "step": 3823
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.420422475575233e-06,
      "loss": 0.1355,
      "step": 3824
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.4149717530840185e-06,
      "loss": 0.1735,
      "step": 3825
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.4095234410862785e-06,
      "loss": 0.1489,
      "step": 3826
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.4040775419335e-06,
      "loss": 0.1359,
      "step": 3827
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.398634057976129e-06,
      "loss": 0.1438,
      "step": 3828
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.393192991563571e-06,
      "loss": 0.1143,
      "step": 3829
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.387754345044179e-06,
      "loss": 0.1548,
      "step": 3830
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.382318120765272e-06,
      "loss": 0.1396,
      "step": 3831
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.376884321073124e-06,
      "loss": 0.1356,
      "step": 3832
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.371452948312951e-06,
      "loss": 0.1303,
      "step": 3833
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.366024004828931e-06,
      "loss": 0.1571,
      "step": 3834
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.360597492964196e-06,
      "loss": 0.1682,
      "step": 3835
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.355173415060817e-06,
      "loss": 0.1522,
      "step": 3836
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.349751773459825e-06,
      "loss": 0.1546,
      "step": 3837
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.3443325705012e-06,
      "loss": 0.1537,
      "step": 3838
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.338915808523856e-06,
      "loss": 0.1585,
      "step": 3839
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.333501489865669e-06,
      "loss": 0.1787,
      "step": 3840
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.328089616863453e-06,
      "loss": 0.1393,
      "step": 3841
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.322680191852969e-06,
      "loss": 0.1657,
      "step": 3842
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.317273217168923e-06,
      "loss": 0.137,
      "step": 3843
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.3118686951449525e-06,
      "loss": 0.1502,
      "step": 3844
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.306466628113651e-06,
      "loss": 0.1447,
      "step": 3845
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.301067018406548e-06,
      "loss": 0.1621,
      "step": 3846
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.295669868354102e-06,
      "loss": 0.1333,
      "step": 3847
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.290275180285725e-06,
      "loss": 0.1488,
      "step": 3848
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.28488295652976e-06,
      "loss": 0.1612,
      "step": 3849
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.279493199413482e-06,
      "loss": 0.1664,
      "step": 3850
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.274105911263108e-06,
      "loss": 0.1361,
      "step": 3851
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.268721094403791e-06,
      "loss": 0.1327,
      "step": 3852
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.263338751159604e-06,
      "loss": 0.1485,
      "step": 3853
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2579588838535676e-06,
      "loss": 0.1518,
      "step": 3854
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2525814948076296e-06,
      "loss": 0.1608,
      "step": 3855
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.247206586342663e-06,
      "loss": 0.1612,
      "step": 3856
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.241834160778478e-06,
      "loss": 0.1625,
      "step": 3857
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.236464220433805e-06,
      "loss": 0.1281,
      "step": 3858
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.231096767626306e-06,
      "loss": 0.1444,
      "step": 3859
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.225731804672574e-06,
      "loss": 0.1379,
      "step": 3860
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.220369333888114e-06,
      "loss": 0.1409,
      "step": 3861
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2150093575873685e-06,
      "loss": 0.1258,
      "step": 3862
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.209651878083703e-06,
      "loss": 0.1369,
      "step": 3863
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.204296897689393e-06,
      "loss": 0.1185,
      "step": 3864
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.198944418715647e-06,
      "loss": 0.1491,
      "step": 3865
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.19359444347259e-06,
      "loss": 0.1424,
      "step": 3866
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.188246974269273e-06,
      "loss": 0.1711,
      "step": 3867
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.182902013413649e-06,
      "loss": 0.1318,
      "step": 3868
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.177559563212604e-06,
      "loss": 0.1487,
      "step": 3869
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.172219625971934e-06,
      "loss": 0.152,
      "step": 3870
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.166882203996351e-06,
      "loss": 0.155,
      "step": 3871
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.161547299589487e-06,
      "loss": 0.1724,
      "step": 3872
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.156214915053873e-06,
      "loss": 0.1238,
      "step": 3873
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.150885052690968e-06,
      "loss": 0.1517,
      "step": 3874
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.145557714801138e-06,
      "loss": 0.1581,
      "step": 3875
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.140232903683649e-06,
      "loss": 0.1795,
      "step": 3876
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.13491062163669e-06,
      "loss": 0.154,
      "step": 3877
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.1295908709573585e-06,
      "loss": 0.112,
      "step": 3878
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.124273653941645e-06,
      "loss": 0.1446,
      "step": 3879
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.1189589728844605e-06,
      "loss": 0.1507,
      "step": 3880
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.113646830079619e-06,
      "loss": 0.1259,
      "step": 3881
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.108337227819833e-06,
      "loss": 0.1365,
      "step": 3882
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.103030168396722e-06,
      "loss": 0.1278,
      "step": 3883
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.097725654100811e-06,
      "loss": 0.1654,
      "step": 3884
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.092423687221523e-06,
      "loss": 0.1419,
      "step": 3885
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.087124270047186e-06,
      "loss": 0.1431,
      "step": 3886
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0818274048650175e-06,
      "loss": 0.1217,
      "step": 3887
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.076533093961141e-06,
      "loss": 0.1274,
      "step": 3888
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.071241339620584e-06,
      "loss": 0.137,
      "step": 3889
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.065952144127255e-06,
      "loss": 0.1433,
      "step": 3890
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.060665509763967e-06,
      "loss": 0.1333,
      "step": 3891
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.055381438812433e-06,
      "loss": 0.148,
      "step": 3892
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.050099933553246e-06,
      "loss": 0.1475,
      "step": 3893
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.044820996265905e-06,
      "loss": 0.139,
      "step": 3894
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.039544629228793e-06,
      "loss": 0.1476,
      "step": 3895
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.034270834719184e-06,
      "loss": 0.1385,
      "step": 3896
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.028999615013244e-06,
      "loss": 0.1372,
      "step": 3897
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.02373097238603e-06,
      "loss": 0.1432,
      "step": 3898
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.018464909111476e-06,
      "loss": 0.1642,
      "step": 3899
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0132014274624175e-06,
      "loss": 0.1483,
      "step": 3900
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.007940529710571e-06,
      "loss": 0.1552,
      "step": 3901
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0026822181265276e-06,
      "loss": 0.1341,
      "step": 3902
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.997426494979775e-06,
      "loss": 0.1677,
      "step": 3903
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.99217336253868e-06,
      "loss": 0.1369,
      "step": 3904
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.986922823070485e-06,
      "loss": 0.1261,
      "step": 3905
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.9816748788413215e-06,
      "loss": 0.142,
      "step": 3906
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.976429532116201e-06,
      "loss": 0.1639,
      "step": 3907
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.971186785159003e-06,
      "loss": 0.1417,
      "step": 3908
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.965946640232499e-06,
      "loss": 0.1206,
      "step": 3909
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.960709099598331e-06,
      "loss": 0.1348,
      "step": 3910
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.955474165517013e-06,
      "loss": 0.1284,
      "step": 3911
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.950241840247939e-06,
      "loss": 0.1234,
      "step": 3912
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.945012126049381e-06,
      "loss": 0.1405,
      "step": 3913
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.93978502517847e-06,
      "loss": 0.1248,
      "step": 3914
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.93456053989123e-06,
      "loss": 0.1357,
      "step": 3915
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.929338672442536e-06,
      "loss": 0.1322,
      "step": 3916
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.924119425086144e-06,
      "loss": 0.1614,
      "step": 3917
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.91890280007468e-06,
      "loss": 0.1447,
      "step": 3918
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.913688799659629e-06,
      "loss": 0.1598,
      "step": 3919
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.908477426091351e-06,
      "loss": 0.1477,
      "step": 3920
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.903268681619076e-06,
      "loss": 0.1707,
      "step": 3921
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.898062568490885e-06,
      "loss": 0.1716,
      "step": 3922
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.892859088953736e-06,
      "loss": 0.124,
      "step": 3923
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.887658245253449e-06,
      "loss": 0.1497,
      "step": 3924
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.8824600396347e-06,
      "loss": 0.1205,
      "step": 3925
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.8772644743410305e-06,
      "loss": 0.1305,
      "step": 3926
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.872071551614846e-06,
      "loss": 0.1447,
      "step": 3927
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.866881273697398e-06,
      "loss": 0.1392,
      "step": 3928
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.861693642828817e-06,
      "loss": 0.1456,
      "step": 3929
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.856508661248078e-06,
      "loss": 0.1439,
      "step": 3930
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.851326331193011e-06,
      "loss": 0.1206,
      "step": 3931
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.846146654900305e-06,
      "loss": 0.1233,
      "step": 3932
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.840969634605509e-06,
      "loss": 0.1864,
      "step": 3933
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.835795272543014e-06,
      "loss": 0.1617,
      "step": 3934
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.830623570946074e-06,
      "loss": 0.1508,
      "step": 3935
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.8254545320467925e-06,
      "loss": 0.1388,
      "step": 3936
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.820288158076119e-06,
      "loss": 0.1362,
      "step": 3937
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.815124451263857e-06,
      "loss": 0.145,
      "step": 3938
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.809963413838661e-06,
      "loss": 0.1233,
      "step": 3939
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.804805048028025e-06,
      "loss": 0.1853,
      "step": 3940
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.799649356058299e-06,
      "loss": 0.173,
      "step": 3941
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.7944963401546776e-06,
      "loss": 0.1576,
      "step": 3942
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.7893460025411877e-06,
      "loss": 0.1315,
      "step": 3943
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.784198345440725e-06,
      "loss": 0.135,
      "step": 3944
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.7790533710750034e-06,
      "loss": 0.1284,
      "step": 3945
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.773911081664594e-06,
      "loss": 0.1663,
      "step": 3946
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7687714794289064e-06,
      "loss": 0.1237,
      "step": 3947
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7636345665861816e-06,
      "loss": 0.1328,
      "step": 3948
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7585003453535117e-06,
      "loss": 0.1477,
      "step": 3949
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7533688179468263e-06,
      "loss": 0.1685,
      "step": 3950
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.748239986580878e-06,
      "loss": 0.1558,
      "step": 3951
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7431138534692736e-06,
      "loss": 0.1487,
      "step": 3952
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.737990420824449e-06,
      "loss": 0.1376,
      "step": 3953
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7328696908576656e-06,
      "loss": 0.1574,
      "step": 3954
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7277516657790323e-06,
      "loss": 0.1385,
      "step": 3955
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7226363477974857e-06,
      "loss": 0.1375,
      "step": 3956
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7175237391207863e-06,
      "loss": 0.1542,
      "step": 3957
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7124138419555357e-06,
      "loss": 0.1544,
      "step": 3958
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7073066585071617e-06,
      "loss": 0.1547,
      "step": 3959
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.7022021909799176e-06,
      "loss": 0.1627,
      "step": 3960
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.697100441576894e-06,
      "loss": 0.1363,
      "step": 3961
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.692001412499994e-06,
      "loss": 0.1304,
      "step": 3962
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.686905105949956e-06,
      "loss": 0.1429,
      "step": 3963
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.6818115241263463e-06,
      "loss": 0.1341,
      "step": 3964
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.6767206692275438e-06,
      "loss": 0.1341,
      "step": 3965
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.6716325434507593e-06,
      "loss": 0.1401,
      "step": 3966
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.666547148992026e-06,
      "loss": 0.1425,
      "step": 3967
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.661464488046197e-06,
      "loss": 0.1236,
      "step": 3968
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.656384562806938e-06,
      "loss": 0.1428,
      "step": 3969
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.651307375466745e-06,
      "loss": 0.1442,
      "step": 3970
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.6462329282169297e-06,
      "loss": 0.1298,
      "step": 3971
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.641161223247611e-06,
      "loss": 0.135,
      "step": 3972
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.6360922627477445e-06,
      "loss": 0.1547,
      "step": 3973
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.6310260489050785e-06,
      "loss": 0.1269,
      "step": 3974
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.625962583906193e-06,
      "loss": 0.13,
      "step": 3975
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.6209018699364752e-06,
      "loss": 0.1513,
      "step": 3976
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.61584390918012e-06,
      "loss": 0.151,
      "step": 3977
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.6107887038201418e-06,
      "loss": 0.1389,
      "step": 3978
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.605736256038367e-06,
      "loss": 0.1496,
      "step": 3979
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.6006865680154203e-06,
      "loss": 0.1334,
      "step": 3980
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.595639641930746e-06,
      "loss": 0.1469,
      "step": 3981
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5905954799625977e-06,
      "loss": 0.145,
      "step": 3982
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5855540842880233e-06,
      "loss": 0.1521,
      "step": 3983
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5805154570828893e-06,
      "loss": 0.1459,
      "step": 3984
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5754796005218663e-06,
      "loss": 0.137,
      "step": 3985
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5704465167784177e-06,
      "loss": 0.1401,
      "step": 3986
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5654162080248233e-06,
      "loss": 0.1463,
      "step": 3987
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.560388676432158e-06,
      "loss": 0.168,
      "step": 3988
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.555363924170303e-06,
      "loss": 0.1454,
      "step": 3989
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5503419534079384e-06,
      "loss": 0.1217,
      "step": 3990
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5453227663125365e-06,
      "loss": 0.1625,
      "step": 3991
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5403063650503776e-06,
      "loss": 0.1276,
      "step": 3992
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.53529275178654e-06,
      "loss": 0.1296,
      "step": 3993
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5302819286848875e-06,
      "loss": 0.1328,
      "step": 3994
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.52527389790809e-06,
      "loss": 0.1538,
      "step": 3995
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.520268661617614e-06,
      "loss": 0.1448,
      "step": 3996
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.515266221973708e-06,
      "loss": 0.1405,
      "step": 3997
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.5102665811354244e-06,
      "loss": 0.156,
      "step": 3998
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.505269741260606e-06,
      "loss": 0.1766,
      "step": 3999
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.50027570450588e-06,
      "loss": 0.1679,
      "step": 4000
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.49528447302667e-06,
      "loss": 0.1454,
      "step": 4001
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4902960489771875e-06,
      "loss": 0.1545,
      "step": 4002
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4853104345104338e-06,
      "loss": 0.1424,
      "step": 4003
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.480327631778194e-06,
      "loss": 0.1418,
      "step": 4004
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.475347642931046e-06,
      "loss": 0.1354,
      "step": 4005
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4703704701183416e-06,
      "loss": 0.1594,
      "step": 4006
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4653961154882276e-06,
      "loss": 0.1475,
      "step": 4007
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.460424581187635e-06,
      "loss": 0.1602,
      "step": 4008
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4554558693622664e-06,
      "loss": 0.1311,
      "step": 4009
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4504899821566186e-06,
      "loss": 0.173,
      "step": 4010
