{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 4089,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.6260162601626018e-07,
      "loss": 1.3881,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.2520325203252037e-07,
      "loss": 1.3847,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.878048780487805e-07,
      "loss": 1.3161,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.504065040650407e-07,
      "loss": 1.361,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.130081300813009e-07,
      "loss": 1.4121,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.75609756097561e-07,
      "loss": 1.3233,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1382113821138213e-06,
      "loss": 1.4575,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3008130081300815e-06,
      "loss": 1.3796,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.4634146341463414e-06,
      "loss": 0.207,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6260162601626018e-06,
      "loss": 1.3267,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.788617886178862e-06,
      "loss": 1.26,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.951219512195122e-06,
      "loss": 1.2246,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.1138211382113824e-06,
      "loss": 1.2048,
      "step": 13
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.2764227642276426e-06,
      "loss": 1.2375,
      "step": 14
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.4390243902439027e-06,
      "loss": 1.2182,
      "step": 15
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.601626016260163e-06,
      "loss": 1.213,
      "step": 16
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.764227642276423e-06,
      "loss": 1.2268,
      "step": 17
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.926829268292683e-06,
      "loss": 1.2122,
      "step": 18
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.0894308943089435e-06,
      "loss": 1.1871,
      "step": 19
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.2520325203252037e-06,
      "loss": 1.1341,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.414634146341464e-06,
      "loss": 1.1182,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.577235772357724e-06,
      "loss": 1.1297,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.7398373983739838e-06,
      "loss": 1.152,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.902439024390244e-06,
      "loss": 1.1474,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.0650406504065046e-06,
      "loss": 1.1088,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.227642276422765e-06,
      "loss": 1.0916,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.390243902439025e-06,
      "loss": 1.0496,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.552845528455285e-06,
      "loss": 1.0622,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.715447154471545e-06,
      "loss": 1.0308,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.8780487804878055e-06,
      "loss": 1.032,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.040650406504065e-06,
      "loss": 1.0401,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.203252032520326e-06,
      "loss": 1.0362,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.365853658536586e-06,
      "loss": 1.0257,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.528455284552846e-06,
      "loss": 1.0427,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.691056910569106e-06,
      "loss": 1.049,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.853658536585366e-06,
      "loss": 1.032,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.016260162601627e-06,
      "loss": 1.0425,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.178861788617887e-06,
      "loss": 1.0491,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.341463414634147e-06,
      "loss": 1.0155,
      "step": 39
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.504065040650407e-06,
      "loss": 0.9523,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.666666666666667e-06,
      "loss": 0.2141,
      "step": 41
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.829268292682928e-06,
      "loss": 1.0642,
      "step": 42
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.991869918699188e-06,
      "loss": 1.0288,
      "step": 43
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.154471544715448e-06,
      "loss": 1.0173,
      "step": 44
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.317073170731707e-06,
      "loss": 0.9765,
      "step": 45
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.4796747967479676e-06,
      "loss": 1.0201,
      "step": 46
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.64227642276423e-06,
      "loss": 0.9567,
      "step": 47
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.804878048780489e-06,
      "loss": 0.9795,
      "step": 48
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.967479674796748e-06,
      "loss": 0.999,
      "step": 49
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.130081300813009e-06,
      "loss": 0.9945,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.292682926829268e-06,
      "loss": 0.9849,
      "step": 51
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.45528455284553e-06,
      "loss": 0.9397,
      "step": 52
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.617886178861789e-06,
      "loss": 1.0358,
      "step": 53
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.78048780487805e-06,
      "loss": 0.9802,
      "step": 54
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.94308943089431e-06,
      "loss": 0.9438,
      "step": 55
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.10569105691057e-06,
      "loss": 1.0267,
      "step": 56
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.268292682926831e-06,
      "loss": 0.9643,
      "step": 57
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.43089430894309e-06,
      "loss": 1.0018,
      "step": 58
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.59349593495935e-06,
      "loss": 0.2024,
      "step": 59
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.756097560975611e-06,
      "loss": 0.9275,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.91869918699187e-06,
      "loss": 1.0207,
      "step": 61
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.008130081300813e-05,
      "loss": 1.0203,
      "step": 62
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.024390243902439e-05,
      "loss": 0.9305,
      "step": 63
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0406504065040652e-05,
      "loss": 0.9629,
      "step": 64
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0569105691056911e-05,
      "loss": 0.9552,
      "step": 65
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0731707317073172e-05,
      "loss": 0.9794,
      "step": 66
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0894308943089431e-05,
      "loss": 0.9567,
      "step": 67
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1056910569105692e-05,
      "loss": 1.003,
      "step": 68
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1219512195121953e-05,
      "loss": 0.2093,
      "step": 69
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1382113821138213e-05,
      "loss": 0.9256,
      "step": 70
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1544715447154474e-05,
      "loss": 0.9688,
      "step": 71
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1707317073170731e-05,
      "loss": 0.9901,
      "step": 72
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1869918699186992e-05,
      "loss": 0.198,
      "step": 73
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2032520325203254e-05,
      "loss": 0.98,
      "step": 74
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2195121951219513e-05,
      "loss": 0.9559,
      "step": 75
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2357723577235774e-05,
      "loss": 0.9648,
      "step": 76
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2520325203252033e-05,
      "loss": 0.9331,
      "step": 77
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2682926829268294e-05,
      "loss": 0.8557,
      "step": 78
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2845528455284555e-05,
      "loss": 0.9468,
      "step": 79
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3008130081300815e-05,
      "loss": 0.9343,
      "step": 80
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3170731707317076e-05,
      "loss": 0.9384,
      "step": 81
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3333333333333333e-05,
      "loss": 0.9239,
      "step": 82
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3495934959349594e-05,
      "loss": 1.011,
      "step": 83
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3658536585365855e-05,
      "loss": 0.9906,
      "step": 84
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3821138211382115e-05,
      "loss": 0.9392,
      "step": 85
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3983739837398376e-05,
      "loss": 0.9442,
      "step": 86
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4146341463414635e-05,
      "loss": 0.9658,
      "step": 87
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4308943089430896e-05,
      "loss": 0.9208,
      "step": 88
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4471544715447157e-05,
      "loss": 0.9988,
      "step": 89
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4634146341463415e-05,
      "loss": 0.2535,
      "step": 90
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4796747967479676e-05,
      "loss": 0.9315,
      "step": 91
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4959349593495935e-05,
      "loss": 0.9862,
      "step": 92
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5121951219512196e-05,
      "loss": 0.9524,
      "step": 93
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.528455284552846e-05,
      "loss": 0.9762,
      "step": 94
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5447154471544717e-05,
      "loss": 0.9369,
      "step": 95
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5609756097560978e-05,
      "loss": 0.9772,
      "step": 96
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5772357723577235e-05,
      "loss": 0.972,
      "step": 97
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5934959349593496e-05,
      "loss": 0.9713,
      "step": 98
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6097560975609757e-05,
      "loss": 0.9746,
      "step": 99
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6260162601626018e-05,
      "loss": 1.0188,
      "step": 100
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.642276422764228e-05,
      "loss": 0.9359,
      "step": 101
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6585365853658537e-05,
      "loss": 0.9412,
      "step": 102
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6747967479674798e-05,
      "loss": 0.9749,
      "step": 103
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.691056910569106e-05,
      "loss": 0.945,
      "step": 104
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7073170731707317e-05,
      "loss": 0.9603,
      "step": 105
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7235772357723578e-05,
      "loss": 0.9789,
      "step": 106
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.739837398373984e-05,
      "loss": 0.9717,
      "step": 107
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.75609756097561e-05,
      "loss": 0.9718,
      "step": 108
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.772357723577236e-05,
      "loss": 0.9287,
      "step": 109
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.788617886178862e-05,
      "loss": 0.9637,
      "step": 110
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.804878048780488e-05,
      "loss": 0.9328,
      "step": 111
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.821138211382114e-05,
      "loss": 0.9774,
      "step": 112
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.83739837398374e-05,
      "loss": 0.9396,
      "step": 113
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8536585365853663e-05,
      "loss": 0.9198,
      "step": 114
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.869918699186992e-05,
      "loss": 0.9172,
      "step": 115
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.886178861788618e-05,
      "loss": 0.225,
      "step": 116
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.902439024390244e-05,
      "loss": 0.2316,
      "step": 117
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.91869918699187e-05,
      "loss": 0.9779,
      "step": 118
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.934959349593496e-05,
      "loss": 0.9471,
      "step": 119
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9512195121951222e-05,
      "loss": 0.9608,
      "step": 120
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9674796747967483e-05,
      "loss": 0.9207,
      "step": 121
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.983739837398374e-05,
      "loss": 0.9713,
      "step": 122
    },
    {
      "epoch": 0.03,
      "learning_rate": 2e-05,
      "loss": 0.9665,
      "step": 123
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999996862640347e-05,
      "loss": 0.9568,
      "step": 124
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999987450563353e-05,
      "loss": 0.9376,
      "step": 125
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999971763774927e-05,
      "loss": 0.9768,
      "step": 126
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999994980228491e-05,
      "loss": 0.2917,
      "step": 127
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999921566107083e-05,
      "loss": 1.0034,
      "step": 128
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999988705525916e-05,
      "loss": 0.9954,
      "step": 129
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999846269762803e-05,
      "loss": 0.9318,
      "step": 130
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999799209643597e-05,
      "loss": 0.9201,
      "step": 131
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999745874931076e-05,
      "loss": 0.9561,
      "step": 132
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999686265658703e-05,
      "loss": 0.9778,
      "step": 133
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999962038186388e-05,
      "loss": 0.9696,
      "step": 134
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999548223587948e-05,
      "loss": 0.966,
      "step": 135
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999469790876187e-05,
      "loss": 0.9193,
      "step": 136
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999938508377781e-05,
      "loss": 0.9585,
      "step": 137
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999294102345965e-05,
      "loss": 0.2509,
      "step": 138
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999196846637744e-05,
      "loss": 0.9472,
      "step": 139
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999093316714175e-05,
      "loss": 0.9341,
      "step": 140
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999898351264021e-05,
      "loss": 0.9696,
      "step": 141
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999886743448476e-05,
      "loss": 0.9278,
      "step": 142
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999874508232065e-05,
      "loss": 0.9516,
      "step": 143
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9998616456224663e-05,
      "loss": 0.9204,
      "step": 144
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.99984815562775e-05,
      "loss": 0.9417,
      "step": 145
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9998340382563814e-05,
      "loss": 0.2682,
      "step": 146
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999819293517218e-05,
      "loss": 0.9355,
      "step": 147
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999803921419512e-05,
      "loss": 0.9339,
      "step": 148
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997879219729088e-05,
      "loss": 0.9716,
      "step": 149
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997712951874484e-05,
      "loss": 0.9528,
      "step": 150
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997540410735626e-05,
      "loss": 0.9302,
      "step": 151
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997361596420783e-05,
      "loss": 0.8877,
      "step": 152
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999717650904216e-05,
      "loss": 0.9514,
      "step": 153
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999698514871589e-05,
      "loss": 0.942,
      "step": 154
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996787515562046e-05,
      "loss": 0.9335,
      "step": 155
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996583609704637e-05,
      "loss": 0.9847,
      "step": 156
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996373431271606e-05,
      "loss": 0.9378,
      "step": 157
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996156980394842e-05,
      "loss": 0.906,
      "step": 158
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9995934257210156e-05,
      "loss": 0.9874,
      "step": 159
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.99957052618573e-05,
      "loss": 1.0117,
      "step": 160
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9995469994479964e-05,
      "loss": 1.0105,
      "step": 161
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9995228455225773e-05,
      "loss": 0.9279,
      "step": 162
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994980644246283e-05,
      "loss": 0.2424,
      "step": 163
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999472656169699e-05,
      "loss": 0.9632,
      "step": 164
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994466207737327e-05,
      "loss": 0.946,
      "step": 165
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999419958253065e-05,
      "loss": 0.9951,
      "step": 166
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999392668624427e-05,
      "loss": 0.9988,
      "step": 167
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999364751904941e-05,
      "loss": 0.9576,
      "step": 168
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999336208112125e-05,
      "loss": 0.2472,
      "step": 169
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999307037263889e-05,
      "loss": 1.0137,
      "step": 170
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9992772393785366e-05,
      "loss": 0.9716,
      "step": 171
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9992468144747656e-05,
      "loss": 0.933,
      "step": 172
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9992157625716662e-05,
      "loss": 0.9285,
      "step": 173
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9991840836887237e-05,
      "loss": 0.9688,
      "step": 174
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9991517778458146e-05,
      "loss": 0.9652,
      "step": 175
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9991188450632102e-05,
      "loss": 0.9406,
      "step": 176
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999085285361575e-05,
      "loss": 0.9195,
      "step": 177
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999051098761967e-05,
      "loss": 0.9883,
      "step": 178
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999016285285837e-05,
      "loss": 0.9518,
      "step": 179
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9989808449550297e-05,
      "loss": 0.9341,
      "step": 180
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9989447777917824e-05,
      "loss": 0.977,
      "step": 181
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.998908083818727e-05,
      "loss": 0.8787,
      "step": 182
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9988707630588876e-05,
      "loss": 0.9947,
      "step": 183
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9988328155356816e-05,
      "loss": 0.969,
      "step": 184
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.99879424127292e-05,
      "loss": 0.9348,
      "step": 185
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9987550402948082e-05,
      "loss": 0.9791,
      "step": 186
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9987152126259425e-05,
      "loss": 0.9271,
      "step": 187
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9986747582913138e-05,
      "loss": 0.9776,
      "step": 188
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9986336773163066e-05,
      "loss": 0.2726,
      "step": 189
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9985919697266975e-05,
      "loss": 0.9318,
      "step": 190
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9985496355486575e-05,
      "loss": 0.95,
      "step": 191
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9985066748087495e-05,
      "loss": 0.9437,
      "step": 192
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.99846308753393e-05,
      "loss": 1.0013,
      "step": 193
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9984188737515495e-05,
      "loss": 0.9574,
      "step": 194
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9983740334893503e-05,
      "loss": 0.9527,
      "step": 195
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9983285667754688e-05,
      "loss": 0.9385,
      "step": 196
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9982824736384342e-05,
      "loss": 0.982,
      "step": 197
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998235754107168e-05,
      "loss": 0.9687,
      "step": 198
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9981884082109858e-05,
      "loss": 0.9254,
      "step": 199
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998140435979596e-05,
      "loss": 0.9256,
      "step": 200
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9980918374430995e-05,
      "loss": 1.0358,
      "step": 201
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998042612631991e-05,
      "loss": 0.2451,
      "step": 202
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997992761577157e-05,
      "loss": 0.9534,
      "step": 203
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9979422843098778e-05,
      "loss": 0.9268,
      "step": 204
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997891180861827e-05,
      "loss": 0.9375,
      "step": 205
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.99783945126507e-05,
      "loss": 0.9596,
      "step": 206
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997787095552066e-05,
      "loss": 0.9345,
      "step": 207
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9977341137556668e-05,
      "loss": 0.2937,
      "step": 208
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997680505909117e-05,
      "loss": 0.9532,
      "step": 209
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9976262720460533e-05,
      "loss": 0.9299,
      "step": 210
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9975714122005065e-05,
      "loss": 0.9334,
      "step": 211
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9975159264069e-05,
      "loss": 0.9203,
      "step": 212
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997459814700049e-05,
      "loss": 0.9763,
      "step": 213
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997403077115162e-05,
      "loss": 0.9418,
      "step": 214
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9973457136878407e-05,
      "loss": 0.8685,
      "step": 215
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9972877244540784e-05,
      "loss": 0.8861,
      "step": 216
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9972291094502623e-05,
      "loss": 0.9691,
      "step": 217
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9971698687131716e-05,
      "loss": 0.9129,
      "step": 218
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9971100022799778e-05,
      "loss": 0.823,
      "step": 219
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997049510188246e-05,
      "loss": 0.95,
      "step": 220
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9969883924759327e-05,
      "loss": 0.9837,
      "step": 221
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9969266491813878e-05,
      "loss": 0.9489,
      "step": 222
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9968642803433534e-05,
      "loss": 0.9586,
      "step": 223
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9968012860009643e-05,
      "loss": 0.8549,
      "step": 224
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996737666193748e-05,
      "loss": 0.9336,
      "step": 225
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9966734209616236e-05,
      "loss": 0.9358,
      "step": 226
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9966085503449032e-05,
      "loss": 0.8984,
      "step": 227
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9965430543842914e-05,
      "loss": 0.9137,
      "step": 228
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9964769331208857e-05,
      "loss": 0.9644,
      "step": 229
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9964101865961742e-05,
      "loss": 1.0038,
      "step": 230
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9963428148520393e-05,
      "loss": 0.9742,
      "step": 231
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9962748179307547e-05,
      "loss": 0.8778,
      "step": 232
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9962061958749864e-05,
      "loss": 0.9248,
      "step": 233
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996136948727793e-05,
      "loss": 0.9587,
      "step": 234
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996067076532625e-05,
      "loss": 0.8925,
      "step": 235
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9959965793333253e-05,
      "loss": 0.9229,
      "step": 236
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9959254571741288e-05,
      "loss": 0.9882,
      "step": 237
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9958537100996627e-05,
      "loss": 0.9772,
      "step": 238
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9957813381549467e-05,
      "loss": 1.01,
      "step": 239
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9957083413853917e-05,
      "loss": 0.913,
      "step": 240
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995634719836801e-05,
      "loss": 0.9577,
      "step": 241
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9955604735553705e-05,