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.445526921713966e-06,
      "loss": 0.1496,
      "step": 4011
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.440566690176356e-06,
      "loss": 0.1466,
      "step": 4012
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4356092896846215e-06,
      "loss": 0.1424,
      "step": 4013
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.430654722378378e-06,
      "loss": 0.134,
      "step": 4014
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.425702990396005e-06,
      "loss": 0.154,
      "step": 4015
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4207540958746686e-06,
      "loss": 0.1544,
      "step": 4016
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.415808040950308e-06,
      "loss": 0.1525,
      "step": 4017
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4108648277576373e-06,
      "loss": 0.1628,
      "step": 4018
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4059244584301444e-06,
      "loss": 0.1529,
      "step": 4019
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.4009869351000836e-06,
      "loss": 0.149,
      "step": 4020
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.396052259898489e-06,
      "loss": 0.1471,
      "step": 4021
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.391120434955164e-06,
      "loss": 0.1514,
      "step": 4022
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.386191462398676e-06,
      "loss": 0.1622,
      "step": 4023
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.381265344356366e-06,
      "loss": 0.1339,
      "step": 4024
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.376342082954348e-06,
      "loss": 0.1564,
      "step": 4025
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3714216803174894e-06,
      "loss": 0.1659,
      "step": 4026
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.366504138569435e-06,
      "loss": 0.167,
      "step": 4027
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.361589459832596e-06,
      "loss": 0.1385,
      "step": 4028
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.356677646228137e-06,
      "loss": 0.1468,
      "step": 4029
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3517686998759946e-06,
      "loss": 0.155,
      "step": 4030
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3468626228948677e-06,
      "loss": 0.1305,
      "step": 4031
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.341959417402213e-06,
      "loss": 0.1365,
      "step": 4032
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3370590855142506e-06,
      "loss": 0.1265,
      "step": 4033
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3321616293459635e-06,
      "loss": 0.1729,
      "step": 4034
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3272670510110838e-06,
      "loss": 0.1365,
      "step": 4035
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.322375352622108e-06,
      "loss": 0.1395,
      "step": 4036
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3174865362902954e-06,
      "loss": 0.1359,
      "step": 4037
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3126006041256454e-06,
      "loss": 0.1626,
      "step": 4038
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.307717558236927e-06,
      "loss": 0.1591,
      "step": 4039
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.3028374007316633e-06,
      "loss": 0.1454,
      "step": 4040
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.297960133716117e-06,
      "loss": 0.1321,
      "step": 4041
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.2930857592953167e-06,
      "loss": 0.1281,
      "step": 4042
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.288214279573043e-06,
      "loss": 0.1656,
      "step": 4043
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.283345696651814e-06,
      "loss": 0.1374,
      "step": 4044
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.2784800126329097e-06,
      "loss": 0.1453,
      "step": 4045
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.273617229616356e-06,
      "loss": 0.1666,
      "step": 4046
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.268757349700925e-06,
      "loss": 0.1154,
      "step": 4047
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.26390037498414e-06,
      "loss": 0.1625,
      "step": 4048
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.2590463075622626e-06,
      "loss": 0.1383,
      "step": 4049
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.2541951495303047e-06,
      "loss": 0.1363,
      "step": 4050
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.2493469029820268e-06,
      "loss": 0.1347,
      "step": 4051
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.2445015700099204e-06,
      "loss": 0.1553,
      "step": 4052
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.239659152705232e-06,
      "loss": 0.1578,
      "step": 4053
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.2348196531579466e-06,
      "loss": 0.1325,
      "step": 4054
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.2299830734567827e-06,
      "loss": 0.1365,
      "step": 4055
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.225149415689207e-06,
      "loss": 0.132,
      "step": 4056
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.2203186819414257e-06,
      "loss": 0.1355,
      "step": 4057
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.2154908742983713e-06,
      "loss": 0.1469,
      "step": 4058
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.210665994843726e-06,
      "loss": 0.1652,
      "step": 4059
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.2058440456599038e-06,
      "loss": 0.1301,
      "step": 4060
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.2010250288280544e-06,
      "loss": 0.1722,
      "step": 4061
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.196208946428062e-06,
      "loss": 0.1288,
      "step": 4062
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1913958005385402e-06,
      "loss": 0.1407,
      "step": 4063
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1865855932368417e-06,
      "loss": 0.1427,
      "step": 4064
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.181778326599049e-06,
      "loss": 0.1664,
      "step": 4065
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1769740026999696e-06,
      "loss": 0.1287,
      "step": 4066
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.172172623613148e-06,
      "loss": 0.1394,
      "step": 4067
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.167374191410856e-06,
      "loss": 0.1294,
      "step": 4068
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1625787081640955e-06,
      "loss": 0.1585,
      "step": 4069
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1577861759425876e-06,
      "loss": 0.1524,
      "step": 4070
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.152996596814788e-06,
      "loss": 0.1344,
      "step": 4071
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1482099728478765e-06,
      "loss": 0.1548,
      "step": 4072
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.143426306107752e-06,
      "loss": 0.1273,
      "step": 4073
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1386455986590436e-06,
      "loss": 0.1248,
      "step": 4074
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.133867852565099e-06,
      "loss": 0.1344,
      "step": 4075
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1290930698879916e-06,
      "loss": 0.1623,
      "step": 4076
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.1243212526885157e-06,
      "loss": 0.1418,
      "step": 4077
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.119552403026177e-06,
      "loss": 0.1545,
      "step": 4078
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.11478652295921e-06,
      "loss": 0.1485,
      "step": 4079
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.1100236145445683e-06,
      "loss": 0.1637,
      "step": 4080
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.105263679837913e-06,
      "loss": 0.1477,
      "step": 4081
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.10050672089363e-06,
      "loss": 0.1122,
      "step": 4082
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0957527397648236e-06,
      "loss": 0.1311,
      "step": 4083
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0910017385032997e-06,
      "loss": 0.125,
      "step": 4084
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.086253719159591e-06,
      "loss": 0.131,
      "step": 4085
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.081508683782942e-06,
      "loss": 0.1502,
      "step": 4086
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0767666344212987e-06,
      "loss": 0.1694,
      "step": 4087
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.072027573121329e-06,
      "loss": 0.1594,
      "step": 4088
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.067291501928411e-06,
      "loss": 0.1485,
      "step": 4089
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.062558422886619e-06,
      "loss": 0.1393,
      "step": 4090
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0578283380387597e-06,
      "loss": 0.1757,
      "step": 4091
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0531012494263233e-06,
      "loss": 0.1281,
      "step": 4092
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.048377159089522e-06,
      "loss": 0.1152,
      "step": 4093
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.04365606906727e-06,
      "loss": 0.1363,
      "step": 4094
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.038937981397181e-06,
      "loss": 0.155,
      "step": 4095
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0342228981155797e-06,
      "loss": 0.1364,
      "step": 4096
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0295108212574965e-06,
      "loss": 0.1348,
      "step": 4097
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.024801752856653e-06,
      "loss": 0.1461,
      "step": 4098
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.020095694945483e-06,
      "loss": 0.1462,
      "step": 4099
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0153926495551177e-06,
      "loss": 0.1606,
      "step": 4100
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.010692618715384e-06,
      "loss": 0.1377,
      "step": 4101
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.0059956044548133e-06,
      "loss": 0.1822,
      "step": 4102
    },
    {
      "epoch": 1.54,
      "learning_rate": 3.001301608800636e-06,
      "loss": 0.127,
      "step": 4103
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.996610633778768e-06,
      "loss": 0.1447,
      "step": 4104
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.9919226814138392e-06,
      "loss": 0.1419,
      "step": 4105
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9872377537291654e-06,
      "loss": 0.1354,
      "step": 4106
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9825558527467514e-06,
      "loss": 0.1727,
      "step": 4107
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9778769804873042e-06,
      "loss": 0.152,
      "step": 4108
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9732011389702253e-06,
      "loss": 0.1489,
      "step": 4109
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.968528330213598e-06,
      "loss": 0.1343,
      "step": 4110
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9638585562342047e-06,
      "loss": 0.1645,
      "step": 4111
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.95919181904752e-06,
      "loss": 0.1354,
      "step": 4112
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9545281206676955e-06,
      "loss": 0.1441,
      "step": 4113
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.949867463107585e-06,
      "loss": 0.1582,
      "step": 4114
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9452098483787274e-06,
      "loss": 0.1432,
      "step": 4115
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.940555278491337e-06,
      "loss": 0.1243,
      "step": 4116
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9359037554543268e-06,
      "loss": 0.1557,
      "step": 4117
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9312552812752926e-06,
      "loss": 0.1394,
      "step": 4118
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.926609857960503e-06,
      "loss": 0.1587,
      "step": 4119
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9219674875149317e-06,
      "loss": 0.1724,
      "step": 4120
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.917328171942212e-06,
      "loss": 0.1661,
      "step": 4121
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9126919132446717e-06,
      "loss": 0.1301,
      "step": 4122
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9080587134233197e-06,
      "loss": 0.1221,
      "step": 4123
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9034285744778334e-06,
      "loss": 0.1476,
      "step": 4124
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.898801498406583e-06,
      "loss": 0.1461,
      "step": 4125
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.89417748720661e-06,
      "loss": 0.1355,
      "step": 4126
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.8895565428736293e-06,
      "loss": 0.1537,
      "step": 4127
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.8849386674020407e-06,
      "loss": 0.1502,
      "step": 4128
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.880323862784917e-06,
      "loss": 0.1387,
      "step": 4129
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.8757121310139992e-06,
      "loss": 0.1507,
      "step": 4130
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.8711034740797084e-06,
      "loss": 0.1346,
      "step": 4131
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.8664978939711397e-06,
      "loss": 0.1444,
      "step": 4132
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.861895392676054e-06,
      "loss": 0.1275,
      "step": 4133
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8572959721808847e-06,
      "loss": 0.1407,
      "step": 4134
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.852699634470747e-06,
      "loss": 0.1232,
      "step": 4135
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8481063815294084e-06,
      "loss": 0.1425,
      "step": 4136
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.843516215339316e-06,
      "loss": 0.1229,
      "step": 4137
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.838929137881584e-06,
      "loss": 0.1219,
      "step": 4138
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8343451511359853e-06,
      "loss": 0.1516,
      "step": 4139
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.829764257080968e-06,
      "loss": 0.1262,
      "step": 4140
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.825186457693645e-06,
      "loss": 0.1375,
      "step": 4141
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8206117549497846e-06,
      "loss": 0.1349,
      "step": 4142
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8160401508238268e-06,
      "loss": 0.129,
      "step": 4143
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8114716472888758e-06,
      "loss": 0.1525,
      "step": 4144
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8069062463166884e-06,
      "loss": 0.139,
      "step": 4145
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.8023439498776904e-06,
      "loss": 0.138,
      "step": 4146
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.797784759940968e-06,
      "loss": 0.1206,
      "step": 4147
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7932286784742544e-06,
      "loss": 0.1652,
      "step": 4148
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.788675707443963e-06,
      "loss": 0.1214,
      "step": 4149
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7841258488151433e-06,
      "loss": 0.1333,
      "step": 4150
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.779579104551513e-06,
      "loss": 0.1602,
      "step": 4151
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7750354766154466e-06,
      "loss": 0.1185,
      "step": 4152
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7704949669679636e-06,
      "loss": 0.1456,
      "step": 4153
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7659575775687462e-06,
      "loss": 0.1239,
      "step": 4154
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7614233103761335e-06,
      "loss": 0.1489,
      "step": 4155
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.756892167347104e-06,
      "loss": 0.1035,
      "step": 4156
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7523641504372967e-06,
      "loss": 0.1443,
      "step": 4157
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.7478392616010052e-06,
      "loss": 0.1571,
      "step": 4158
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.7433175027911608e-06,
      "loss": 0.1234,
      "step": 4159
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.738798875959353e-06,
      "loss": 0.157,
      "step": 4160
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.734283383055822e-06,
      "loss": 0.1284,
      "step": 4161
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.729771026029443e-06,
      "loss": 0.1347,
      "step": 4162
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.725261806827749e-06,
      "loss": 0.1488,
      "step": 4163
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.7207557273969165e-06,
      "loss": 0.1576,
      "step": 4164
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.7162527896817636e-06,
      "loss": 0.1245,
      "step": 4165
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.7117529956257594e-06,
      "loss": 0.151,
      "step": 4166
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.7072563471710033e-06,
      "loss": 0.1458,
      "step": 4167
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.702762846258249e-06,
      "loss": 0.1507,
      "step": 4168
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.69827249482689e-06,
      "loss": 0.1461,
      "step": 4169
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6937852948149534e-06,
      "loss": 0.1345,
      "step": 4170
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.689301248159113e-06,
      "loss": 0.1639,
      "step": 4171
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6848203567946806e-06,
      "loss": 0.1943,
      "step": 4172
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6803426226556063e-06,
      "loss": 0.1411,
      "step": 4173
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6758680476744715e-06,
      "loss": 0.1638,
      "step": 4174
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6713966337825025e-06,
      "loss": 0.1394,
      "step": 4175
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6669283829095593e-06,
      "loss": 0.1445,
      "step": 4176
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.66246329698413e-06,
      "loss": 0.147,
      "step": 4177
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6580013779333456e-06,
      "loss": 0.1417,
      "step": 4178
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.653542627682966e-06,
      "loss": 0.1662,
      "step": 4179
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.649087048157384e-06,
      "loss": 0.1399,
      "step": 4180
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6446346412796266e-06,
      "loss": 0.1692,
      "step": 4181
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.640185408971343e-06,
      "loss": 0.1532,
      "step": 4182
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.635739353152822e-06,
      "loss": 0.1481,
      "step": 4183
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6312964757429793e-06,
      "loss": 0.1525,
      "step": 4184
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.6268567786593524e-06,
      "loss": 0.1296,
      "step": 4185
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.622420263818112e-06,
      "loss": 0.146,
      "step": 4186
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.617986933134059e-06,
      "loss": 0.143,
      "step": 4187
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.613556788520607e-06,
      "loss": 0.1358,
      "step": 4188
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.6091298318898086e-06,
      "loss": 0.1229,
      "step": 4189
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.6047060651523347e-06,
      "loss": 0.1482,
      "step": 4190
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.6002854902174757e-06,
      "loss": 0.1652,
      "step": 4191
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.595868108993149e-06,
      "loss": 0.1242,
      "step": 4192
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5914539233858937e-06,
      "loss": 0.1405,
      "step": 4193
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5870429353008686e-06,
      "loss": 0.1332,
      "step": 4194
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5826351466418543e-06,
      "loss": 0.1469,
      "step": 4195
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5782305593112445e-06,
      "loss": 0.141,
      "step": 4196
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5738291752100564e-06,
      "loss": 0.1649,