      "loss": 0.258,
      "step": 242
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9954856025876874e-05,
      "loss": 0.9361,
      "step": 243
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995410106980731e-05,
      "loss": 0.9458,
      "step": 244
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995333986781873e-05,
      "loss": 0.9286,
      "step": 245
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9952572420388763e-05,
      "loss": 0.8949,
      "step": 246
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995179872799897e-05,
      "loss": 0.9611,
      "step": 247
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9951018791134805e-05,
      "loss": 0.9935,
      "step": 248
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995023261028567e-05,
      "loss": 0.9629,
      "step": 249
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994944018594487e-05,
      "loss": 0.9557,
      "step": 250
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9948641518609623e-05,
      "loss": 0.2426,
      "step": 251
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9947836608781072e-05,
      "loss": 0.901,
      "step": 252
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994702545696428e-05,
      "loss": 0.9656,
      "step": 253
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9946208063668218e-05,
      "loss": 0.9579,
      "step": 254
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994538442940578e-05,
      "loss": 0.9201,
      "step": 255
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994455455469377e-05,
      "loss": 0.9305,
      "step": 256
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994371844005291e-05,
      "loss": 0.9235,
      "step": 257
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9942876086007847e-05,
      "loss": 0.8641,
      "step": 258
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9942027493087122e-05,
      "loss": 0.9251,
      "step": 259
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9941172661823215e-05,
      "loss": 0.9592,
      "step": 260
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.99403115927525e-05,
      "loss": 0.2485,
      "step": 261
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.993944428641528e-05,
      "loss": 0.26,
      "step": 262
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9938570743355763e-05,
      "loss": 0.922,
      "step": 263
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9937690964122072e-05,
      "loss": 0.9661,
      "step": 264
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9936804949266242e-05,
      "loss": 0.9288,
      "step": 265
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9935912699344228e-05,
      "loss": 0.9711,
      "step": 266
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9935014214915884e-05,
      "loss": 0.8855,
      "step": 267
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9934109496544992e-05,
      "loss": 0.2542,
      "step": 268
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.993319854479923e-05,
      "loss": 0.9569,
      "step": 269
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.99322813602502e-05,
      "loss": 0.9555,
      "step": 270
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.993135794347341e-05,
      "loss": 0.9461,
      "step": 271
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9930428295048272e-05,
      "loss": 0.2415,
      "step": 272
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992949241555812e-05,
      "loss": 0.2451,
      "step": 273
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992855030559019e-05,
      "loss": 0.9212,
      "step": 274
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992760196573563e-05,
      "loss": 0.9341,
      "step": 275
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9926647396589498e-05,
      "loss": 0.9461,
      "step": 276
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9925686598750754e-05,
      "loss": 0.9096,
      "step": 277
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992471957282228e-05,
      "loss": 0.9023,
      "step": 278
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9923746319410852e-05,
      "loss": 0.9592,
      "step": 279
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992276683912716e-05,
      "loss": 0.9137,
      "step": 280
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9921781132585797e-05,
      "loss": 0.9118,
      "step": 281
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9920789200405272e-05,
      "loss": 0.9325,
      "step": 282
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9919791043207995e-05,
      "loss": 0.9068,
      "step": 283
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9918786661620276e-05,
      "loss": 0.9042,
      "step": 284
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991777605627234e-05,
      "loss": 0.9206,
      "step": 285
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9916759227798313e-05,
      "loss": 0.9445,
      "step": 286
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991573617683623e-05,
      "loss": 0.9296,
      "step": 287
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991470690402802e-05,
      "loss": 0.9051,
      "step": 288
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991367141001952e-05,
      "loss": 0.9406,
      "step": 289
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9912629695460486e-05,
      "loss": 0.898,
      "step": 290
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9911581761004556e-05,
      "loss": 0.9454,
      "step": 291
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991052760730928e-05,
      "loss": 0.2553,
      "step": 292
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9909467235036114e-05,
      "loss": 0.9677,
      "step": 293
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.990840064485041e-05,
      "loss": 0.9343,
      "step": 294
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9907327837421416e-05,
      "loss": 0.8837,
      "step": 295
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.99062488134223e-05,
      "loss": 0.9367,
      "step": 296
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9905163573530114e-05,
      "loss": 0.2503,
      "step": 297
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9904072118425812e-05,
      "loss": 0.9502,
      "step": 298
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.990297444879426e-05,
      "loss": 0.9357,
      "step": 299
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9901870565324207e-05,
      "loss": 0.8882,
      "step": 300
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9900760468708312e-05,
      "loss": 0.9196,
      "step": 301
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.989964415964313e-05,
      "loss": 0.9624,
      "step": 302
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9898521638829115e-05,
      "loss": 0.8918,
      "step": 303
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.989739290697061e-05,
      "loss": 0.9209,
      "step": 304
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9896257964775874e-05,
      "loss": 0.9626,
      "step": 305
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.989511681295704e-05,
      "loss": 0.9699,
      "step": 306
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9893969452230157e-05,
      "loss": 0.92,
      "step": 307
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9892815883315158e-05,
      "loss": 0.2617,
      "step": 308
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9891656106935878e-05,
      "loss": 0.9577,
      "step": 309
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9890490123820037e-05,
      "loss": 0.9223,
      "step": 310
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9889317934699264e-05,
      "loss": 0.9579,
      "step": 311
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9888139540309073e-05,
      "loss": 0.9643,
      "step": 312
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9886954941388872e-05,
      "loss": 0.9772,
      "step": 313
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9885764138681964e-05,
      "loss": 0.8645,
      "step": 314
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9884567132935545e-05,
      "loss": 0.9015,
      "step": 315
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9883363924900702e-05,
      "loss": 0.8674,
      "step": 316
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9882154515332414e-05,
      "loss": 0.9285,
      "step": 317
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9880938904989556e-05,
      "loss": 0.9727,
      "step": 318
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987971709463488e-05,
      "loss": 0.9899,
      "step": 319
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9878489085035046e-05,
      "loss": 0.9471,
      "step": 320
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987725487696059e-05,
      "loss": 0.9067,
      "step": 321
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987601447118595e-05,
      "loss": 0.9698,
      "step": 322
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9874767868489435e-05,
      "loss": 0.9191,
      "step": 323
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987351506965326e-05,
      "loss": 0.9271,
      "step": 324
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9872256075463526e-05,
      "loss": 0.9738,
      "step": 325
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9870990886710206e-05,
      "loss": 0.931,
      "step": 326
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986971950418718e-05,
      "loss": 0.9351,
      "step": 327
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.98684419286922e-05,
      "loss": 0.232,
      "step": 328
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9867158161026903e-05,
      "loss": 0.9544,
      "step": 329
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986586820199683e-05,
      "loss": 0.9509,
      "step": 330
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9864572052411383e-05,
      "loss": 1.0073,
      "step": 331
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9863269713083867e-05,
      "loss": 0.966,
      "step": 332
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9861961184831456e-05,
      "loss": 0.23,
      "step": 333
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986064646847522e-05,
      "loss": 0.9144,
      "step": 334
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985932556484011e-05,
      "loss": 0.949,
      "step": 335
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985799847475495e-05,
      "loss": 0.9353,
      "step": 336
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985666519905245e-05,
      "loss": 0.8579,
      "step": 337
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9855325738569207e-05,
      "loss": 0.9065,
      "step": 338
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9853980094145698e-05,
      "loss": 0.9427,
      "step": 339
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9852628266626273e-05,
      "loss": 0.902,
      "step": 340
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9851270256859166e-05,
      "loss": 0.9154,
      "step": 341
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9849906065696486e-05,
      "loss": 0.9596,
      "step": 342
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.984853569399423e-05,
      "loss": 0.8386,
      "step": 343
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9847159142612267e-05,
      "loss": 0.9067,
      "step": 344
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.984577641241435e-05,
      "loss": 0.8298,
      "step": 345
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9844387504268092e-05,
      "loss": 0.9212,
      "step": 346
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9842992419044998e-05,
      "loss": 0.9182,
      "step": 347
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9841591157620446e-05,
      "loss": 0.9815,
      "step": 348
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9840183720873692e-05,
      "loss": 0.8994,
      "step": 349
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9838770109687858e-05,
      "loss": 0.8718,
      "step": 350
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9837350324949947e-05,
      "loss": 0.2844,
      "step": 351
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.983592436755083e-05,
      "loss": 0.9121,
      "step": 352
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9834492238385262e-05,
      "loss": 0.9633,
      "step": 353
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9833053938351864e-05,
      "loss": 0.9424,
      "step": 354
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9831609468353122e-05,
      "loss": 0.9181,
      "step": 355
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9830158829295404e-05,
      "loss": 0.8977,
      "step": 356
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9828702022088943e-05,
      "loss": 0.9212,
      "step": 357
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9827239047647852e-05,
      "loss": 0.9209,
      "step": 358
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.98257699068901e-05,
      "loss": 0.2482,
      "step": 359
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.982429460073753e-05,
      "loss": 0.9768,
      "step": 360
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9822813130115863e-05,
      "loss": 0.9426,
      "step": 361
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.982132549595467e-05,
      "loss": 0.9235,
      "step": 362
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.981983169918741e-05,
      "loss": 0.9048,
      "step": 363
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9818331740751394e-05,
      "loss": 0.9815,
      "step": 364
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.98168256215878e-05,
      "loss": 0.9199,
      "step": 365
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.981531334264168e-05,
      "loss": 0.945,
      "step": 366
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9813794904861947e-05,
      "loss": 0.9165,
      "step": 367
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9812270309201376e-05,
      "loss": 0.9144,
      "step": 368
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.981073955661661e-05,
      "loss": 0.9337,
      "step": 369
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.980920264806815e-05,
      "loss": 0.895,
      "step": 370
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9807659584520365e-05,
      "loss": 0.9117,
      "step": 371
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9806110366941484e-05,
      "loss": 0.8974,
      "step": 372
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9804554996303596e-05,
      "loss": 0.9297,
      "step": 373
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9802993473582657e-05,
      "loss": 0.8984,
      "step": 374
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9801425799758475e-05,
      "loss": 0.9437,
      "step": 375
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9799851975814718e-05,
      "loss": 0.9695,
      "step": 376
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9798272002738925e-05,
      "loss": 0.8712,
      "step": 377
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.979668588152248e-05,
      "loss": 0.8921,
      "step": 378
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.979509361316063e-05,
      "loss": 0.9077,
      "step": 379
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9793495198652475e-05,
      "loss": 0.9454,
      "step": 380
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9791890639000976e-05,
      "loss": 0.9332,
      "step": 381
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9790279935212958e-05,
      "loss": 0.8922,
      "step": 382
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9788663088299086e-05,
      "loss": 0.9442,
      "step": 383
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.978704009927388e-05,
      "loss": 0.9009,
      "step": 384
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.978541096915573e-05,
      "loss": 0.9348,
      "step": 385
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.978377569896687e-05,
      "loss": 0.9595,
      "step": 386
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9782134289733374e-05,
      "loss": 0.9507,
      "step": 387
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.978048674248519e-05,
      "loss": 0.9451,
      "step": 388
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.977883305825611e-05,
      "loss": 0.9605,
      "step": 389
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9777173238083764e-05,
      "loss": 0.9303,
      "step": 390
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9775507283009655e-05,
      "loss": 0.9216,
      "step": 391
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.977383519407911e-05,
      "loss": 0.9248,
      "step": 392
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.977215697234133e-05,
      "loss": 0.9266,
      "step": 393
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9770472618849342e-05,
      "loss": 0.9096,
      "step": 394
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9768782134660038e-05,
      "loss": 0.2661,
      "step": 395
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9767085520834147e-05,
      "loss": 0.9194,
      "step": 396
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9765382778436245e-05,
      "loss": 0.8902,
      "step": 397
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9763673908534757e-05,
      "loss": 0.8572,
      "step": 398
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9761958912201948e-05,
      "loss": 0.9257,
      "step": 399
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9760237790513934e-05,
      "loss": 0.9511,
      "step": 400
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.975851054455067e-05,
      "loss": 0.9588,
      "step": 401
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.975677717539595e-05,
      "loss": 0.9013,
      "step": 402
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9755037684137416e-05,
      "loss": 0.9308,
      "step": 403
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9753292071866555e-05,
      "loss": 0.9053,
      "step": 404
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9751540339678686e-05,
      "loss": 0.9259,
      "step": 405
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9749782488672968e-05,
      "loss": 0.2457,
      "step": 406
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9748018519952412e-05,
      "loss": 0.8709,
      "step": 407
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9746248434623853e-05,
      "loss": 0.9184,
      "step": 408
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9744472233797967e-05,
      "loss": 0.9014,
      "step": 409
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9742689918589277e-05,
      "loss": 0.9398,
      "step": 410
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9740901490116132e-05,
      "loss": 0.9624,
      "step": 411
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.973910694950072e-05,
      "loss": 0.9447,
      "step": 412
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.973730629786907e-05,
      "loss": 0.9541,
      "step": 413
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9735499536351032e-05,
      "loss": 0.9262,
      "step": 414
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9733686666080308e-05,
      "loss": 0.2494,
      "step": 415
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.973186768819441e-05,
      "loss": 0.9576,
      "step": 416
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9730042603834715e-05,
      "loss": 0.8897,
      "step": 417
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.972821141414639e-05,
      "loss": 0.8459,
      "step": 418
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9726374120278472e-05,
      "loss": 0.9106,
      "step": 419
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9724530723383802e-05,
      "loss": 0.9438,
      "step": 420
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9722681224619065e-05,
      "loss": 0.8793,
      "step": 421
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9720825625144767e-05,
      "loss": 0.863,
      "step": 422
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9718963926125246e-05,
      "loss": 0.9225,
      "step": 423
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9717096128728663e-05,
      "loss": 0.9174,
      "step": 424
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9715222234127012e-05,
      "loss": 0.8383,
      "step": 425
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9713342243496105e-05,
      "loss": 0.9266,
      "step": 426
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9711456158015588e-05,
      "loss": 0.9335,
      "step": 427
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9709563978868927e-05,
      "loss": 0.9073,
      "step": 428
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9707665707243406e-05,
      "loss": 0.9652,
      "step": 429
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9705761344330142e-05,
      "loss": 0.926,
      "step": 430
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9703850891324062e-05,
      "loss": 0.9074,
      "step": 431
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9701934349423935e-05,
      "loss": 0.8578,
      "step": 432
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9700011719832327e-05,
      "loss": 0.888,
      "step": 433
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9698083003755633e-05,
      "loss": 0.9271,
      "step": 434
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9696148202404073e-05,
      "loss": 0.942,
      "step": 435
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.969420731699168e-05,
      "loss": 0.9173,
      "step": 436
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9692260348736307e-05,
      "loss": 0.8881,
      "step": 437
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9690307298859617e-05,
      "loss": 0.9119,
      "step": 438
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9688348168587096e-05,
      "loss": 0.2741,
      "step": 439
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9686382959148043e-05,
      "loss": 0.9012,
      "step": 440
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.968441167177557e-05,
      "loss": 0.9175,
      "step": 441
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.968243430770661e-05,
      "loss": 0.9407,
      "step": 442
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9680450868181902e-05,
      "loss": 0.956,
      "step": 443
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9678461354445993e-05,
      "loss": 0.9159,
      "step": 444
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9676465767747256e-05,
      "loss": 0.8864,
      "step": 445
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9674464109337856e-05,
      "loss": 0.8912,
      "step": 446
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9672456380473784e-05,
      "loss": 0.9113,
      "step": 447
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.967044258241483e-05,
      "loss": 0.9393,
      "step": 448
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9668422716424597e-05,
      "loss": 0.9253,
      "step": 449
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9666396783770497e-05,
      "loss": 0.2304,
      "step": 450
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.966436478572374e-05,
      "loss": 0.9606,
      "step": 451
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9662326723559353e-05,
      "loss": 0.8697,
      "step": 452
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.966028259855616e-05,
      "loss": 0.8404,
      "step": 453
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9658232411996793e-05,
      "loss": 0.9029,
      "step": 454
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9656176165167685e-05,
      "loss": 0.9623,
      "step": 455
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.965411385935908e-05,
      "loss": 0.9223,
      "step": 456
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9652045495865004e-05,
      "loss": 0.8947,
      "step": 457
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9649971075983308e-05,
      "loss": 0.9643,
      "step": 458
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.964789060101563e-05,
      "loss": 0.9379,
      "step": 459
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.964580407226741e-05,
      "loss": 0.9348,
      "step": 460
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.964371149104788e-05,
      "loss": 0.9152,
      "step": 461
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9641612858670083e-05,
      "loss": 0.9182,
      "step": 462
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9639508176450853e-05,
      "loss": 0.9536,
      "step": 463
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.963739744571081e-05,
      "loss": 0.9165,
      "step": 464
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.963528066777439e-05,
      "loss": 0.8869,
      "step": 465
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.96331578439698e-05,
      "loss": 0.8726,
      "step": 466
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.963102897562906e-05,
      "loss": 0.9593,
      "step": 467
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9628894064087974e-05,
      "loss": 0.8721,
      "step": 468
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9626753110686137e-05,
      "loss": 0.9115,
      "step": 469
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.962460611676694e-05,
      "loss": 0.8846,
      "step": 470
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.962245308367756e-05,
      "loss": 0.9289,
      "step": 471
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.962029401276896e-05,
      "loss": 0.9064,
      "step": 472
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9618128905395907e-05,
      "loss": 0.9068,
      "step": 473
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9615957762916937e-05,
      "loss": 0.8946,
      "step": 474
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9613780586694383e-05,
      "loss": 0.8881,
      "step": 475
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.961159737809436e-05,
      "loss": 0.9062,
      "step": 476
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9609408138486777e-05,
      "loss": 0.2797,
      "step": 477