      "step": 4197
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5694309962379283e-06,
      "loss": 0.1641,
      "step": 4198
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5650360242931037e-06,
      "loss": 0.1304,
      "step": 4199
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5606442612724524e-06,
      "loss": 0.152,
      "step": 4200
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.556255709071458e-06,
      "loss": 0.1411,
      "step": 4201
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5518703695842105e-06,
      "loss": 0.1382,
      "step": 4202
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.54748824470342e-06,
      "loss": 0.1398,
      "step": 4203
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.543109336320414e-06,
      "loss": 0.1392,
      "step": 4204
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5387336463251165e-06,
      "loss": 0.1562,
      "step": 4205
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5343611766060762e-06,
      "loss": 0.139,
      "step": 4206
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.529991929050447e-06,
      "loss": 0.1575,
      "step": 4207
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5256259055439936e-06,
      "loss": 0.1359,
      "step": 4208
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.521263107971086e-06,
      "loss": 0.11,
      "step": 4209
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5169035382147088e-06,
      "loss": 0.1215,
      "step": 4210
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.512547198156441e-06,
      "loss": 0.1442,
      "step": 4211
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.50819408967648e-06,
      "loss": 0.1655,
      "step": 4212
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.503844214653626e-06,
      "loss": 0.1536,
      "step": 4213
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.499497574965274e-06,
      "loss": 0.1315,
      "step": 4214
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.495154172487435e-06,
      "loss": 0.1543,
      "step": 4215
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4908140090947197e-06,
      "loss": 0.1554,
      "step": 4216
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4864770866603337e-06,
      "loss": 0.1596,
      "step": 4217
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4821434070560913e-06,
      "loss": 0.1356,
      "step": 4218
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.477812972152409e-06,
      "loss": 0.1028,
      "step": 4219
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4734857838182913e-06,
      "loss": 0.1343,
      "step": 4220
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.469161843921354e-06,
      "loss": 0.1404,
      "step": 4221
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.464841154327804e-06,
      "loss": 0.1741,
      "step": 4222
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.460523716902449e-06,
      "loss": 0.1262,
      "step": 4223
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4562095335086933e-06,
      "loss": 0.1389,
      "step": 4224
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.451898606008528e-06,
      "loss": 0.1264,
      "step": 4225
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4475909362625504e-06,
      "loss": 0.1346,
      "step": 4226
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4432865261299497e-06,
      "loss": 0.1546,
      "step": 4227
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.438985377468499e-06,
      "loss": 0.1232,
      "step": 4228
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.434687492134574e-06,
      "loss": 0.1342,
      "step": 4229
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4303928719831415e-06,
      "loss": 0.1343,
      "step": 4230
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4261015188677494e-06,
      "loss": 0.1596,
      "step": 4231
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4218134346405463e-06,
      "loss": 0.142,
      "step": 4232
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.4175286211522674e-06,
      "loss": 0.1508,
      "step": 4233
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.413247080252228e-06,
      "loss": 0.1555,
      "step": 4234
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.408968813788343e-06,
      "loss": 0.1432,
      "step": 4235
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.404693823607106e-06,
      "loss": 0.123,
      "step": 4236
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.400422111553601e-06,
      "loss": 0.1229,
      "step": 4237
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.3961536794714935e-06,
      "loss": 0.1534,
      "step": 4238
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3918885292030403e-06,
      "loss": 0.1268,
      "step": 4239
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3876266625890697e-06,
      "loss": 0.1533,
      "step": 4240
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3833680814690018e-06,
      "loss": 0.1768,
      "step": 4241
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3791127876808406e-06,
      "loss": 0.1531,
      "step": 4242
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3748607830611626e-06,
      "loss": 0.1164,
      "step": 4243
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3706120694451296e-06,
      "loss": 0.1443,
      "step": 4244
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.366366648666487e-06,
      "loss": 0.1396,
      "step": 4245
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3621245225575497e-06,
      "loss": 0.1591,
      "step": 4246
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.357885692949218e-06,
      "loss": 0.143,
      "step": 4247
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3536501616709708e-06,
      "loss": 0.1344,
      "step": 4248
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3494179305508513e-06,
      "loss": 0.1422,
      "step": 4249
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.345189001415493e-06,
      "loss": 0.1472,
      "step": 4250
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3409633760900974e-06,
      "loss": 0.1468,
      "step": 4251
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3367410563984403e-06,
      "loss": 0.152,
      "step": 4252
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3325220441628747e-06,
      "loss": 0.1603,
      "step": 4253
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.328306341204316e-06,
      "loss": 0.1219,
      "step": 4254
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.324093949342261e-06,
      "loss": 0.1369,
      "step": 4255
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3198848703947783e-06,
      "loss": 0.1427,
      "step": 4256
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.315679106178497e-06,
      "loss": 0.1206,
      "step": 4257
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3114766585086234e-06,
      "loss": 0.1159,
      "step": 4258
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3072775291989333e-06,
      "loss": 0.1393,
      "step": 4259
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3030817200617616e-06,
      "loss": 0.1539,
      "step": 4260
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.29888923290802e-06,
      "loss": 0.1538,
      "step": 4261
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.2947000695471854e-06,
      "loss": 0.1174,
      "step": 4262
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.2905142317872896e-06,
      "loss": 0.1204,
      "step": 4263
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.28633172143494e-06,
      "loss": 0.1475,
      "step": 4264
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2821525402953093e-06,
      "loss": 0.1663,
      "step": 4265
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2779766901721166e-06,
      "loss": 0.1567,
      "step": 4266
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.273804172867671e-06,
      "loss": 0.147,
      "step": 4267
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.269634990182815e-06,
      "loss": 0.1551,
      "step": 4268
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2654691439169684e-06,
      "loss": 0.1217,
      "step": 4269
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2613066358681112e-06,
      "loss": 0.1432,
      "step": 4270
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.257147467832771e-06,
      "loss": 0.1409,
      "step": 4271
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.252991641606045e-06,
      "loss": 0.1256,
      "step": 4272
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2488391589815828e-06,
      "loss": 0.1454,
      "step": 4273
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2446900217515977e-06,
      "loss": 0.1149,
      "step": 4274
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.240544231706846e-06,
      "loss": 0.1561,
      "step": 4275
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.236401790636651e-06,
      "loss": 0.1384,
      "step": 4276
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2322627003288898e-06,
      "loss": 0.1526,
      "step": 4277
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2281269625699842e-06,
      "loss": 0.1269,
      "step": 4278
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.223994579144917e-06,
      "loss": 0.1598,
      "step": 4279
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2198655518372237e-06,
      "loss": 0.1333,
      "step": 4280
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2157398824289865e-06,
      "loss": 0.162,
      "step": 4281
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.211617572700846e-06,
      "loss": 0.1586,
      "step": 4282
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2074986244319797e-06,
      "loss": 0.1501,
      "step": 4283
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.2033830394001264e-06,
      "loss": 0.1472,
      "step": 4284
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.1992708193815714e-06,
      "loss": 0.1489,
      "step": 4285
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.1951619661511412e-06,
      "loss": 0.1428,
      "step": 4286
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.1910564814822142e-06,
      "loss": 0.1733,
      "step": 4287
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.1869543671467177e-06,
      "loss": 0.1245,
      "step": 4288
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.1828556249151143e-06,
      "loss": 0.1559,
      "step": 4289
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.1787602565564223e-06,
      "loss": 0.1359,
      "step": 4290
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.174668263838202e-06,
      "loss": 0.1559,
      "step": 4291
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.170579648526546e-06,
      "loss": 0.1552,
      "step": 4292
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.166494412386102e-06,
      "loss": 0.1244,
      "step": 4293
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.162412557180057e-06,
      "loss": 0.1599,
      "step": 4294
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1583340846701286e-06,
      "loss": 0.1789,
      "step": 4295
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1542589966165906e-06,
      "loss": 0.1525,
      "step": 4296
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.150187294778242e-06,
      "loss": 0.1495,
      "step": 4297
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.146118980912427e-06,
      "loss": 0.1308,
      "step": 4298
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1420540567750305e-06,
      "loss": 0.1326,
      "step": 4299
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1379925241204626e-06,
      "loss": 0.1132,
      "step": 4300
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1339343847016823e-06,
      "loss": 0.1403,
      "step": 4301
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1298796402701815e-06,
      "loss": 0.1572,
      "step": 4302
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1258282925759788e-06,
      "loss": 0.1508,
      "step": 4303
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1217803433676353e-06,
      "loss": 0.1416,
      "step": 4304
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1177357943922437e-06,
      "loss": 0.1449,
      "step": 4305
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.113694647395426e-06,
      "loss": 0.1426,
      "step": 4306
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1096569041213356e-06,
      "loss": 0.1454,
      "step": 4307
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.1056225663126662e-06,
      "loss": 0.1504,
      "step": 4308
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.101591635710627e-06,
      "loss": 0.136,
      "step": 4309
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.097564114054964e-06,
      "loss": 0.1392,
      "step": 4310
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.0935400030839616e-06,
      "loss": 0.1462,
      "step": 4311
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.089519304534414e-06,
      "loss": 0.1293,
      "step": 4312
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.0855020201416543e-06,
      "loss": 0.1472,
      "step": 4313
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.08148815163954e-06,
      "loss": 0.1588,
      "step": 4314
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.077477700760451e-06,
      "loss": 0.1643,
      "step": 4315
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.0734706692352947e-06,
      "loss": 0.1619,
      "step": 4316
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.0694670587935064e-06,
      "loss": 0.1481,
      "step": 4317
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.0654668711630345e-06,
      "loss": 0.1458,
      "step": 4318
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.06147010807036e-06,
      "loss": 0.1413,
      "step": 4319
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0574767712404865e-06,
      "loss": 0.1334,
      "step": 4320
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.053486862396927e-06,
      "loss": 0.1586,
      "step": 4321
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.049500383261728e-06,
      "loss": 0.1689,
      "step": 4322
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0455173355554504e-06,
      "loss": 0.1577,
      "step": 4323
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0415377209971686e-06,
      "loss": 0.1412,
      "step": 4324
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.03756154130449e-06,
      "loss": 0.1584,
      "step": 4325
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.033588798193523e-06,
      "loss": 0.1405,
      "step": 4326
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0296194933789027e-06,
      "loss": 0.1424,
      "step": 4327
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0256536285737817e-06,
      "loss": 0.1492,
      "step": 4328
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0216912054898165e-06,
      "loss": 0.1546,
      "step": 4329
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.017732225837189e-06,
      "loss": 0.166,
      "step": 4330
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0137766913245935e-06,
      "loss": 0.136,
      "step": 4331
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0098246036592316e-06,
      "loss": 0.1519,
      "step": 4332
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.005875964546823e-06,
      "loss": 0.155,
      "step": 4333
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.0019307756915996e-06,
      "loss": 0.1666,
      "step": 4334
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.9979890387962953e-06,
      "loss": 0.1905,
      "step": 4335
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.9940507555621646e-06,
      "loss": 0.1524,
      "step": 4336
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.990115927688969e-06,
      "loss": 0.1292,
      "step": 4337
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.9861845568749728e-06,
      "loss": 0.1394,
      "step": 4338
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.98225664481695e-06,
      "loss": 0.1639,
      "step": 4339
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.978332193210194e-06,
      "loss": 0.1392,
      "step": 4340
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.974411203748485e-06,
      "loss": 0.1532,
      "step": 4341
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.970493678124125e-06,
      "loss": 0.1548,
      "step": 4342
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.966579618027913e-06,
      "loss": 0.1426,
      "step": 4343
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.962669025149151e-06,
      "loss": 0.1133,
      "step": 4344
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.958761901175651e-06,
      "loss": 0.1237,
      "step": 4345
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.954858247793725e-06,
      "loss": 0.1256,
      "step": 4346
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9509580666881823e-06,
      "loss": 0.1355,
      "step": 4347
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9470613595423405e-06,
      "loss": 0.1332,
      "step": 4348
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9431681280380177e-06,
      "loss": 0.1386,
      "step": 4349
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9392783738555245e-06,
      "loss": 0.1385,
      "step": 4350
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9353920986736776e-06,
      "loss": 0.1431,
      "step": 4351
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.931509304169792e-06,
      "loss": 0.1288,
      "step": 4352
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9276299920196752e-06,
      "loss": 0.1683,
      "step": 4353
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9237541638976355e-06,
      "loss": 0.169,
      "step": 4354
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.919881821476478e-06,
      "loss": 0.129,
      "step": 4355
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.916012966427503e-06,
      "loss": 0.1315,
      "step": 4356
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9121476004205063e-06,
      "loss": 0.1361,
      "step": 4357
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.908285725123773e-06,
      "loss": 0.1378,
      "step": 4358
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.904427342204086e-06,
      "loss": 0.168,
      "step": 4359
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.9005724533267256e-06,
      "loss": 0.1288,
      "step": 4360
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8967210601554497e-06,
      "loss": 0.1382,
      "step": 4361
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8928731643525222e-06,
      "loss": 0.1259,
      "step": 4362
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8890287675786922e-06,
      "loss": 0.1571,
      "step": 4363
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8851878714931937e-06,
      "loss": 0.1355,
      "step": 4364
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8813504777537561e-06,
      "loss": 0.1395,
      "step": 4365
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8775165880165992e-06,
      "loss": 0.1286,
      "step": 4366
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.87368620393642e-06,
      "loss": 0.138,
      "step": 4367
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8698593271664134e-06,
      "loss": 0.1496,
      "step": 4368
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8660359593582533e-06,
      "loss": 0.1418,
      "step": 4369
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8622161021621055e-06,
      "loss": 0.1295,
      "step": 4370
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.8583997572266177e-06,
      "loss": 0.133,
      "step": 4371
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8545869261989157e-06,
      "loss": 0.1522,
      "step": 4372
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8507776107246168e-06,
      "loss": 0.1475,
      "step": 4373
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8469718124478198e-06,
      "loss": 0.1727,
      "step": 4374
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8431695330111056e-06,
      "loss": 0.143,
      "step": 4375
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.839370774055529e-06,
      "loss": 0.1452,
      "step": 4376
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8355755372206353e-06,
      "loss": 0.1558,
      "step": 4377
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8317838241444464e-06,
      "loss": 0.1429,
      "step": 4378
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8279956364634587e-06,
      "loss": 0.1403,
      "step": 4379
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8242109758126525e-06,
      "loss": 0.1644,
      "step": 4380
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8204298438254875e-06,
      "loss": 0.1523,
      "step": 4381
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8166522421338916e-06,
      "loss": 0.1273,
      "step": 4382
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8128781723682764e-06,
      "loss": 0.1347,
      "step": 4383
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8091076361575277e-06,
      "loss": 0.1247,
      "step": 4384