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9607212869245313e-05,
      "loss": 0.8967,
      "step": 478
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.960501157174744e-05,
      "loss": 0.2756,
      "step": 479
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.960280424737441e-05,
      "loss": 0.9868,
      "step": 480
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9600590897511262e-05,
      "loss": 0.9227,
      "step": 481
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9598371523546798e-05,
      "loss": 0.8696,
      "step": 482
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.959614612687363e-05,
      "loss": 0.2526,
      "step": 483
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9593914708888122e-05,
      "loss": 0.8778,
      "step": 484
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.959167727099043e-05,
      "loss": 0.912,
      "step": 485
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9589433814584473e-05,
      "loss": 0.9008,
      "step": 486
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9587184341077973e-05,
      "loss": 0.8335,
      "step": 487
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.95849288518824e-05,
      "loss": 0.9375,
      "step": 488
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9582667348413014e-05,
      "loss": 0.8912,
      "step": 489
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9580399832088847e-05,
      "loss": 0.2687,
      "step": 490
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.95781263043327e-05,
      "loss": 0.8643,
      "step": 491
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9575846766571142e-05,
      "loss": 0.9138,
      "step": 492
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9573561220234528e-05,
      "loss": 0.8992,
      "step": 493
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9571269666756972e-05,
      "loss": 0.9098,
      "step": 494
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9568972107576358e-05,
      "loss": 0.9332,
      "step": 495
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9566668544134337e-05,
      "loss": 0.9221,
      "step": 496
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9564358977876336e-05,
      "loss": 0.8753,
      "step": 497
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.956204341025154e-05,
      "loss": 0.2768,
      "step": 498
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9559721842712902e-05,
      "loss": 0.9434,
      "step": 499
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.955739427671714e-05,
      "loss": 0.9054,
      "step": 500
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.955506071372474e-05,
      "loss": 0.8568,
      "step": 501
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9552721155199944e-05,
      "loss": 0.8813,
      "step": 502
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9550375602610765e-05,
      "loss": 0.8877,
      "step": 503
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.954802405742896e-05,
      "loss": 0.8815,
      "step": 504
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.954566652113007e-05,
      "loss": 0.925,
      "step": 505
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9543302995193377e-05,
      "loss": 0.9067,
      "step": 506
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9540933481101925e-05,
      "loss": 0.9474,
      "step": 507
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.953855798034252e-05,
      "loss": 0.9208,
      "step": 508
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9536176494405724e-05,
      "loss": 0.8502,
      "step": 509
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9533789024785847e-05,
      "loss": 0.2714,
      "step": 510
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.953139557298097e-05,
      "loss": 0.9742,
      "step": 511
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.95289961404929e-05,
      "loss": 0.9443,
      "step": 512
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9526590728827232e-05,
      "loss": 0.921,
      "step": 513
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.952417933949328e-05,
      "loss": 0.9051,
      "step": 514
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.952176197400413e-05,
      "loss": 0.9096,
      "step": 515
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9519338633876613e-05,
      "loss": 0.2593,
      "step": 516
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9516909320631302e-05,
      "loss": 0.9116,
      "step": 517
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9514474035792522e-05,
      "loss": 0.2697,
      "step": 518
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9512032780888347e-05,
      "loss": 0.913,
      "step": 519
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9509585557450603e-05,
      "loss": 0.9357,
      "step": 520
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9507132367014842e-05,
      "loss": 0.9205,
      "step": 521
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9504673211120383e-05,
      "loss": 0.8902,
      "step": 522
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.950220809131027e-05,
      "loss": 0.9032,
      "step": 523
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9499737009131303e-05,
      "loss": 0.9473,
      "step": 524
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9497259966134008e-05,
      "loss": 0.9462,
      "step": 525
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9494776963872668e-05,
      "loss": 0.9039,
      "step": 526
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9492288003905293e-05,
      "loss": 0.923,
      "step": 527
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9489793087793635e-05,
      "loss": 0.8763,
      "step": 528
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9487292217103188e-05,
      "loss": 0.8157,
      "step": 529
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9484785393403178e-05,
      "loss": 0.8935,
      "step": 530
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9482272618266558e-05,
      "loss": 0.8811,
      "step": 531
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9479753893270034e-05,
      "loss": 0.8828,
      "step": 532
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.947722921999403e-05,
      "loss": 0.9628,
      "step": 533
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9474698600022706e-05,
      "loss": 0.9388,
      "step": 534
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9472162034943964e-05,
      "loss": 0.9116,
      "step": 535
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9469619526349417e-05,
      "loss": 0.9057,
      "step": 536
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9467071075834422e-05,
      "loss": 0.8868,
      "step": 537
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9464516684998062e-05,
      "loss": 0.8619,
      "step": 538
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.946195635544314e-05,
      "loss": 0.9006,
      "step": 539
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9459390088776196e-05,
      "loss": 0.8589,
      "step": 540
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.945681788660749e-05,
      "loss": 0.894,
      "step": 541
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9454239750551006e-05,
      "loss": 0.9359,
      "step": 542
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.945165568222445e-05,
      "loss": 0.9045,
      "step": 543
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9449065683249258e-05,
      "loss": 0.9196,
      "step": 544
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9446469755250574e-05,
      "loss": 0.8828,
      "step": 545
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9443867899857274e-05,
      "loss": 0.9533,
      "step": 546
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9441260118701947e-05,
      "loss": 0.2993,
      "step": 547
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9438646413420908e-05,
      "loss": 0.9239,
      "step": 548
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9436026785654175e-05,
      "loss": 0.8558,
      "step": 549
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.94334012370455e-05,
      "loss": 0.8902,
      "step": 550
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9430769769242338e-05,
      "loss": 0.8766,
      "step": 551
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9428132383895856e-05,
      "loss": 0.9029,
      "step": 552
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9425489082660945e-05,
      "loss": 0.8963,
      "step": 553
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.94228398671962e-05,
      "loss": 0.9064,
      "step": 554
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9420184739163932e-05,
      "loss": 0.8749,
      "step": 555
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9417523700230155e-05,
      "loss": 0.877,
      "step": 556
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9414856752064598e-05,
      "loss": 0.9126,
      "step": 557
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.94121838963407e-05,
      "loss": 0.8869,
      "step": 558
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9409505134735598e-05,
      "loss": 0.9064,
      "step": 559
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.940682046893014e-05,
      "loss": 0.9484,
      "step": 560
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9404129900608884e-05,
      "loss": 0.9315,
      "step": 561
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9401433431460078e-05,
      "loss": 0.8713,
      "step": 562
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9398731063175687e-05,
      "loss": 0.9289,
      "step": 563
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9396022797451364e-05,
      "loss": 0.854,
      "step": 564
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9393308635986477e-05,
      "loss": 0.8766,
      "step": 565
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9390588580484087e-05,
      "loss": 0.887,
      "step": 566
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9387862632650945e-05,
      "loss": 0.9135,
      "step": 567
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9385130794197513e-05,
      "loss": 0.9055,
      "step": 568
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.938239306683794e-05,
      "loss": 0.9091,
      "step": 569
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9379649452290075e-05,
      "loss": 0.8895,
      "step": 570
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9376899952275457e-05,
      "loss": 0.9305,
      "step": 571
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.937414456851932e-05,
      "loss": 0.9131,
      "step": 572
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9371383302750594e-05,
      "loss": 0.9019,
      "step": 573
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.936861615670189e-05,
      "loss": 0.8972,
      "step": 574
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.936584313210952e-05,
      "loss": 0.881,
      "step": 575
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9363064230713475e-05,
      "loss": 0.9666,
      "step": 576
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9360279454257438e-05,
      "loss": 0.9031,
      "step": 577
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9357488804488782e-05,
      "loss": 0.8942,
      "step": 578
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9354692283158555e-05,
      "loss": 0.2364,
      "step": 579
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9351889892021503e-05,
      "loss": 0.8962,
      "step": 580
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.934908163283604e-05,
      "loss": 0.9025,
      "step": 581
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9346267507364276e-05,
      "loss": 0.8727,
      "step": 582
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9343447517371994e-05,
      "loss": 0.8897,
      "step": 583
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9340621664628657e-05,
      "loss": 0.8567,
      "step": 584
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.933778995090741e-05,
      "loss": 0.8799,
      "step": 585
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9334952377985074e-05,
      "loss": 0.9226,
      "step": 586
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9332108947642144e-05,
      "loss": 0.9099,
      "step": 587
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932925966166279e-05,
      "loss": 0.8568,
      "step": 588
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932640452183487e-05,
      "loss": 0.9512,
      "step": 589
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9323543529949894e-05,
      "loss": 0.8809,
      "step": 590
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932067668780306e-05,
      "loss": 0.9254,
      "step": 591
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9317803997193225e-05,
      "loss": 0.9257,
      "step": 592
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9314925459922928e-05,
      "loss": 0.8782,
      "step": 593
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9312041077798366e-05,
      "loss": 0.92,
      "step": 594
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.930915085262941e-05,
      "loss": 0.2834,
      "step": 595
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9306254786229598e-05,
      "loss": 0.8712,
      "step": 596
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.930335288041612e-05,
      "loss": 0.9178,
      "step": 597
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.930044513700985e-05,
      "loss": 0.2669,
      "step": 598
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9297531557835314e-05,
      "loss": 0.8987,
      "step": 599
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9294612144720698e-05,
      "loss": 0.2647,
      "step": 600
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.929168689949786e-05,
      "loss": 0.9069,
      "step": 601
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9288755824002295e-05,
      "loss": 0.8846,
      "step": 602
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.928581892007318e-05,
      "loss": 0.9749,
      "step": 603
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9282876189553338e-05,
      "loss": 0.9374,
      "step": 604
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9279927634289254e-05,
      "loss": 0.9029,
      "step": 605
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9276973256131055e-05,
      "loss": 0.2841,
      "step": 606
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9274013056932538e-05,
      "loss": 0.9052,
      "step": 607
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.927104703855114e-05,
      "loss": 0.9466,
      "step": 608
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.926807520284796e-05,
      "loss": 0.9088,
      "step": 609
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9265097551687732e-05,
      "loss": 0.9288,
      "step": 610
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.926211408693886e-05,
      "loss": 0.9014,
      "step": 611
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9259124810473374e-05,
      "loss": 0.8849,
      "step": 612
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.925612972416697e-05,
      "loss": 0.8512,
      "step": 613
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9253128829898972e-05,
      "loss": 0.884,
      "step": 614
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9250122129552365e-05,
      "loss": 0.8611,
      "step": 615
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9247109625013762e-05,
      "loss": 0.899,
      "step": 616
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9244091318173432e-05,
      "loss": 0.8892,
      "step": 617
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9241067210925272e-05,
      "loss": 0.8996,
      "step": 618
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9238037305166827e-05,
      "loss": 0.9012,
      "step": 619
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9235001602799278e-05,
      "loss": 0.8595,
      "step": 620
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.923196010572744e-05,
      "loss": 0.9066,
      "step": 621
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9228912815859774e-05,
      "loss": 0.9291,
      "step": 622
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.922585973510836e-05,
      "loss": 0.8758,
      "step": 623
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9222800865388925e-05,
      "loss": 0.8942,
      "step": 624
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9219736208620827e-05,
      "loss": 0.8771,
      "step": 625
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9216665766727047e-05,
      "loss": 0.8762,
      "step": 626
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9213589541634203e-05,
      "loss": 0.8666,
      "step": 627
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.921050753527254e-05,
      "loss": 0.8381,
      "step": 628
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9207419749575932e-05,
      "loss": 0.866,
      "step": 629
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9204326186481872e-05,
      "loss": 0.855,
      "step": 630
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.920122684793149e-05,
      "loss": 0.8794,
      "step": 631
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.919812173586953e-05,
      "loss": 0.9339,
      "step": 632
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9195010852244367e-05,
      "loss": 0.862,
      "step": 633
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9191894199007994e-05,
      "loss": 0.9318,
      "step": 634
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9188771778116018e-05,
      "loss": 0.8921,
      "step": 635
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.918564359152767e-05,
      "loss": 0.2817,
      "step": 636
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9182509641205804e-05,
      "loss": 0.8661,
      "step": 637
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9179369929116887e-05,
      "loss": 0.9898,
      "step": 638
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.917622445723099e-05,
      "loss": 0.9312,
      "step": 639
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9173073227521822e-05,
      "loss": 0.8761,
      "step": 640
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9169916241966686e-05,
      "loss": 0.913,
      "step": 641
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9166753502546496e-05,
      "loss": 0.9015,
      "step": 642
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9163585011245785e-05,
      "loss": 0.9222,
      "step": 643
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9160410770052697e-05,
      "loss": 0.8812,
      "step": 644
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.915723078095898e-05,
      "loss": 0.9422,
      "step": 645
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9154045045959977e-05,
      "loss": 0.8552,
      "step": 646
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9150853567054658e-05,
      "loss": 0.857,
      "step": 647
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.914765634624558e-05,
      "loss": 0.962,
      "step": 648
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9144453385538915e-05,
      "loss": 0.8725,
      "step": 649
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9141244686944428e-05,
      "loss": 0.8847,
      "step": 650
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9138030252475487e-05,
      "loss": 0.9056,
      "step": 651
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9134810084149063e-05,
      "loss": 0.8735,
      "step": 652
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9131584183985715e-05,
      "loss": 0.8867,
      "step": 653
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9128352554009607e-05,
      "loss": 0.8235,
      "step": 654
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.91251151962485e-05,
      "loss": 0.9066,
      "step": 655
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.912187211273374e-05,
      "loss": 0.9149,
      "step": 656
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.911862330550027e-05,
      "loss": 0.8638,
      "step": 657
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9115368776586632e-05,
      "loss": 0.861,
      "step": 658
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.911210852803495e-05,
      "loss": 0.8805,
      "step": 659
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9108842561890934e-05,
      "loss": 0.9031,
      "step": 660
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9105570880203886e-05,
      "loss": 0.874,
      "step": 661
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.91022934850267e-05,
      "loss": 0.3035,
      "step": 662
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9099010378415847e-05,
      "loss": 0.903,
      "step": 663
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.909572156243138e-05,
      "loss": 0.8774,
      "step": 664
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9092427039136945e-05,
      "loss": 0.8411,
      "step": 665
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.908912681059976e-05,
      "loss": 0.8761,
      "step": 666
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9085820878890626e-05,
      "loss": 0.9226,
      "step": 667
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.908250924608392e-05,
      "loss": 0.8415,
      "step": 668
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9079191914257606e-05,
      "loss": 0.9032,
      "step": 669
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9075868885493207e-05,
      "loss": 0.9174,
      "step": 670
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9072540161875837e-05,
      "loss": 0.888,
      "step": 671
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9069205745494172e-05,
      "loss": 0.9093,
      "step": 672
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.906586563844047e-05,
      "loss": 0.9308,
      "step": 673
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.906251984281055e-05,
      "loss": 0.2827,
      "step": 674
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9059168360703805e-05,
      "loss": 0.9063,
      "step": 675
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.90558111942232e-05,
      "loss": 0.9113,
      "step": 676
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.905244834547526e-05,
      "loss": 0.9407,
      "step": 677
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9049079816570078e-05,
      "loss": 0.9235,
      "step": 678
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.904570560962131e-05,
      "loss": 0.8764,
      "step": 679
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9042325726746182e-05,
      "loss": 0.8925,
      "step": 680
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9038940170065467e-05,
      "loss": 0.8944,
      "step": 681
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.903554894170351e-05,
      "loss": 0.8904,
      "step": 682
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9032152043788218e-05,
      "loss": 0.938,
      "step": 683
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.902874947845104e-05,
      "loss": 0.9447,
      "step": 684
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9025341247826995e-05,
      "loss": 0.8923,
      "step": 685
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.902192735405465e-05,
      "loss": 0.8687,
      "step": 686
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9018507799276132e-05,
      "loss": 0.9211,
      "step": 687
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9015082585637114e-05,
      "loss": 0.8862,
      "step": 688
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9011651715286814e-05,
      "loss": 0.9306,
      "step": 689
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9008215190378014e-05,
      "loss": 0.9051,
      "step": 690
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9004773013067036e-05,
      "loss": 0.8414,
      "step": 691
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.900132518551375e-05,
      "loss": 0.8943,
      "step": 692
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.899787170988157e-05,
      "loss": 0.9065,
      "step": 693
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8994412588337456e-05,
      "loss": 0.8673,
      "step": 694
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.899094782305191e-05,
      "loss": 0.2901,
      "step": 695
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.898747741619897e-05,
      "loss": 0.882,
      "step": 696
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.898400136995622e-05,
      "loss": 0.8848,
      "step": 697
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8980519686504792e-05,
      "loss": 0.9103,
      "step": 698
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8977032368029333e-05,
      "loss": 0.9009,
      "step": 699
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8973539416718042e-05,
      "loss": 0.254,
      "step": 700
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8970040834762647e-05,
      "loss": 0.8666,
      "step": 701
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.896653662435841e-05,
      "loss": 0.932,
      "step": 702
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8963026787704123e-05,
      "loss": 0.9666,
      "step": 703
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8959511327002115e-05,
      "loss": 0.8978,
      "step": 704
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8955990244458234e-05,
      "loss": 0.8655,
      "step": 705
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8952463542281864e-05,
      "loss": 0.9544,
      "step": 706
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8948931222685906e-05,
      "loss": 0.2467,
      "step": 707
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.89453932878868e-05,
      "loss": 0.8773,
      "step": 708
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8941849740104493e-05,
      "loss": 0.9319,
      "step": 709
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8938300581562465e-05,
      "loss": 0.9029,
      "step": 710
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8934745814487715e-05,
      "loss": 0.2895,
      "step": 711
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8931185441110752e-05,
      "loss": 0.8903,
      "step": 712