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8053406351290059e-06,
      "loss": 0.129,
      "step": 4385
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.8015771709085483e-06,
      "loss": 0.1224,
      "step": 4386
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7978172451204567e-06,
      "loss": 0.1497,
      "step": 4387
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7940608593875154e-06,
      "loss": 0.1371,
      "step": 4388
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7903080153309794e-06,
      "loss": 0.138,
      "step": 4389
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7865587145705687e-06,
      "loss": 0.1271,
      "step": 4390
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.782812958724479e-06,
      "loss": 0.1352,
      "step": 4391
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7790707494093807e-06,
      "loss": 0.1428,
      "step": 4392
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7753320882404012e-06,
      "loss": 0.1538,
      "step": 4393
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.771596976831146e-06,
      "loss": 0.1412,
      "step": 4394
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7678654167936893e-06,
      "loss": 0.1407,
      "step": 4395
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7641374097385645e-06,
      "loss": 0.1541,
      "step": 4396
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7604129572747775e-06,
      "loss": 0.1383,
      "step": 4397
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7566920610097993e-06,
      "loss": 0.1595,
      "step": 4398
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7529747225495664e-06,
      "loss": 0.1511,
      "step": 4399
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7492609434984808e-06,
      "loss": 0.1335,
      "step": 4400
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7455507254594018e-06,
      "loss": 0.1713,
      "step": 4401
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.741844070033658e-06,
      "loss": 0.1416,
      "step": 4402
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7381409788210412e-06,
      "loss": 0.1681,
      "step": 4403
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7344414534197995e-06,
      "loss": 0.1415,
      "step": 4404
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7307454954266456e-06,
      "loss": 0.1574,
      "step": 4405
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7270531064367546e-06,
      "loss": 0.1657,
      "step": 4406
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7233642880437528e-06,
      "loss": 0.1258,
      "step": 4407
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7196790418397357e-06,
      "loss": 0.1372,
      "step": 4408
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.715997369415253e-06,
      "loss": 0.1405,
      "step": 4409
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7123192723593075e-06,
      "loss": 0.1385,
      "step": 4410
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7086447522593642e-06,
      "loss": 0.1491,
      "step": 4411
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7049738107013435e-06,
      "loss": 0.1508,
      "step": 4412
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.701306449269622e-06,
      "loss": 0.1421,
      "step": 4413
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6976426695470272e-06,
      "loss": 0.137,
      "step": 4414
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.693982473114848e-06,
      "loss": 0.1294,
      "step": 4415
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.690325861552815e-06,
      "loss": 0.1528,
      "step": 4416
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.686672836439124e-06,
      "loss": 0.1236,
      "step": 4417
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6830233993504174e-06,
      "loss": 0.1206,
      "step": 4418
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.679377551861786e-06,
      "loss": 0.1244,
      "step": 4419
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6757352955467754e-06,
      "loss": 0.1562,
      "step": 4420
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6720966319773835e-06,
      "loss": 0.1443,
      "step": 4421
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6684615627240509e-06,
      "loss": 0.1309,
      "step": 4422
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.66483008935567e-06,
      "loss": 0.1621,
      "step": 4423
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6612022134395865e-06,
      "loss": 0.181,
      "step": 4424
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6575779365415834e-06,
      "loss": 0.1352,
      "step": 4425
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6539572602258968e-06,
      "loss": 0.1678,
      "step": 4426
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.650340186055208e-06,
      "loss": 0.1309,
      "step": 4427
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6467267155906442e-06,
      "loss": 0.1439,
      "step": 4428
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6431168503917772e-06,
      "loss": 0.1261,
      "step": 4429
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6395105920166165e-06,
      "loss": 0.1316,
      "step": 4430
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.635907942021625e-06,
      "loss": 0.1493,
      "step": 4431
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6323089019617045e-06,
      "loss": 0.1298,
      "step": 4432
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6287134733901921e-06,
      "loss": 0.1442,
      "step": 4433
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6251216578588747e-06,
      "loss": 0.143,
      "step": 4434
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6215334569179797e-06,
      "loss": 0.1382,
      "step": 4435
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6179488721161674e-06,
      "loss": 0.1512,
      "step": 4436
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6143679050005411e-06,
      "loss": 0.1285,
      "step": 4437
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6107905571166494e-06,
      "loss": 0.1308,
      "step": 4438
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6072168300084645e-06,
      "loss": 0.1576,
      "step": 4439
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6036467252184085e-06,
      "loss": 0.1459,
      "step": 4440
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6000802442873365e-06,
      "loss": 0.1222,
      "step": 4441
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5965173887545326e-06,
      "loss": 0.1314,
      "step": 4442
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5929581601577303e-06,
      "loss": 0.1279,
      "step": 4443
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5894025600330876e-06,
      "loss": 0.1511,
      "step": 4444
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5858505899151943e-06,
      "loss": 0.1265,
      "step": 4445
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.582302251337081e-06,
      "loss": 0.1685,
      "step": 4446
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5787575458302084e-06,
      "loss": 0.1798,
      "step": 4447
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5752164749244657e-06,
      "loss": 0.1345,
      "step": 4448
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5716790401481773e-06,
      "loss": 0.1394,
      "step": 4449
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.5681452430280997e-06,
      "loss": 0.1318,
      "step": 4450
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5646150850894115e-06,
      "loss": 0.1345,
      "step": 4451
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5610885678557287e-06,
      "loss": 0.1571,
      "step": 4452
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5575656928490957e-06,
      "loss": 0.1134,
      "step": 4453
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5540464615899764e-06,
      "loss": 0.1497,
      "step": 4454
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5505308755972725e-06,
      "loss": 0.1382,
      "step": 4455
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.547018936388307e-06,
      "loss": 0.1443,
      "step": 4456
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5435106454788295e-06,
      "loss": 0.1422,
      "step": 4457
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5400060043830178e-06,
      "loss": 0.1376,
      "step": 4458
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5365050146134674e-06,
      "loss": 0.1506,
      "step": 4459
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5330076776812053e-06,
      "loss": 0.1363,
      "step": 4460
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5295139950956817e-06,
      "loss": 0.1291,
      "step": 4461
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5260239683647616e-06,
      "loss": 0.1218,
      "step": 4462
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5225375989947399e-06,
      "loss": 0.1365,
      "step": 4463
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5190548884903334e-06,
      "loss": 0.1424,
      "step": 4464
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5155758383546726e-06,
      "loss": 0.1222,
      "step": 4465
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5121004500893144e-06,
      "loss": 0.1734,
      "step": 4466
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5086287251942356e-06,
      "loss": 0.1479,
      "step": 4467
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5051606651678264e-06,
      "loss": 0.1673,
      "step": 4468
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.5016962715068995e-06,
      "loss": 0.1257,
      "step": 4469
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4982355457066866e-06,
      "loss": 0.1228,
      "step": 4470
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4947784892608276e-06,
      "loss": 0.1364,
      "step": 4471
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4913251036613952e-06,
      "loss": 0.1453,
      "step": 4472
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4878753903988596e-06,
      "loss": 0.1557,
      "step": 4473
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4844293509621167e-06,
      "loss": 0.137,
      "step": 4474
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.480986986838474e-06,
      "loss": 0.1474,
      "step": 4475
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4775482995136558e-06,
      "loss": 0.1375,
      "step": 4476
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4741132904717913e-06,
      "loss": 0.1225,
      "step": 4477
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4706819611954315e-06,
      "loss": 0.1422,
      "step": 4478
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.467254313165535e-06,
      "loss": 0.1321,
      "step": 4479
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.463830347861469e-06,
      "loss": 0.1245,
      "step": 4480
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4604100667610154e-06,
      "loss": 0.1178,
      "step": 4481
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4569934713403678e-06,
      "loss": 0.1311,
      "step": 4482
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4535805630741218e-06,
      "loss": 0.1294,
      "step": 4483
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.450171343435286e-06,
      "loss": 0.1496,
      "step": 4484
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.44676581389528e-06,
      "loss": 0.1458,
      "step": 4485
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4433639759239203e-06,
      "loss": 0.1618,
      "step": 4486
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.439965830989448e-06,
      "loss": 0.1796,
      "step": 4487
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4365713805584902e-06,
      "loss": 0.17,
      "step": 4488
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4331806260960935e-06,
      "loss": 0.1603,
      "step": 4489
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4297935690657062e-06,
      "loss": 0.1657,
      "step": 4490
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4264102109291733e-06,
      "loss": 0.1329,
      "step": 4491
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4230305531467537e-06,
      "loss": 0.1319,
      "step": 4492
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4196545971771058e-06,
      "loss": 0.1289,
      "step": 4493
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4162823444772855e-06,
      "loss": 0.1537,
      "step": 4494
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.412913796502755e-06,
      "loss": 0.1314,
      "step": 4495
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4095489547073815e-06,
      "loss": 0.1448,
      "step": 4496
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.406187820543421e-06,
      "loss": 0.1522,
      "step": 4497
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4028303954615398e-06,
      "loss": 0.1242,
      "step": 4498
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.3994766809108018e-06,
      "loss": 0.1264,
      "step": 4499
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.3961266783386595e-06,
      "loss": 0.1295,
      "step": 4500
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.392780389190983e-06,
      "loss": 0.1555,
      "step": 4501
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.389437814912018e-06,
      "loss": 0.1228,
      "step": 4502
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.3860989569444195e-06,
      "loss": 0.1465,
      "step": 4503
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.38276381672924e-06,
      "loss": 0.1469,
      "step": 4504
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3794323957059164e-06,
      "loss": 0.1364,
      "step": 4505
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3761046953122913e-06,
      "loss": 0.1554,
      "step": 4506
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3727807169845987e-06,
      "loss": 0.1537,
      "step": 4507
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3694604621574604e-06,
      "loss": 0.1551,
      "step": 4508
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.366143932263898e-06,
      "loss": 0.1393,
      "step": 4509
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3628311287353247e-06,
      "loss": 0.1578,
      "step": 4510
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3595220530015407e-06,
      "loss": 0.1188,
      "step": 4511
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3562167064907405e-06,
      "loss": 0.146,
      "step": 4512
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3529150906295118e-06,
      "loss": 0.0971,
      "step": 4513
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3496172068428259e-06,
      "loss": 0.1558,
      "step": 4514
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3463230565540452e-06,
      "loss": 0.1506,
      "step": 4515
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3430326411849304e-06,
      "loss": 0.1605,
      "step": 4516
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.339745962155613e-06,
      "loss": 0.1377,
      "step": 4517
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3364630208846253e-06,
      "loss": 0.1488,
      "step": 4518
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3331838187888835e-06,
      "loss": 0.1492,
      "step": 4519
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3299083572836835e-06,
      "loss": 0.1398,
      "step": 4520
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3266366377827145e-06,
      "loss": 0.1191,
      "step": 4521
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.32336866169805e-06,
      "loss": 0.1476,
      "step": 4522
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3201044304401423e-06,
      "loss": 0.1542,
      "step": 4523
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3168439454178307e-06,
      "loss": 0.1431,
      "step": 4524
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3135872080383416e-06,
      "loss": 0.153,
      "step": 4525
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3103342197072754e-06,
      "loss": 0.1544,
      "step": 4526
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3070849818286214e-06,
      "loss": 0.1222,
      "step": 4527
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3038394958047496e-06,
      "loss": 0.1318,
      "step": 4528
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.3005977630364052e-06,
      "loss": 0.1306,
      "step": 4529
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.2973597849227192e-06,
      "loss": 0.1308,
      "step": 4530
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2941255628612004e-06,
      "loss": 0.158,
      "step": 4531
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2908950982477364e-06,
      "loss": 0.1354,
      "step": 4532
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2876683924765955e-06,
      "loss": 0.1794,
      "step": 4533
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2844454469404167e-06,
      "loss": 0.1456,
      "step": 4534
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.281226263030222e-06,
      "loss": 0.1593,
      "step": 4535
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.278010842135411e-06,
      "loss": 0.122,
      "step": 4536
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2747991856437524e-06,
      "loss": 0.1496,
      "step": 4537
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2715912949413967e-06,
      "loss": 0.1222,
      "step": 4538
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2683871714128682e-06,
      "loss": 0.1552,
      "step": 4539
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2651868164410609e-06,
      "loss": 0.1296,
      "step": 4540
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2619902314072475e-06,
      "loss": 0.1501,
      "step": 4541
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.258797417691071e-06,
      "loss": 0.1016,
      "step": 4542
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2556083766705462e-06,
      "loss": 0.142,
      "step": 4543
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2524231097220585e-06,
      "loss": 0.1604,
      "step": 4544
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2492416182203736e-06,
      "loss": 0.1532,
      "step": 4545
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2460639035386135e-06,
      "loss": 0.1213,
      "step": 4546
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2428899670482808e-06,
      "loss": 0.1372,
      "step": 4547
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2397198101192432e-06,
      "loss": 0.1214,
      "step": 4548
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.236553434119736e-06,
      "loss": 0.1334,
      "step": 4549
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2333908404163653e-06,
      "loss": 0.1343,
      "step": 4550
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2302320303741067e-06,
      "loss": 0.1328,
      "step": 4551
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2270770053562953e-06,
      "loss": 0.1349,
      "step": 4552
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2239257667246396e-06,
      "loss": 0.149,
      "step": 4553
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2207783158392127e-06,
      "loss": 0.1284,
      "step": 4554
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2176346540584484e-06,
      "loss": 0.1522,
      "step": 4555
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2144947827391507e-06,
      "loss": 0.1369,
      "step": 4556
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2113587032364882e-06,
      "loss": 0.1273,
      "step": 4557
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2082264169039837e-06,
      "loss": 0.1279,
      "step": 4558
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2050979250935335e-06,
      "loss": 0.1391,
      "step": 4559
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.201973229155391e-06,
      "loss": 0.1496,
      "step": 4560
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1988523304381716e-06,
      "loss": 0.1288,
      "step": 4561
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.195735230288857e-06,
      "loss": 0.1512,
      "step": 4562
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1926219300527796e-06,
      "loss": 0.1365,
      "step": 4563
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1895124310736384e-06,
      "loss": 0.1574,
      "step": 4564
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1864067346934937e-06,
      "loss": 0.1653,
      "step": 4565
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.183304842252756e-06,
      "loss": 0.1645,
      "step": 4566
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.180206755090203e-06,
      "loss": 0.1433,
      "step": 4567
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1771124745429685e-06,
      "loss": 0.1495,
      "step": 4568
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1740220019465364e-06,
      "loss": 0.1725,
      "step": 4569
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1709353386347544e-06,
      "loss": 0.1436,
      "step": 4570
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1678524859398254e-06,
      "loss": 0.1303,
      "step": 4571
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1647734451923031e-06,
      "loss": 0.1468,
      "step": 4572
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1616982177211001e-06,
      "loss": 0.1199,
      "step": 4573
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.158626804853482e-06,
      "loss": 0.1204,
      "step": 4574
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1555592079150691e-06,
      "loss": 0.1452,
      "step": 4575
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1524954282298317e-06,
      "loss": 0.1483,
      "step": 4576
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.149435467120097e-06,
      "loss": 0.1355,
      "step": 4577
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1463793259065391e-06,
      "loss": 0.1781,
      "step": 4578
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.143327005908187e-06,
      "loss": 0.16,
      "step": 4579
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1402785084424206e-06,
      "loss": 0.1412,
      "step": 4580
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1372338348249667e-06,
      "loss": 0.1382,
      "step": 4581
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.1341929863699042e-06,
      "loss": 0.1462,
      "step": 4582
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.131155964389663e-06,
      "loss": 0.1521,
      "step": 4583
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1281227701950158e-06,
      "loss": 0.1254,
      "step": 4584
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1250934050950891e-06,
      "loss": 0.1399,
      "step": 4585
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1220678703973554e-06,
      "loss": 0.144,
      "step": 4586
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1190461674076292e-06,
      "loss": 0.1303,
      "step": 4587
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1160282974300761e-06,
      "loss": 0.1493,
      "step": 4588
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1130142617672079e-06,
      "loss": 0.1383,
      "step": 4589
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1100040617198782e-06,
      "loss": 0.154,
      "step": 4590
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.106997698587291e-06,
      "loss": 0.1414,
      "step": 4591
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1039951736669829e-06,
      "loss": 0.135,
      "step": 4592
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.1009964882548463e-06,
      "loss": 0.1339,
      "step": 4593
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0980016436451113e-06,
      "loss": 0.1711,
      "step": 4594
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0950106411303474e-06,
      "loss": 0.1272,
      "step": 4595
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0920234820014697e-06,
      "loss": 0.1283,
      "step": 4596
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0890401675477357e-06,
      "loss": 0.1386,
      "step": 4597
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.086060699056739e-06,
      "loss": 0.1448,
      "step": 4598
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0830850778144165e-06,
      "loss": 0.111,
      "step": 4599
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.080113305105045e-06,
      "loss": 0.134,
      "step": 4600
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0771453822112366e-06,
      "loss": 0.1602,
      "step": 4601
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0741813104139441e-06,
      "loss": 0.1582,
      "step": 4602
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0712210909924614e-06,
      "loss": 0.1513,
      "step": 4603
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0682647252244149e-06,
      "loss": 0.1549,
      "step": 4604
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0653122143857719e-06,
      "loss": 0.1456,
      "step": 4605
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0623635597508297e-06,
      "loss": 0.1369,
      "step": 4606
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.059418762592227e-06,
      "loss": 0.1177,
      "step": 4607
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.056477824180937e-06,
      "loss": 0.1364,
      "step": 4608
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0535407457862645e-06,
      "loss": 0.1367,
      "step": 4609
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.0506075286758488e-06,
      "loss": 0.1318,
      "step": 4610
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0476781741156695e-06,
      "loss": 0.1506,
      "step": 4611
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.044752683370026e-06,
      "loss": 0.1417,
      "step": 4612
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0418310577015633e-06,
      "loss": 0.129,
      "step": 4613
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0389132983712512e-06,
      "loss": 0.1203,
      "step": 4614
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0359994066383904e-06,
      "loss": 0.139,
      "step": 4615
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0330893837606149e-06,
      "loss": 0.1221,
      "step": 4616
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.030183230993891e-06,
      "loss": 0.1453,
      "step": 4617
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.027280949592504e-06,
      "loss": 0.1346,
      "step": 4618
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0243825408090847e-06,
      "loss": 0.1411,
      "step": 4619
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0214880058945819e-06,
      "loss": 0.1452,
      "step": 4620
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0185973460982723e-06,
      "loss": 0.16,
      "step": 4621
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.015710562667762e-06,
      "loss": 0.1279,
      "step": 4622
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0128276568489881e-06,
      "loss": 0.1324,
      "step": 4623
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0099486298862048e-06,
      "loss": 0.1552,
      "step": 4624
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0070734830220007e-06,
      "loss": 0.1294,
      "step": 4625
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0042022174972888e-06,
      "loss": 0.1263,
      "step": 4626
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.0013348345513007e-06,
      "loss": 0.1478,
      "step": 4627
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.98471335421598e-07,
      "loss": 0.1317,
      "step": 4628
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.956117213440674e-07,
      "loss": 0.1492,
      "step": 4629
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.927559935529107e-07,
      "loss": 0.1365,
      "step": 4630
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.899041532806607e-07,
      "loss": 0.1286,
      "step": 4631
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.870562017581687e-07,
      "loss": 0.1434,
      "step": 4632
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.842121402146087e-07,
      "loss": 0.1308,
      "step": 4633
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.813719698774771e-07,
      "loss": 0.1618,
      "step": 4634
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.785356919725842e-07,
      "loss": 0.156,
      "step": 4635
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.757033077240673e-07,
      "loss": 0.1285,
      "step": 4636
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.728748183543846e-07,
      "loss": 0.1334,
      "step": 4637
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.700502250843035e-07,
      "loss": 0.1471,
      "step": 4638
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.672295291329203e-07,
      "loss": 0.1527,
      "step": 4639
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.644127317176454e-07,
      "loss": 0.1592,
      "step": 4640
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.615998340542032e-07,
      "loss": 0.1543,
      "step": 4641
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.58790837356639e-07,
      "loss": 0.1382,
      "step": 4642
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.559857428373153e-07,
      "loss": 0.149,
      "step": 4643
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.531845517069061e-07,
      "loss": 0.1395,
      "step": 4644
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.503872651744028e-07,
      "loss": 0.1177,
      "step": 4645
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.475938844471166e-07,
      "loss": 0.1333,
      "step": 4646
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.448044107306598e-07,
      "loss": 0.1349,
      "step": 4647
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.420188452289747e-07,
      "loss": 0.1471,
      "step": 4648
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.392371891443075e-07,
      "loss": 0.1277,
      "step": 4649
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.364594436772157e-07,
      "loss": 0.1415,
      "step": 4650
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.336856100265734e-07,
      "loss": 0.1651,
      "step": 4651
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.309156893895677e-07,
      "loss": 0.1491,
      "step": 4652
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.281496829616888e-07,
      "loss": 0.1352,
      "step": 4653
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.253875919367461e-07,
      "loss": 0.1634,
      "step": 4654
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.226294175068562e-07,
      "loss": 0.1309,
      "step": 4655
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.198751608624412e-07,
      "loss": 0.1582,
      "step": 4656
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.171248231922391e-07,
      "loss": 0.1356,
      "step": 4657
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.143784056832939e-07,
      "loss": 0.1349,
      "step": 4658
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.116359095209548e-07,
      "loss": 0.1417,
      "step": 4659
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.088973358888809e-07,
      "loss": 0.1871,
      "step": 4660
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.061626859690409e-07,
      "loss": 0.166,
      "step": 4661
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.034319609417031e-07,
      "loss": 0.1392,
      "step": 4662
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.007051619854513e-07,
      "loss": 0.1372,
      "step": 4663
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.979822902771651e-07,
      "loss": 0.1637,
      "step": 4664
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.952633469920357e-07,
      "loss": 0.1602,
      "step": 4665
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.925483333035578e-07,
      "loss": 0.1293,
      "step": 4666
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.898372503835273e-07,
      "loss": 0.1424,
      "step": 4667
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.871300994020449e-07,
      "loss": 0.1556,
      "step": 4668
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.84426881527517e-07,
      "loss": 0.1359,
      "step": 4669
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.817275979266471e-07,
      "loss": 0.1405,
      "step": 4670
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.790322497644455e-07,
      "loss": 0.151,
      "step": 4671
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.763408382042238e-07,
      "loss": 0.1296,
      "step": 4672
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.736533644075895e-07,
      "loss": 0.1569,
      "step": 4673
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.709698295344549e-07,
      "loss": 0.1339,
      "step": 4674
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.682902347430344e-07,
      "loss": 0.1136,
      "step": 4675
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.656145811898331e-07,
      "loss": 0.1471,
      "step": 4676
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.629428700296661e-07,
      "loss": 0.1462,
      "step": 4677
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.60275102415642e-07,
      "loss": 0.1265,
      "step": 4678
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.57611279499162e-07,
      "loss": 0.1436,
      "step": 4679
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.549514024299343e-07,
      "loss": 0.1485,
      "step": 4680
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.522954723559585e-07,
      "loss": 0.1177,
      "step": 4681
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.496434904235306e-07,
      "loss": 0.1432,
      "step": 4682
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.469954577772432e-07,
      "loss": 0.1432,
      "step": 4683
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.443513755599897e-07,
      "loss": 0.1428,
      "step": 4684
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.417112449129472e-07,
      "loss": 0.1499,
      "step": 4685
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.390750669755976e-07,
      "loss": 0.129,
      "step": 4686
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.364428428857141e-07,
      "loss": 0.148,
      "step": 4687
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.33814573779359e-07,
      "loss": 0.1459,
      "step": 4688
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.311902607908917e-07,
      "loss": 0.1391,
      "step": 4689
    },
    {
      "epoch": 1.76,
      "learning_rate": 8.285699050529672e-07,
      "loss": 0.1429,
      "step": 4690
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.25953507696523e-07,
      "loss": 0.1341,
      "step": 4691
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.233410698508005e-07,
      "loss": 0.1548,
      "step": 4692
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.207325926433196e-07,
      "loss": 0.1823,
      "step": 4693
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.181280771999e-07,
      "loss": 0.1444,
      "step": 4694
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.155275246446504e-07,
      "loss": 0.1458,
      "step": 4695
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.129309360999627e-07,
      "loss": 0.1279,
      "step": 4696
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.103383126865239e-07,
      "loss": 0.1592,
      "step": 4697
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.0774965552331e-07,
      "loss": 0.1235,
      "step": 4698
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.051649657275795e-07,
      "loss": 0.1436,
      "step": 4699
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.025842444148857e-07,
      "loss": 0.162,
      "step": 4700
    },
    {
      "epoch": 1.77,
      "learning_rate": 8.000074926990653e-07,
      "loss": 0.1543,
      "step": 4701
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.974347116922398e-07,
      "loss": 0.1499,
      "step": 4702
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.948659025048211e-07,
      "loss": 0.1489,
      "step": 4703
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.923010662455066e-07,
      "loss": 0.1423,
      "step": 4704
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.897402040212743e-07,
      "loss": 0.1412,
      "step": 4705
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.87183316937391e-07,
      "loss": 0.1536,
      "step": 4706
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.846304060974086e-07,
      "loss": 0.1483,
      "step": 4707
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.820814726031589e-07,
      "loss": 0.1609,
      "step": 4708
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.795365175547631e-07,
      "loss": 0.1535,
      "step": 4709
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.769955420506181e-07,
      "loss": 0.181,
      "step": 4710
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.744585471874078e-07,
      "loss": 0.1306,
      "step": 4711
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.719255340600995e-07,
      "loss": 0.1472,
      "step": 4712
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.693965037619344e-07,
      "loss": 0.1511,
      "step": 4713
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.668714573844438e-07,
      "loss": 0.1109,
      "step": 4714
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.643503960174358e-07,
      "loss": 0.133,
      "step": 4715
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.618333207489959e-07,
      "loss": 0.1295,
      "step": 4716
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.593202326654925e-07,
      "loss": 0.1291,
      "step": 4717
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.568111328515737e-07,
      "loss": 0.1247,
      "step": 4718
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.543060223901632e-07,
      "loss": 0.1436,
      "step": 4719
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.518049023624629e-07,
      "loss": 0.164,
      "step": 4720
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.493077738479593e-07,
      "loss": 0.1425,
      "step": 4721
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.468146379244079e-07,
      "loss": 0.1356,
      "step": 4722
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.443254956678426e-07,
      "loss": 0.117,
      "step": 4723
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.418403481525793e-07,
      "loss": 0.1171,
      "step": 4724
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.393591964512003e-07,
      "loss": 0.1392,
      "step": 4725
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.368820416345724e-07,
      "loss": 0.1452,
      "step": 4726
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.344088847718334e-07,
      "loss": 0.1542,
      "step": 4727
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.31939726930393e-07,
      "loss": 0.1251,
      "step": 4728
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.294745691759397e-07,
      "loss": 0.1272,
      "step": 4729
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.270134125724349e-07,
      "loss": 0.1564,
      "step": 4730
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.245562581821075e-07,
      "loss": 0.1318,
      "step": 4731
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.221031070654661e-07,
      "loss": 0.1371,
      "step": 4732
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.196539602812901e-07,
      "loss": 0.1531,
      "step": 4733
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.172088188866255e-07,
      "loss": 0.1398,
      "step": 4734
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.147676839367945e-07,
      "loss": 0.1463,
      "step": 4735
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.123305564853899e-07,
      "loss": 0.1287,
      "step": 4736
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.098974375842738e-07,
      "loss": 0.1262,
      "step": 4737
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.074683282835792e-07,
      "loss": 0.1171,
      "step": 4738
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.050432296317045e-07,
      "loss": 0.1278,
      "step": 4739
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.026221426753221e-07,
      "loss": 0.1317,
      "step": 4740
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.002050684593731e-07,
      "loss": 0.1355,
      "step": 4741
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.977920080270606e-07,
      "loss": 0.1377,
      "step": 4742
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.953829624198616e-07,
      "loss": 0.1313,
      "step": 4743
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.929779326775187e-07,
      "loss": 0.1404,
      "step": 4744
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.905769198380396e-07,
      "loss": 0.1453,
      "step": 4745
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.881799249376997e-07,
      "loss": 0.1245,
      "step": 4746
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.857869490110413e-07,
      "loss": 0.1228,
      "step": 4747
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.833979930908685e-07,
      "loss": 0.1691,
      "step": 4748
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.810130582082531e-07,
      "loss": 0.1561,
      "step": 4749
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.786321453925316e-07,
      "loss": 0.132,
      "step": 4750
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.762552556713042e-07,
      "loss": 0.134,
      "step": 4751
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.73882390070435e-07,
      "loss": 0.1241,
      "step": 4752
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.715135496140501e-07,
      "loss": 0.1508,
      "step": 4753
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.691487353245385e-07,
      "loss": 0.1525,
      "step": 4754
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.667879482225525e-07,
      "loss": 0.1703,
      "step": 4755
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.644311893270072e-07,
      "loss": 0.1123,
      "step": 4756
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.620784596550756e-07,
      "loss": 0.1302,
      "step": 4757
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.597297602221942e-07,
      "loss": 0.1517,
      "step": 4758