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8927619463665622e-05,
      "loss": 0.9049,
      "step": 713
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8924047884389867e-05,
      "loss": 0.8764,
      "step": 714
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8920470705524558e-05,
      "loss": 0.8717,
      "step": 715
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8916887929314273e-05,
      "loss": 0.9173,
      "step": 716
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8913299558007096e-05,
      "loss": 0.8998,
      "step": 717
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.890970559385464e-05,
      "loss": 0.8794,
      "step": 718
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8906106039112012e-05,
      "loss": 0.8916,
      "step": 719
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8902500896037832e-05,
      "loss": 0.8283,
      "step": 720
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8898890166894224e-05,
      "loss": 0.9018,
      "step": 721
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.889527385394682e-05,
      "loss": 0.8723,
      "step": 722
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.889165195946476e-05,
      "loss": 0.8815,
      "step": 723
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8888024485720674e-05,
      "loss": 0.9303,
      "step": 724
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8884391434990704e-05,
      "loss": 0.2814,
      "step": 725
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.888075280955449e-05,
      "loss": 0.8495,
      "step": 726
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.887710861169516e-05,
      "loss": 0.8559,
      "step": 727
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.887345884369935e-05,
      "loss": 0.2363,
      "step": 728
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8869803507857186e-05,
      "loss": 0.8949,
      "step": 729
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.886614260646229e-05,
      "loss": 0.8975,
      "step": 730
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8862476141811773e-05,
      "loss": 0.8738,
      "step": 731
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8858804116206243e-05,
      "loss": 0.8726,
      "step": 732
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8855126531949786e-05,
      "loss": 0.8464,
      "step": 733
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.885144339134999e-05,
      "loss": 0.8162,
      "step": 734
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.884775469671791e-05,
      "loss": 0.8693,
      "step": 735
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8844060450368118e-05,
      "loss": 0.8628,
      "step": 736
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8840360654618636e-05,
      "loss": 0.8786,
      "step": 737
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8836655311790984e-05,
      "loss": 0.8512,
      "step": 738
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.883294442421016e-05,
      "loss": 0.8247,
      "step": 739
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8829227994204647e-05,
      "loss": 0.2583,
      "step": 740
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.88255060241064e-05,
      "loss": 0.8807,
      "step": 741
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.882177851625084e-05,
      "loss": 0.8555,
      "step": 742
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8818045472976887e-05,
      "loss": 0.9164,
      "step": 743
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8814306896626916e-05,
      "loss": 0.8625,
      "step": 744
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8810562789546775e-05,
      "loss": 0.8766,
      "step": 745
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.880681315408579e-05,
      "loss": 0.8046,
      "step": 746
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.880305799259675e-05,
      "loss": 0.9222,
      "step": 747
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8799297307435914e-05,
      "loss": 0.9415,
      "step": 748
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8795531100963008e-05,
      "loss": 0.799,
      "step": 749
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8791759375541222e-05,
      "loss": 0.8686,
      "step": 750
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.87879821335372e-05,
      "loss": 0.9644,
      "step": 751
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8784199377321065e-05,
      "loss": 0.8979,
      "step": 752
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8780411109266387e-05,
      "loss": 0.9172,
      "step": 753
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8776617331750197e-05,
      "loss": 0.8439,
      "step": 754
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.877281804715298e-05,
      "loss": 0.8839,
      "step": 755
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8769013257858688e-05,
      "loss": 0.958,
      "step": 756
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8765202966254715e-05,
      "loss": 0.8921,
      "step": 757
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8761387174731913e-05,
      "loss": 0.8843,
      "step": 758
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8757565885684584e-05,
      "loss": 0.9072,
      "step": 759
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.875373910151048e-05,
      "loss": 0.2972,
      "step": 760
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.87499068246108e-05,
      "loss": 0.8919,
      "step": 761
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8746069057390194e-05,
      "loss": 0.8725,
      "step": 762
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8742225802256746e-05,
      "loss": 0.8839,
      "step": 763
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8738377061621996e-05,
      "loss": 0.2519,
      "step": 764
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8734522837900917e-05,
      "loss": 0.8899,
      "step": 765
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.873066313351193e-05,
      "loss": 0.9264,
      "step": 766
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.872679795087689e-05,
      "loss": 0.8725,
      "step": 767
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.872292729242109e-05,
      "loss": 0.8942,
      "step": 768
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8719051160573257e-05,
      "loss": 0.8892,
      "step": 769
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.871516955776556e-05,
      "loss": 0.9245,
      "step": 770
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8711282486433595e-05,
      "loss": 0.8945,
      "step": 771
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8707389949016385e-05,
      "loss": 0.2478,
      "step": 772
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8703491947956394e-05,
      "loss": 0.853,
      "step": 773
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8699588485699504e-05,
      "loss": 0.8498,
      "step": 774
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8695679564695033e-05,
      "loss": 0.844,
      "step": 775
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8691765187395713e-05,
      "loss": 0.8812,
      "step": 776
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8687845356257708e-05,
      "loss": 0.8938,
      "step": 777
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8683920073740605e-05,
      "loss": 0.8769,
      "step": 778
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8679989342307407e-05,
      "loss": 0.8484,
      "step": 779
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8676053164424535e-05,
      "loss": 0.9239,
      "step": 780
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8672111542561835e-05,
      "loss": 0.8414,
      "step": 781
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.866816447919256e-05,
      "loss": 0.2571,
      "step": 782
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8664211976793384e-05,
      "loss": 0.9066,
      "step": 783
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.866025403784439e-05,
      "loss": 0.8234,
      "step": 784
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.865629066482907e-05,
      "loss": 0.8528,
      "step": 785
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8652321860234337e-05,
      "loss": 0.8847,
      "step": 786
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.86483476265505e-05,
      "loss": 0.8807,
      "step": 787
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8644367966271276e-05,
      "loss": 0.8981,
      "step": 788
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.86403828818938e-05,
      "loss": 0.8912,
      "step": 789
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8636392375918587e-05,
      "loss": 0.8712,
      "step": 790
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.863239645084958e-05,
      "loss": 0.8777,
      "step": 791
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8628395109194106e-05,
      "loss": 0.9119,
      "step": 792
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.862438835346289e-05,
      "loss": 0.8893,
      "step": 793
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8620376186170062e-05,
      "loss": 0.8668,
      "step": 794
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8616358609833146e-05,
      "loss": 0.9254,
      "step": 795
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8612335626973056e-05,
      "loss": 0.8718,
      "step": 796
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.86083072401141e-05,
      "loss": 0.8302,
      "step": 797
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.860427345178398e-05,
      "loss": 0.9067,
      "step": 798
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8600234264513785e-05,
      "loss": 0.8796,
      "step": 799
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.859618968083799e-05,
      "loss": 0.8695,
      "step": 800
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.859213970329446e-05,
      "loss": 0.8784,
      "step": 801
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8588084334424442e-05,
      "loss": 0.8832,
      "step": 802
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.858402357677256e-05,
      "loss": 0.8693,
      "step": 803
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8579957432886834e-05,
      "loss": 0.9052,
      "step": 804
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.857588590531865e-05,
      "loss": 0.8717,
      "step": 805
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8571808996622777e-05,
      "loss": 0.8983,
      "step": 806
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8567726709357367e-05,
      "loss": 0.2762,
      "step": 807
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.856363904608393e-05,
      "loss": 0.8721,
      "step": 808
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8559546009367375e-05,
      "loss": 0.9098,
      "step": 809
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8555447601775954e-05,
      "loss": 0.8722,
      "step": 810
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.855134382588131e-05,
      "loss": 0.8781,
      "step": 811
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.854723468425844e-05,
      "loss": 0.8885,
      "step": 812
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.854312017948572e-05,
      "loss": 0.8346,
      "step": 813
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.853900031414489e-05,
      "loss": 0.8231,
      "step": 814
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.853487509082104e-05,
      "loss": 0.8602,
      "step": 815
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8530744512102644e-05,
      "loss": 0.8672,
      "step": 816
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8526608580581517e-05,
      "loss": 0.8728,
      "step": 817
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.852246729885284e-05,
      "loss": 0.9033,
      "step": 818
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.851832066951515e-05,
      "loss": 0.907,
      "step": 819
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8514168695170344e-05,
      "loss": 0.8613,
      "step": 820
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8510011378423666e-05,
      "loss": 0.836,
      "step": 821
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.850584872188372e-05,
      "loss": 0.9819,
      "step": 822
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8501680728162452e-05,
      "loss": 0.8339,
      "step": 823
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.849750739987516e-05,
      "loss": 0.229,
      "step": 824
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8493328739640497e-05,
      "loss": 0.899,
      "step": 825
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.848914475008045e-05,
      "loss": 0.8757,
      "step": 826
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8484955433820352e-05,
      "loss": 0.8497,
      "step": 827
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8480760793488893e-05,
      "loss": 0.8824,
      "step": 828
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.847656083171808e-05,
      "loss": 0.8637,
      "step": 829
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8472355551143275e-05,
      "loss": 0.87,