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.57385092042061e-07,
      "loss": 0.145,
      "step": 4759
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.550444561266311e-07,
      "loss": 0.1296,
      "step": 4760
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.52707853486122e-07,
      "loss": 0.1721,
      "step": 4761
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.503752851290102e-07,
      "loss": 0.1489,
      "step": 4762
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.480467520620282e-07,
      "loss": 0.1377,
      "step": 4763
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.457222552901698e-07,
      "loss": 0.1443,
      "step": 4764
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.434017958166861e-07,
      "loss": 0.144,
      "step": 4765
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.410853746430856e-07,
      "loss": 0.1378,
      "step": 4766
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.387729927691356e-07,
      "loss": 0.1464,
      "step": 4767
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.364646511928529e-07,
      "loss": 0.1433,
      "step": 4768
    },
    {
      "epoch": 1.79,
      "learning_rate": 6.341603509105198e-07,
      "loss": 0.1347,
      "step": 4769
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.318600929166719e-07,
      "loss": 0.1366,
      "step": 4770
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.295638782040936e-07,
      "loss": 0.1184,
      "step": 4771
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.272717077638324e-07,
      "loss": 0.1353,
      "step": 4772
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.24983582585188e-07,
      "loss": 0.1792,
      "step": 4773
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.226995036557104e-07,
      "loss": 0.13,
      "step": 4774
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.204194719612077e-07,
      "loss": 0.1713,
      "step": 4775
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.181434884857407e-07,
      "loss": 0.1582,
      "step": 4776
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.1587155421162e-07,
      "loss": 0.1151,
      "step": 4777
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.136036701194114e-07,
      "loss": 0.1538,
      "step": 4778
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.113398371879308e-07,
      "loss": 0.1463,
      "step": 4779
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.090800563942479e-07,
      "loss": 0.1513,
      "step": 4780
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.068243287136821e-07,
      "loss": 0.1334,
      "step": 4781
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.045726551198039e-07,
      "loss": 0.1507,
      "step": 4782
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.023250365844324e-07,
      "loss": 0.1448,
      "step": 4783
    },
    {
      "epoch": 1.8,
      "learning_rate": 6.000814740776384e-07,
      "loss": 0.1559,
      "step": 4784
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.978419685677428e-07,
      "loss": 0.1587,
      "step": 4785
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.956065210213114e-07,
      "loss": 0.1273,
      "step": 4786
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.933751324031633e-07,
      "loss": 0.1411,
      "step": 4787
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.911478036763663e-07,
      "loss": 0.1559,
      "step": 4788
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.889245358022289e-07,
      "loss": 0.1589,
      "step": 4789
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.867053297403136e-07,
      "loss": 0.1581,
      "step": 4790
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.844901864484309e-07,
      "loss": 0.1285,
      "step": 4791
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.822791068826294e-07,
      "loss": 0.1112,
      "step": 4792
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.800720919972125e-07,
      "loss": 0.1376,
      "step": 4793
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.778691427447247e-07,
      "loss": 0.1694,
      "step": 4794
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.756702600759578e-07,
      "loss": 0.1316,
      "step": 4795
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.734754449399494e-07,
      "loss": 0.1288,
      "step": 4796
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.71284698283977e-07,
      "loss": 0.141,
      "step": 4797
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.690980210535646e-07,
      "loss": 0.1518,
      "step": 4798
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.669154141924849e-07,
      "loss": 0.1658,
      "step": 4799
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.647368786427432e-07,
      "loss": 0.1327,
      "step": 4800
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.625624153445964e-07,
      "loss": 0.1312,
      "step": 4801
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.603920252365425e-07,
      "loss": 0.1531,
      "step": 4802
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.582257092553167e-07,
      "loss": 0.1418,
      "step": 4803
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.560634683359012e-07,
      "loss": 0.1595,
      "step": 4804
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.539053034115194e-07,
      "loss": 0.1501,
      "step": 4805
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.517512154136284e-07,
      "loss": 0.1336,
      "step": 4806
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.496012052719335e-07,
      "loss": 0.113,
      "step": 4807
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.474552739143779e-07,
      "loss": 0.1244,
      "step": 4808
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.45313422267143e-07,
      "loss": 0.1512,
      "step": 4809
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.431756512546526e-07,
      "loss": 0.1467,
      "step": 4810
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.410419617995621e-07,
      "loss": 0.1212,
      "step": 4811
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.389123548227726e-07,
      "loss": 0.1304,
      "step": 4812
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.367868312434221e-07,
      "loss": 0.1185,
      "step": 4813
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.346653919788825e-07,
      "loss": 0.1442,
      "step": 4814
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.325480379447645e-07,
      "loss": 0.1437,
      "step": 4815
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.304347700549206e-07,
      "loss": 0.13,
      "step": 4816
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.2832558922143e-07,
      "loss": 0.1416,
      "step": 4817
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.262204963546169e-07,
      "loss": 0.1729,
      "step": 4818
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.241194923630378e-07,
      "loss": 0.1145,
      "step": 4819
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.220225781534815e-07,
      "loss": 0.1326,
      "step": 4820
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.199297546309767e-07,
      "loss": 0.1379,
      "step": 4821
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.178410226987851e-07,
      "loss": 0.1368,
      "step": 4822
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.157563832583967e-07,
      "loss": 0.1459,
      "step": 4823
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.136758372095452e-07,
      "loss": 0.126,
      "step": 4824
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.115993854501922e-07,
      "loss": 0.1428,
      "step": 4825
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.095270288765298e-07,
      "loss": 0.1357,
      "step": 4826
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.074587683829868e-07,
      "loss": 0.1479,
      "step": 4827
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.053946048622238e-07,
      "loss": 0.1472,
      "step": 4828
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.033345392051304e-07,
      "loss": 0.1332,
      "step": 4829
    },
    {
      "epoch": 1.82,
      "learning_rate": 5.012785723008285e-07,
      "loss": 0.1525,
      "step": 4830
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.992267050366751e-07,
      "loss": 0.1586,
      "step": 4831
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.971789382982506e-07,
      "loss": 0.1448,
      "step": 4832
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.951352729693703e-07,
      "loss": 0.1133,
      "step": 4833
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.93095709932081e-07,
      "loss": 0.1727,
      "step": 4834
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.910602500666528e-07,
      "loss": 0.1446,
      "step": 4835
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.890288942515897e-07,
      "loss": 0.1668,
      "step": 4836
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.870016433636259e-07,
      "loss": 0.136,
      "step": 4837
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.84978498277715e-07,
      "loss": 0.1355,
      "step": 4838
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.829594598670517e-07,
      "loss": 0.1401,
      "step": 4839
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.809445290030479e-07,
      "loss": 0.1276,
      "step": 4840
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.789337065553445e-07,
      "loss": 0.1654,
      "step": 4841
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.769269933918152e-07,
      "loss": 0.1479,
      "step": 4842
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.749243903785505e-07,
      "loss": 0.141,
      "step": 4843
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.729258983798757e-07,
      "loss": 0.1768,
      "step": 4844
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.709315182583385e-07,
      "loss": 0.1497,
      "step": 4845
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.689412508747082e-07,
      "loss": 0.1605,
      "step": 4846
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.669550970879855e-07,
      "loss": 0.1379,
      "step": 4847
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.649730577553935e-07,
      "loss": 0.1024,
      "step": 4848
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.629951337323757e-07,
      "loss": 0.157,
      "step": 4849
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.6102132587260373e-07,
      "loss": 0.1126,
      "step": 4850
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.590516350279728e-07,
      "loss": 0.1528,
      "step": 4851
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.5708606204859617e-07,
      "loss": 0.1591,
      "step": 4852
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.551246077828175e-07,
      "loss": 0.1364,
      "step": 4853
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.5316727307719745e-07,
      "loss": 0.1464,
      "step": 4854
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.51214058776519e-07,
      "loss": 0.1542,
      "step": 4855
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.4926496572378907e-07,
      "loss": 0.1561,
      "step": 4856
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.4731999476023466e-07,
      "loss": 0.1311,
      "step": 4857
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.4537914672530214e-07,
      "loss": 0.1408,
      "step": 4858
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.4344242245666135e-07,
      "loss": 0.1515,
      "step": 4859
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.415098227902004e-07,
      "loss": 0.1182,
      "step": 4860
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.395813485600264e-07,
      "loss": 0.1553,
      "step": 4861
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.3765700059846926e-07,
      "loss": 0.1832,
      "step": 4862
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.357367797360745e-07,
      "loss": 0.14,
      "step": 4863
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.3382068680160816e-07,
      "loss": 0.1568,
      "step": 4864
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.319087226220553e-07,
      "loss": 0.1398,
      "step": 4865
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.3000088802261695e-07,
      "loss": 0.1335,
      "step": 4866
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.2809718382671117e-07,
      "loss": 0.1473,
      "step": 4867
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.261976108559795e-07,
      "loss": 0.1338,
      "step": 4868
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.24302169930273e-07,
      "loss": 0.1652,
      "step": 4869
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.224108618676626e-07,
      "loss": 0.1385,
      "step": 4870
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.205236874844376e-07,
      "loss": 0.1471,
      "step": 4871
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.186406475950966e-07,
      "loss": 0.1273,
      "step": 4872
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.167617430123605e-07,
      "loss": 0.1327,
      "step": 4873
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.148869745471651e-07,
      "loss": 0.1578,
      "step": 4874
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.1301634300865424e-07,
      "loss": 0.1389,
      "step": 4875
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.1114984920419323e-07,
      "loss": 0.1508,
      "step": 4876
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0928749393936096e-07,
      "loss": 0.1522,
      "step": 4877
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0742927801794563e-07,
      "loss": 0.1189,
      "step": 4878
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0557520224195234e-07,
      "loss": 0.1274,
      "step": 4879
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.037252674115999e-07,
      "loss": 0.1545,
      "step": 4880
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.018794743253174e-07,
      "loss": 0.1243,
      "step": 4881
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.000378237797453e-07,
      "loss": 0.1267,
      "step": 4882
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.982003165697446e-07,
      "loss": 0.1429,
      "step": 4883
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.963669534883763e-07,
      "loss": 0.1436,
      "step": 4884
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.9453773532692086e-07,
      "loss": 0.1339,
      "step": 4885
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.927126628748679e-07,
      "loss": 0.1411,
      "step": 4886
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.90891736919915e-07,
      "loss": 0.134,
      "step": 4887
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.8907495824797246e-07,
      "loss": 0.1369,
      "step": 4888
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.872623276431631e-07,
      "loss": 0.1307,
      "step": 4889
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.8545384588781433e-07,
      "loss": 0.1166,
      "step": 4890
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.836495137624652e-07,
      "loss": 0.1288,
      "step": 4891
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.8184933204586494e-07,
      "loss": 0.1706,
      "step": 4892
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.800533015149699e-07,
      "loss": 0.153,
      "step": 4893
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.782614229449455e-07,
      "loss": 0.146,
      "step": 4894
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.764736971091665e-07,
      "loss": 0.1308,
      "step": 4895
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.7469012477921007e-07,
      "loss": 0.1408,
      "step": 4896
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.729107067248694e-07,
      "loss": 0.1266,
      "step": 4897
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.7113544371413566e-07,
      "loss": 0.1554,
      "step": 4898
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.693643365132138e-07,
      "loss": 0.1605,
      "step": 4899
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.6759738588651117e-07,
      "loss": 0.122,
      "step": 4900
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.6583459259664e-07,
      "loss": 0.1518,
      "step": 4901
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.6407595740442394e-07,
      "loss": 0.1369,
      "step": 4902
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.6232148106888686e-07,
      "loss": 0.1628,
      "step": 4903
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.6057116434725867e-07,
      "loss": 0.1518,
      "step": 4904
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.588250079949751e-07,
      "loss": 0.1306,
      "step": 4905
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.570830127656766e-07,
      "loss": 0.1798,
      "step": 4906
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.5534517941120506e-07,
      "loss": 0.1723,
      "step": 4907
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.5361150868160833e-07,
      "loss": 0.1423,
      "step": 4908
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.5188200132514006e-07,
      "loss": 0.1643,
      "step": 4909
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.5015665808824983e-07,
      "loss": 0.1406,
      "step": 4910
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.4843547971559755e-07,
      "loss": 0.1346,
      "step": 4911
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.4671846695004006e-07,
      "loss": 0.1466,
      "step": 4912
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.450056205326402e-07,
      "loss": 0.1317,
      "step": 4913
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.432969412026621e-07,
      "loss": 0.1922,
      "step": 4914
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.41592429697567e-07,
      "loss": 0.1471,
      "step": 4915
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.398920867530231e-07,
      "loss": 0.1329,
      "step": 4916
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.3819591310289666e-07,
      "loss": 0.1473,
      "step": 4917
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.365039094792544e-07,
      "loss": 0.1453,
      "step": 4918
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.3481607661236203e-07,
      "loss": 0.1764,
      "step": 4919
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.3313241523069137e-07,
      "loss": 0.1553,
      "step": 4920
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.314529260609045e-07,
      "loss": 0.1362,
      "step": 4921
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.2977760982786934e-07,
      "loss": 0.1649,
      "step": 4922
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.281064672546519e-07,
      "loss": 0.122,
      "step": 4923
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.2643949906251414e-07,
      "loss": 0.1664,
      "step": 4924
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.247767059709184e-07,
      "loss": 0.1175,
      "step": 4925
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.23118088697526e-07,
      "loss": 0.1474,
      "step": 4926
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.214636479581945e-07,
      "loss": 0.1596,
      "step": 4927
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.198133844669782e-07,
      "loss": 0.1284,
      "step": 4928
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.1816729893613064e-07,
      "loss": 0.1206,
      "step": 4929
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.1652539207610025e-07,
      "loss": 0.139,
      "step": 4930
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.1488766459553233e-07,
      "loss": 0.1648,
      "step": 4931
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.132541172012693e-07,
      "loss": 0.143,
      "step": 4932
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.116247505983483e-07,
      "loss": 0.1201,
      "step": 4933
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.099995654900023e-07,
      "loss": 0.1404,
      "step": 4934
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.083785625776614e-07,
      "loss": 0.1597,
      "step": 4935
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.0676174256094703e-07,
      "loss": 0.1189,
      "step": 4936
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.051491061376788e-07,
      "loss": 0.1277,
      "step": 4937
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.035406540038699e-07,
      "loss": 0.1494,
      "step": 4938
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.0193638685372396e-07,
      "loss": 0.1516,
      "step": 4939
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.0033630537964376e-07,
      "loss": 0.1581,
      "step": 4940
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.9874041027222246e-07,
      "loss": 0.1255,
      "step": 4941
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.971487022202468e-07,
      "loss": 0.1394,
      "step": 4942
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.955611819106985e-07,
      "loss": 0.1547,
      "step": 4943
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.939778500287471e-07,
      "loss": 0.1407,
      "step": 4944
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.9239870725775834e-07,
      "loss": 0.1566,
      "step": 4945
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.9082375427929023e-07,
      "loss": 0.1333,
      "step": 4946
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.8925299177308906e-07,
      "loss": 0.1296,
      "step": 4947
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.8768642041709595e-07,
      "loss": 0.1284,
      "step": 4948
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.8612404088744215e-07,
      "loss": 0.1188,
      "step": 4949
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.8456585385844836e-07,
      "loss": 0.1396,
      "step": 4950
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.830118600026266e-07,
      "loss": 0.151,
      "step": 4951
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.8146205999068035e-07,
      "loss": 0.1343,
      "step": 4952
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.7991645449150117e-07,
      "loss": 0.1421,
      "step": 4953
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.783750441721711e-07,
      "loss": 0.1252,
      "step": 4954
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.7683782969796123e-07,
      "loss": 0.1405,
      "step": 4955
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.7530481173233316e-07,
      "loss": 0.1116,
      "step": 4956
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.737759909369353e-07,
      "loss": 0.1303,
      "step": 4957
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.7225136797160656e-07,
      "loss": 0.1458,
      "step": 4958
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.707309434943717e-07,
      "loss": 0.136,
      "step": 4959
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.692147181614446e-07,
      "loss": 0.1576,
      "step": 4960
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.677026926272275e-07,
      "loss": 0.1266,
      "step": 4961
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.6619486754430936e-07,
      "loss": 0.1401,
      "step": 4962
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.646912435634652e-07,
      "loss": 0.1418,
      "step": 4963
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.631918213336593e-07,
      "loss": 0.1455,
      "step": 4964
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.6169660150203947e-07,
      "loss": 0.1442,
      "step": 4965
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.6020558471394284e-07,
      "loss": 0.1088,
      "step": 4966
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.587187716128903e-07,
      "loss": 0.1501,
      "step": 4967
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.572361628405884e-07,
      "loss": 0.1329,
      "step": 4968
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.557577590369309e-07,
      "loss": 0.1571,
      "step": 4969
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.542835608399952e-07,
      "loss": 0.1297,
      "step": 4970
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.5281356888604445e-07,
      "loss": 0.1283,
      "step": 4971
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.5134778380952573e-07,
      "loss": 0.1466,
      "step": 4972
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.498862062430696e-07,
      "loss": 0.1312,
      "step": 4973
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.4842883681749365e-07,
      "loss": 0.1702,
      "step": 4974
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.4697567616179583e-07,
      "loss": 0.1392,
      "step": 4975
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.455267249031601e-07,
      "loss": 0.1481,
      "step": 4976
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.4408198366695056e-07,
      "loss": 0.1292,
      "step": 4977
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.426414530767196e-07,
      "loss": 0.1691,
      "step": 4978
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.412051337541954e-07,
      "loss": 0.1495,
      "step": 4979
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.397730263192943e-07,
      "loss": 0.1243,
      "step": 4980
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.3834513139011305e-07,
      "loss": 0.1551,
      "step": 4981
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.3692144958292863e-07,
      "loss": 0.14,
      "step": 4982
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.355019815122006e-07,
      "loss": 0.1471,
      "step": 4983
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.3408672779057117e-07,
      "loss": 0.1695,
      "step": 4984
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.3267568902886283e-07,
      "loss": 0.1273,
      "step": 4985
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.3126886583607845e-07,
      "loss": 0.1621,
      "step": 4986
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2986625881940227e-07,
      "loss": 0.131,
      "step": 4987
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2846786858419566e-07,
      "loss": 0.151,
      "step": 4988
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2707369573400584e-07,
      "loss": 0.1354,
      "step": 4989
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2568374087055588e-07,
      "loss": 0.1597,
      "step": 4990
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2429800459374818e-07,
      "loss": 0.137,
      "step": 4991
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.229164875016654e-07,
      "loss": 0.1329,
      "step": 4992
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2153919019057056e-07,
      "loss": 0.1155,
      "step": 4993
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2016611325490268e-07,
      "loss": 0.0968,
      "step": 4994
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.1879725728727986e-07,
      "loss": 0.1421,
      "step": 4995
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.1743262287850175e-07,
      "loss": 0.1482,
      "step": 4996
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.1607221061754168e-07,
      "loss": 0.143,
      "step": 4997
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.1471602109155222e-07,
      "loss": 0.1162,
      "step": 4998
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.1336405488586397e-07,
      "loss": 0.1489,
      "step": 4999
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.120163125839847e-07,
      "loss": 0.1527,
      "step": 5000
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.1067279476759906e-07,
      "loss": 0.1334,
      "step": 5001
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0933350201656765e-07,
      "loss": 0.1423,
      "step": 5002
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0799843490892813e-07,
      "loss": 0.1504,
      "step": 5003
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.066675940208951e-07,
      "loss": 0.1492,
      "step": 5004
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0534097992685687e-07,
      "loss": 0.1693,
      "step": 5005
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0401859319937988e-07,
      "loss": 0.151,
      "step": 5006
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.027004344092065e-07,
      "loss": 0.1528,
      "step": 5007
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0138650412524941e-07,
      "loss": 0.1242,
      "step": 5008
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.0007680291460275e-07,
      "loss": 0.1449,
      "step": 5009
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9877133134253211e-07,
      "loss": 0.1466,
      "step": 5010
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9747008997247685e-07,
      "loss": 0.1513,
      "step": 5011
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9617307936605213e-07,
      "loss": 0.1236,
      "step": 5012
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.948803000830479e-07,
      "loss": 0.1416,
      "step": 5013
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.935917526814235e-07,
      "loss": 0.1748,
      "step": 5014
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.923074377173195e-07,
      "loss": 0.1392,
      "step": 5015
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9102735574504038e-07,
      "loss": 0.1341,
      "step": 5016
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8975150731707082e-07,
      "loss": 0.1496,
      "step": 5017
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8847989298406698e-07,
      "loss": 0.1621,
      "step": 5018
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8721251329485323e-07,
      "loss": 0.1813,
      "step": 5019
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.859493687964331e-07,
      "loss": 0.125,
      "step": 5020
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8469046003397716e-07,
      "loss": 0.1507,
      "step": 5021
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.834357875508286e-07,
      "loss": 0.1339,
      "step": 5022
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8218535188850306e-07,
      "loss": 0.1459,
      "step": 5023
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8093915358669001e-07,
      "loss": 0.1773,
      "step": 5024
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7969719318324474e-07,
      "loss": 0.1286,
      "step": 5025
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7845947121419739e-07,
      "loss": 0.1568,
      "step": 5026
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7722598821374838e-07,
      "loss": 0.1447,
      "step": 5027
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.759967447142652e-07,
      "loss": 0.1675,
      "step": 5028
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7477174124629236e-07,
      "loss": 0.1395,
      "step": 5029
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7355097833853807e-07,
      "loss": 0.171,
      "step": 5030
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7233445651788195e-07,
      "loss": 0.1339,
      "step": 5031
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.7112217630937512e-07,
      "loss": 0.1238,
      "step": 5032
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6991413823623682e-07,
      "loss": 0.135,
      "step": 5033
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6871034281985444e-07,
      "loss": 0.1511,
      "step": 5034
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6751079057978458e-07,
      "loss": 0.1232,
      "step": 5035
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6631548203375537e-07,
      "loss": 0.1289,
      "step": 5036
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.651244176976574e-07,
      "loss": 0.1305,
      "step": 5037
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6393759808555509e-07,
      "loss": 0.1241,
      "step": 5038
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6275502370967976e-07,
      "loss": 0.1124,
      "step": 5039
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.615766950804276e-07,
      "loss": 0.1138,
      "step": 5040
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6040261270636516e-07,
      "loss": 0.1847,
      "step": 5041
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5923277709422592e-07,
      "loss": 0.1288,
      "step": 5042
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5806718874890715e-07,
      "loss": 0.1273,
      "step": 5043
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5690584817347975e-07,
      "loss": 0.1678,
      "step": 5044
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5574875586917504e-07,
      "loss": 0.1435,
      "step": 5045
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.545959123353935e-07,
      "loss": 0.1381,
      "step": 5046
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.534473180697016e-07,
      "loss": 0.1503,
      "step": 5047
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5230297356783164e-07,
      "loss": 0.1486,
      "step": 5048
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5116287932368078e-07,
      "loss": 0.1206,
      "step": 5049
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.5002703582931433e-07,
      "loss": 0.141,
      "step": 5050
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.4889544357496123e-07,
      "loss": 0.1334,
      "step": 5051
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.4776810304901412e-07,
      "loss": 0.1233,
      "step": 5052
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.4664501473803383e-07,
      "loss": 0.149,
      "step": 5053
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.4552617912674373e-07,
      "loss": 0.1298,
      "step": 5054
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.4441159669803085e-07,
      "loss": 0.1339,
      "step": 5055
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.433012679329515e-07,
      "loss": 0.1254,
      "step": 5056
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.421951933107202e-07,
      "loss": 0.122,
      "step": 5057
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.410933733087161e-07,
      "loss": 0.1538,
      "step": 5058
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3999580840248882e-07,
      "loss": 0.1417,
      "step": 5059
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3890249906574171e-07,
      "loss": 0.1624,
      "step": 5060
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3781344577034948e-07,
      "loss": 0.1531,
      "step": 5061
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.367286489863451e-07,
      "loss": 0.1356,
      "step": 5062
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.3564810918192507e-07,
      "loss": 0.1398,
      "step": 5063
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.3457182682344972e-07,
      "loss": 0.1479,
      "step": 5064
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.3349980237544415e-07,
      "loss": 0.1575,
      "step": 5065
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.3243203630059042e-07,
      "loss": 0.1537,
      "step": 5066
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.3136852905973663e-07,
      "loss": 0.1548,
      "step": 5067
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.303092811118911e-07,
      "loss": 0.1406,
      "step": 5068
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2925429291422487e-07,
      "loss": 0.1321,
      "step": 5069
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.282035649220692e-07,
      "loss": 0.1538,
      "step": 5070
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2715709758891804e-07,
      "loss": 0.155,
      "step": 5071
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2611489136642562e-07,
      "loss": 0.1513,
      "step": 5072
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2507694670440773e-07,
      "loss": 0.1463,
      "step": 5073
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2404326405083934e-07,
      "loss": 0.1234,
      "step": 5074
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.23013843851858e-07,
      "loss": 0.1606,
      "step": 5075
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2198868655176054e-07,
      "loss": 0.1336,
      "step": 5076
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2096779259300307e-07,
      "loss": 0.1628,
      "step": 5077
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1995116241620308e-07,
      "loss": 0.1483,
      "step": 5078
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.189387964601385e-07,
      "loss": 0.1276,
      "step": 5079
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1793069516174427e-07,
      "loss": 0.1546,
      "step": 5080
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1692685895611566e-07,
      "loss": 0.1486,
      "step": 5081
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1592728827650835e-07,
      "loss": 0.1299,
      "step": 5082
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.149319835543361e-07,
      "loss": 0.1305,
      "step": 5083
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1394094521917199e-07,
      "loss": 0.142,
      "step": 5084
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1295417369874606e-07,
      "loss": 0.1508,
      "step": 5085
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1197166941894988e-07,
      "loss": 0.134,
      "step": 5086
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1099343280382868e-07,
      "loss": 0.1393,
      "step": 5087
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.100194642755914e-07,
      "loss": 0.1638,
      "step": 5088
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0904976425460068e-07,
      "loss": 0.136,
      "step": 5089
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.080843331593795e-07,
      "loss": 0.1324,
      "step": 5090
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0712317140660567e-07,
      "loss": 0.1177,
      "step": 5091
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0616627941111623e-07,
      "loss": 0.1488,
      "step": 5092
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0521365758590418e-07,
      "loss": 0.1508,
      "step": 5093
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0426530634212173e-07,
      "loss": 0.1474,
      "step": 5094
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.033212260890759e-07,
      "loss": 0.1509,
      "step": 5095
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0238141723423079e-07,
      "loss": 0.1443,
      "step": 5096
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0144588018320745e-07,
      "loss": 0.1449,
      "step": 5097
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.0051461533978179e-07,
      "loss": 0.1168,
      "step": 5098
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.958762310588788e-08,
      "loss": 0.1499,
      "step": 5099
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.866490388161564e-08,
      "loss": 0.1502,
      "step": 5100
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.774645806520877e-08,
      "loss": 0.1482,
      "step": 5101
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.683228605306682e-08,
      "loss": 0.1323,
      "step": 5102
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.592238823974754e-08,
      "loss": 0.1546,
      "step": 5103
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.50167650179601e-08,
      "loss": 0.1289,
      "step": 5104
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.411541677857294e-08,
      "loss": 0.1266,
      "step": 5105
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.321834391060602e-08,
      "loss": 0.1603,
      "step": 5106
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.232554680123518e-08,
      "loss": 0.1329,
      "step": 5107
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.143702583579106e-08,
      "loss": 0.1238,
      "step": 5108
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.055278139775803e-08,
      "loss": 0.1602,
      "step": 5109
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.967281386877636e-08,
      "loss": 0.1417,
      "step": 5110
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.87971236286389e-08,
      "loss": 0.1156,
      "step": 5111
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.792571105529113e-08,
      "loss": 0.146,
      "step": 5112
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.705857652483662e-08,
      "loss": 0.1338,
      "step": 5113
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.619572041152935e-08,
      "loss": 0.1427,
      "step": 5114
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.533714308777475e-08,
      "loss": 0.1546,
      "step": 5115
    },
    {
      "epoch": 1.93,
      "learning_rate": 8.44828449241375e-08,
      "loss": 0.1412,
      "step": 5116
    },
    {
      "epoch": 1.93,
      "learning_rate": 8.363282628932934e-08,
      "loss": 0.1214,
      "step": 5117
    },
    {
      "epoch": 1.93,
      "learning_rate": 8.278708755021902e-08,
      "loss": 0.136,
      "step": 5118
    },
    {
      "epoch": 1.93,
      "learning_rate": 8.19456290718279e-08,
      "loss": 0.1409,
      "step": 5119
    },
    {
      "epoch": 1.93,
      "learning_rate": 8.110845121732546e-08,
      "loss": 0.1442,
      "step": 5120
    },
    {
      "epoch": 1.93,
      "learning_rate": 8.027555434803935e-08,
      "loss": 0.1415,
      "step": 5121
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.944693882344645e-08,
      "loss": 0.1216,
      "step": 5122
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.862260500117625e-08,
      "loss": 0.1449,
      "step": 5123
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.780255323701081e-08,
      "loss": 0.1603,
      "step": 5124
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.698678388488368e-08,
      "loss": 0.1417,
      "step": 5125
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.6175297296881e-08,
      "loss": 0.1397,
      "step": 5126
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.536809382323817e-08,
      "loss": 0.1645,
      "step": 5127
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.45651738123454e-08,
      "loss": 0.1382,
      "step": 5128
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.376653761074104e-08,
      "loss": 0.1205,
      "step": 5129
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.297218556311602e-08,
      "loss": 0.1637,
      "step": 5130
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.218211801231167e-08,
      "loss": 0.1323,
      "step": 5131
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.139633529932077e-08,
      "loss": 0.1461,
      "step": 5132
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.061483776328649e-08,
      "loss": 0.1293,
      "step": 5133
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.983762574150344e-08,
      "loss": 0.1375,
      "step": 5134
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.906469956941441e-08,
      "loss": 0.1493,
      "step": 5135
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.82960595806137e-08,
      "loss": 0.1228,
      "step": 5136
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.7531706106847e-08,
      "loss": 0.1514,
      "step": 5137
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.677163947800602e-08,
      "loss": 0.0989,
      "step": 5138
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.601586002213834e-08,
      "loss": 0.1391,
      "step": 5139
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.526436806543523e-08,
      "loss": 0.177,
      "step": 5140
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.451716393224061e-08,
      "loss": 0.1406,
      "step": 5141
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.37742479450465e-08,
      "loss": 0.1298,
      "step": 5142
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.30356204244964e-08,
      "loss": 0.1222,
      "step": 5143
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.230128168938088e-08,
      "loss": 0.1478,
      "step": 5144
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.157123205663862e-08,
      "loss": 0.1159,
      "step": 5145
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.08454718413598e-08,
      "loss": 0.155,
      "step": 5146
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.012400135678054e-08,
      "loss": 0.1447,
      "step": 5147
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.940682091428951e-08,
      "loss": 0.1448,
      "step": 5148
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.869393082341801e-08,
      "loss": 0.1359,
      "step": 5149
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.798533139185103e-08,
      "loss": 0.1597,
      "step": 5150
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.728102292541837e-08,
      "loss": 0.124,
      "step": 5151
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.6581005728099106e-08,
      "loss": 0.1513,
      "step": 5152
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.588528010202044e-08,
      "loss": 0.1281,
      "step": 5153
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.519384634745773e-08,
      "loss": 0.121,
      "step": 5154
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.4506704762831154e-08,
      "loss": 0.1386,
      "step": 5155
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.382385564471126e-08,
      "loss": 0.1452,
      "step": 5156
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.3145299287816756e-08,
      "loss": 0.1494,
      "step": 5157
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.247103598500891e-08,
      "loss": 0.1448,
      "step": 5158
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.1801066027301615e-08,
      "loss": 0.139,
      "step": 5159
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.1135389703853564e-08,
      "loss": 0.1471,
      "step": 5160
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.0474007301968276e-08,
      "loss": 0.1706,
      "step": 5161
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.9816919107098514e-08,
      "loss": 0.131,
      "step": 5162
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.91641254028441e-08,
      "loss": 0.1363,
      "step": 5163
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.851562647094965e-08,
      "loss": 0.1223,
      "step": 5164
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.787142259130573e-08,
      "loss": 0.1403,
      "step": 5165
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.7231514041951033e-08,
      "loss": 0.136,
      "step": 5166
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.65959010990702e-08,
      "loss": 0.1487,
      "step": 5167
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.596458403699156e-08,
      "loss": 0.1574,
      "step": 5168
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.533756312819271e-08,
      "loss": 0.1415,
      "step": 5169
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.4714838643293844e-08,
      "loss": 0.1601,
      "step": 5170
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.409641085106331e-08,
      "loss": 0.1359,
      "step": 5171
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.348228001841426e-08,
      "loss": 0.1185,
      "step": 5172
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.2872446410404667e-08,
      "loss": 0.1349,
      "step": 5173
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.226691029023733e-08,
      "loss": 0.1422,
      "step": 5174
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.166567191926319e-08,
      "loss": 0.1296,
      "step": 5175
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.1068731556973554e-08,
      "loss": 0.1525,
      "step": 5176
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.047608946101011e-08,
      "loss": 0.1281,
      "step": 5177
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.988774588715383e-08,
      "loss": 0.1666,
      "step": 5178
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.930370108933601e-08,
      "loss": 0.1372,
      "step": 5179
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.872395531962836e-08,
      "loss": 0.1344,
      "step": 5180
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.81485088282485e-08,
      "loss": 0.1576,
      "step": 5181
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.757736186355998e-08,
      "loss": 0.1691,
      "step": 5182
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.701051467206784e-08,
      "loss": 0.1322,
      "step": 5183
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.6447967498424166e-08,
      "loss": 0.1481,
      "step": 5184
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.58897205854225e-08,
      "loss": 0.1396,
      "step": 5185
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.533577417400347e-08,
      "loss": 0.157,
      "step": 5186
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.478612850324803e-08,
      "loss": 0.1395,
      "step": 5187
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.42407838103842e-08,
      "loss": 0.1365,
      "step": 5188
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.3699740330781494e-08,
      "loss": 0.1146,
      "step": 5189
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.316299829795311e-08,
      "loss": 0.125,
      "step": 5190
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.2630557943558185e-08,
      "loss": 0.1367,
      "step": 5191
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.210241949739623e-08,
      "loss": 0.1575,
      "step": 5192
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.1578583187412694e-08,
      "loss": 0.137,
      "step": 5193
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.105904923969338e-08,
      "loss": 0.1178,
      "step": 5194
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.054381787847005e-08,
      "loss": 0.1285,
      "step": 5195
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.0032889326114814e-08,
      "loss": 0.1525,
      "step": 5196
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.9526263803144607e-08,
      "loss": 0.1855,
      "step": 5197
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.9023941528218967e-08,
      "loss": 0.1166,
      "step": 5198
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.852592271813892e-08,
      "loss": 0.1466,
      "step": 5199
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.8032207587849192e-08,
      "loss": 0.1396,
      "step": 5200
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.7542796350438218e-08,
      "loss": 0.1393,
      "step": 5201
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.7057689217133696e-08,
      "loss": 0.1504,
      "step": 5202
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.6576886397308156e-08,
      "loss": 0.1583,
      "step": 5203
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.6100388098474482e-08,
      "loss": 0.1707,
      "step": 5204
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.5628194526291506e-08,
      "loss": 0.1377,
      "step": 5205
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.5160305884556202e-08,
      "loss": 0.1642,
      "step": 5206
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.469672237520815e-08,
      "loss": 0.1556,
      "step": 5207
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.423744419832952e-08,
      "loss": 0.1476,
      "step": 5208
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.378247155214619e-08,
      "loss": 0.137,
      "step": 5209
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.3331804633023314e-08,
      "loss": 0.1205,
      "step": 5210
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.2885443635467518e-08,
      "loss": 0.1616,
      "step": 5211
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.2443388752128036e-08,
      "loss": 0.1273,
      "step": 5212
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.2005640173795583e-08,
      "loss": 0.1264,
      "step": 5213
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.1572198089402362e-08,
      "loss": 0.1383,
      "step": 5214
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.1143062686022065e-08,
      "loss": 0.1865,
      "step": 5215
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.0718234148867644e-08,
      "loss": 0.1294,
      "step": 5216
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.0297712661295766e-08,
      "loss": 0.1377,
      "step": 5217
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9881498404802357e-08,
      "loss": 0.1243,
      "step": 5218
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9469591559025946e-08,
      "loss": 0.1259,
      "step": 5219
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9061992301745435e-08,
      "loss": 0.1715,
      "step": 5220
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8658700808880104e-08,
      "loss": 0.1543,
      "step": 5221
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8259717254489605e-08,
      "loss": 0.1429,
      "step": 5222
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.7865041810775085e-08,
      "loss": 0.1768,
      "step": 5223
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.747467464807917e-08,
      "loss": 0.1456,
      "step": 5224
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.7088615934881535e-08,
      "loss": 0.1196,
      "step": 5225
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.6706865837807784e-08,
      "loss": 0.1323,
      "step": 5226
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.6329424521619452e-08,
      "loss": 0.1345,
      "step": 5227
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.5956292149219565e-08,
      "loss": 0.114,
      "step": 5228
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.5587468881652636e-08,
      "loss": 0.1298,
      "step": 5229
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.5222954878101327e-08,
      "loss": 0.1812,
      "step": 5230
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.4862750295889794e-08,
      "loss": 0.1558,
      "step": 5231
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.450685529048368e-08,
      "loss": 0.1317,
      "step": 5232
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.4155270015485667e-08,
      "loss": 0.1354,
      "step": 5233
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.380799462263882e-08,
      "loss": 0.1204,
      "step": 5234
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.3465029261828799e-08,
      "loss": 0.1336,
      "step": 5235
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.3126374081078308e-08,
      "loss": 0.1298,
      "step": 5236
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.2792029226550429e-08,
      "loss": 0.1463,
      "step": 5237
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.2461994842547508e-08,
      "loss": 0.143,
      "step": 5238
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.2136271071513384e-08,
      "loss": 0.1155,
      "step": 5239
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1814858054030043e-08,
      "loss": 0.1258,
      "step": 5240
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1497755928817634e-08,
      "loss": 0.1271,
      "step": 5241
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1184964832738898e-08,
      "loss": 0.1229,
      "step": 5242
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0876484900793627e-08,
      "loss": 0.1526,
      "step": 5243
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0572316266120874e-08,
      "loss": 0.1588,
      "step": 5244
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0272459060000073e-08,
      "loss": 0.1201,
      "step": 5245
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.97691341184992e-09,
      "loss": 0.1436,
      "step": 5246
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.68567944922727e-09,
      "loss": 0.1093,
      "step": 5247
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.398757297827133e-09,
      "loss": 0.1147,
      "step": 5248
    },
    {
      "epoch": 1.98,
      "learning_rate": 9.116147081487115e-09,
      "loss": 0.1314,
      "step": 5249
    },
    {
      "epoch": 1.98,
      "learning_rate": 8.837848922179648e-09,
      "loss": 0.1168,
      "step": 5250
    },
    {
      "epoch": 1.98,
      "learning_rate": 8.563862940018653e-09,
      "loss": 0.1155,
      "step": 5251
    },
    {
      "epoch": 1.98,
      "learning_rate": 8.294189253257313e-09,
      "loss": 0.1362,
      "step": 5252
    },
    {
      "epoch": 1.98,
      "learning_rate": 8.028827978284747e-09,
      "loss": 0.1362,
      "step": 5253
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.767779229631566e-09,
      "loss": 0.1431,
      "step": 5254
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.511043119966533e-09,
      "loss": 0.1286,
      "step": 5255
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.258619760095453e-09,
      "loss": 0.1227,
      "step": 5256
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.010509258965625e-09,
      "loss": 0.1392,
      "step": 5257
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.76671172366139e-09,
      "loss": 0.1261,
      "step": 5258
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.5272272594041345e-09,
      "loss": 0.1667,
      "step": 5259
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.2920559695556215e-09,
      "loss": 0.1462,
      "step": 5260
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.061197955615772e-09,
      "loss": 0.1544,
      "step": 5261
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.834653317222661e-09,
      "loss": 0.1558,
      "step": 5262
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.612422152152519e-09,
      "loss": 0.113,
      "step": 5263
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.394504556320845e-09,
      "loss": 0.1157,
      "step": 5264
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.180900623780183e-09,
      "loss": 0.1284,
      "step": 5265
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.971610446722341e-09,
      "loss": 0.1699,
      "step": 5266
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.766634115476177e-09,
      "loss": 0.162,
      "step": 5267
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.565971718508699e-09,
      "loss": 0.135,
      "step": 5268
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.369623342427298e-09,
      "loss": 0.1341,
      "step": 5269
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.177589071974186e-09,
      "loss": 0.178,
      "step": 5270
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.9898689900330635e-09,
      "loss": 0.1602,
      "step": 5271
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.806463177621345e-09,
      "loss": 0.1091,
      "step": 5272
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.6273717138990416e-09,
      "loss": 0.1189,
      "step": 5273
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.4525946761598816e-09,
      "loss": 0.1195,
      "step": 5274
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.2821321398390784e-09,
      "loss": 0.1338,
      "step": 5275
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.1159841785066704e-09,
      "loss": 0.1093,
      "step": 5276
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.9541508638741834e-09,
      "loss": 0.1472,
      "step": 5277
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.796632265785748e-09,
      "loss": 0.1375,
      "step": 5278
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.643428452229202e-09,
      "loss": 0.1459,
      "step": 5279
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.4945394893260976e-09,
      "loss": 0.1361,
      "step": 5280
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.349965441336144e-09,
      "loss": 0.1336,
      "step": 5281
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.2097063706572053e-09,
      "loss": 0.1669,
      "step": 5282
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.0737623378253026e-09,
      "loss": 0.1287,
      "step": 5283
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.942133401515722e-09,
      "loss": 0.1289,
      "step": 5284
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8148196185363565e-09,
      "loss": 0.1607,
      "step": 5285
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.6918210438376936e-09,
      "loss": 0.1556,
      "step": 5286
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.5731377305050478e-09,
      "loss": 0.1545,
      "step": 5287
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.4587697297618886e-09,
      "loss": 0.1301,
      "step": 5288
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.3487170909698421e-09,
      "loss": 0.1311,
      "step": 5289
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.2429798616275802e-09,
      "loss": 0.1489,
      "step": 5290
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1415580873708198e-09,
      "loss": 0.1406,
      "step": 5291
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.0444518119734348e-09,
      "loss": 0.1525,
      "step": 5292
    },
    {
      "epoch": 1.99,
      "learning_rate": 9.51661077345234e-10,
      "loss": 0.1444,
      "step": 5293
    },
    {
      "epoch": 1.99,
      "learning_rate": 8.631859235364026e-10,
      "loss": 0.1412,
      "step": 5294
    },
    {
      "epoch": 1.99,
      "learning_rate": 7.790263887319516e-10,
      "loss": 0.1239,
      "step": 5295
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.991825092561577e-10,
      "loss": 0.1434,
      "step": 5296
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.236543195670131e-10,
      "loss": 0.1471,
      "step": 5297
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.524418522639963e-10,
      "loss": 0.1377,
      "step": 5298
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.855451380825216e-10,
      "loss": 0.1438,
      "step": 5299
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.2296420589504894e-10,
      "loss": 0.1481,
      "step": 5300
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.646990827110841e-10,
      "loss": 0.1484,
      "step": 5301
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.1074979367828884e-10,
      "loss": 0.1292,
      "step": 5302
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.6111636208137056e-10,
      "loss": 0.1142,
      "step": 5303
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.157988093409724e-10,
      "loss": 0.1447,
      "step": 5304
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.7479715501589334e-10,
      "loss": 0.1238,
      "step": 5305
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.381114168041986e-10,
      "loss": 0.1407,
      "step": 5306
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.0574161053877874e-10,
      "loss": 0.1536,
      "step": 5307
    },
    {
      "epoch": 2.0,
      "learning_rate": 7.768775018845987e-11,
      "loss": 0.1253,
      "step": 5308
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.394984786466495e-11,
      "loss": 0.1172,
      "step": 5309
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.452791381031162e-11,
      "loss": 0.1297,
      "step": 5310
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.942195640758371e-11,
      "loss": 0.1199,
      "step": 5311
    },
    {
      "epoch": 2.0,
      "learning_rate": 8.631982177931265e-12,
      "loss": 0.1217,
      "step": 5312
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.157995777629651e-12,
      "loss": 0.159,
      "step": 5313
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0,
      "loss": 0.1405,
      "step": 5314
    },
    {
      "epoch": 2.0,
      "step": 5314,
      "total_flos": 4.602945792393609e+18,
      "train_loss": 0.19471471927608588,
      "train_runtime": 109696.5668,
      "train_samples_per_second": 3.1,
      "train_steps_per_second": 0.048
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 5314,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 100000.0,
  "total_flos": 4.602945792393609e+18,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}