      "step": 830
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8468144954403176e-05,
      "loss": 0.9288,
      "step": 831
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.846392904413981e-05,
      "loss": 0.9178,
      "step": 832
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.845970782299854e-05,
      "loss": 0.8955,
      "step": 833
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8455481293628075e-05,
      "loss": 0.8848,
      "step": 834
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8451249458680438e-05,
      "loss": 0.867,
      "step": 835
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8447012320810985e-05,
      "loss": 0.8265,
      "step": 836
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.84427698826784e-05,
      "loss": 0.9357,
      "step": 837
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8438522146944695e-05,
      "loss": 0.9051,
      "step": 838
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8434269116275204e-05,
      "loss": 0.9045,
      "step": 839
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.843001079333859e-05,
      "loss": 0.2491,
      "step": 840
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8425747180806824e-05,
      "loss": 0.9041,
      "step": 841
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8421478281355207e-05,
      "loss": 0.8556,
      "step": 842
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.841720409766235e-05,
      "loss": 0.8772,
      "step": 843
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.841292463241019e-05,
      "loss": 0.876,
      "step": 844
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8408639888283965e-05,
      "loss": 0.8892,
      "step": 845
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8404349867972235e-05,
      "loss": 0.9049,
      "step": 846
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8400054574166865e-05,
      "loss": 0.8776,
      "step": 847
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8395754009563034e-05,
      "loss": 0.8817,
      "step": 848
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8391448176859222e-05,
      "loss": 0.8834,
      "step": 849
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8387137078757225e-05,
      "loss": 0.8784,
      "step": 850
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.838282071796213e-05,
      "loss": 0.2903,
      "step": 851
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.837849909718233e-05,
      "loss": 0.8541,
      "step": 852
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8374172219129525e-05,
      "loss": 0.902,
      "step": 853
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.836984008651871e-05,
      "loss": 0.8666,
      "step": 854
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.836550270206818e-05,
      "loss": 0.8679,
      "step": 855
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.836116006849951e-05,
      "loss": 0.9037,
      "step": 856
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8356812188537594e-05,
      "loss": 0.8256,
      "step": 857
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8352459064910597e-05,
      "loss": 0.858,
      "step": 858
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8348100700349983e-05,
      "loss": 0.9334,
      "step": 859
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.83437370975905e-05,
      "loss": 0.2939,
      "step": 860
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8339368259370198e-05,
      "loss": 0.8656,
      "step": 861
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8334994188430387e-05,
      "loss": 0.8651,
      "step": 862
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8330614887515686e-05,
      "loss": 0.8556,
      "step": 863
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.832623035937397e-05,
      "loss": 0.8771,
      "step": 864
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.832184060675642e-05,
      "loss": 0.8565,
      "step": 865
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8317445632417473e-05,
      "loss": 0.8608,
      "step": 866
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8313045439114855e-05,
      "loss": 0.9102,
      "step": 867
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8308640029609567e-05,
      "loss": 0.8544,
      "step": 868
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8304229406665874e-05,
      "loss": 0.8908,
      "step": 869
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8299813573051324e-05,
      "loss": 0.969,
      "step": 870
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8295392531536728e-05,
      "loss": 0.8827,
      "step": 871
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.829096628489616e-05,
      "loss": 0.897,
      "step": 872
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.828653483590697e-05,
      "loss": 0.9119,
      "step": 873
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.828209818734977e-05,
      "loss": 0.2494,
      "step": 874
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8277656342008427e-05,
      "loss": 0.9156,
      "step": 875
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8273209302670077e-05,
      "loss": 0.9463,
      "step": 876
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8268757072125112e-05,
      "loss": 0.8925,
      "step": 877
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8264299653167178e-05,
      "loss": 0.8788,
      "step": 878
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.825983704859319e-05,
      "loss": 0.8838,
      "step": 879
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8255369261203297e-05,
      "loss": 0.8592,
      "step": 880
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8250896293800917e-05,
      "loss": 0.9022,
      "step": 881
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8246418149192705e-05,
      "loss": 0.9163,
      "step": 882
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.824193483018858e-05,
      "loss": 0.826,
      "step": 883
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.823744633960169e-05,
      "loss": 0.8741,
      "step": 884
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.823295268024844e-05,
      "loss": 0.9002,
      "step": 885
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8228453854948482e-05,
      "loss": 0.908,
      "step": 886
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8223949866524687e-05,
      "loss": 0.8698,
      "step": 887
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8219440717803197e-05,
      "loss": 0.9253,
      "step": 888
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.821492641161337e-05,
      "loss": 0.8993,
      "step": 889
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.82104069507878e-05,
      "loss": 0.8566,
      "step": 890
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8205882338162334e-05,
      "loss": 0.8407,
      "step": 891
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.820135257657603e-05,
      "loss": 0.8538,
      "step": 892
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8196817668871193e-05,
      "loss": 0.8434,
      "step": 893
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8192277617893346e-05,
      "loss": 0.9028,
      "step": 894
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8187732426491243e-05,
      "loss": 0.8969,
      "step": 895
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8183182097516863e-05,
      "loss": 0.8922,
      "step": 896
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8178626633825417e-05,
      "loss": 0.9361,
      "step": 897
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8174066038275328e-05,
      "loss": 0.8596,
      "step": 898
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8169500313728233e-05,
      "loss": 0.2773,
      "step": 899
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8164929463049004e-05,
      "loss": 0.9159,
      "step": 900
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8160353489105726e-05,
      "loss": 0.8908,
      "step": 901
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.815577239476968e-05,
      "loss": 0.8948,
      "step": 902
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8151186182915384e-05,
      "loss": 0.9206,
      "step": 903
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.814659485642056e-05,
      "loss": 0.9309,
      "step": 904
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.814199841816613e-05,
      "loss": 0.8804,
      "step": 905
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8137396871036234e-05,
      "loss": 0.266,
      "step": 906
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.813279021791821e-05,
      "loss": 0.8969,
      "step": 907
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8128178461702603e-05,
      "loss": 0.9271,
      "step": 908
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8123561605283165e-05,
      "loss": 0.9137,
      "step": 909
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.811893965155684e-05,
      "loss": 0.8725,
      "step": 910
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8114312603423778e-05,
      "loss": 0.8552,
      "step": 911
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.810968046378732e-05,
      "loss": 0.8258,
      "step": 912
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8105043235554e-05,
      "loss": 0.8445,
      "step": 913
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8100400921633555e-05,
      "loss": 0.8529,
      "step": 914
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8095753524938903e-05,
      "loss": 0.8817,
      "step": 915
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8091101048386154e-05,
      "loss": 0.868,
      "step": 916
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8086443494894606e-05,
      "loss": 0.8918,
      "step": 917
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8081780867386755e-05,
      "loss": 0.8333,
      "step": 918
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.807711316878825e-05,
      "loss": 0.2513,
      "step": 919
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8072440402027952e-05,
      "loss": 0.8449,
      "step": 920
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.806776257003789e-05,
      "loss": 0.8738,
      "step": 921
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8063079675753266e-05,
      "loss": 0.8891,
      "step": 922
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8058391722112472e-05,
      "loss": 0.8824,
      "step": 923
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.805369871205707e-05,
      "loss": 0.8536,
      "step": 924
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8049000648531782e-05,
      "loss": 0.9008,
      "step": 925
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8044297534484516e-05,
      "loss": 0.8223,
      "step": 926
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.803958937286635e-05,
      "loss": 0.8896,
      "step": 927
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.803487616663151e-05,
      "loss": 0.8675,
      "step": 928
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.803015791873741e-05,
      "loss": 0.8704,
      "step": 929
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.802543463214462e-05,
      "loss": 0.8748,
      "step": 930
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.802070630981686e-05,
      "loss": 0.8428,
      "step": 931
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.801597295472103e-05,
      "loss": 0.8427,
      "step": 932
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8011234569827173e-05,
      "loss": 0.8688,
      "step": 933
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.800649115810849e-05,
      "loss": 0.8944,
      "step": 934
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8001742722541343e-05,
      "loss": 0.8806,
      "step": 935
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7996989266105243e-05,
      "loss": 0.8395,
      "step": 936
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.799223079178284e-05,
      "loss": 0.8146,
      "step": 937
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.798746730255996e-05,
      "loss": 0.2571,
      "step": 938
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7982698801425543e-05,
      "loss": 0.9127,
      "step": 939
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7977925291371698e-05,
      "loss": 0.8909,
      "step": 940
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7973146775393667e-05,
      "loss": 0.8668,
      "step": 941
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7968363256489834e-05,
      "loss": 0.9186,
      "step": 942
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7963574737661723e-05,
      "loss": 0.9299,
      "step": 943
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7958781221913995e-05,
      "loss": 0.23,
      "step": 944
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.795398271225445e-05,
      "loss": 0.2714,
      "step": 945
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.794917921169401e-05,
      "loss": 0.8121,
      "step": 946
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7944370723246744e-05,
      "loss": 0.8323,
      "step": 947
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7939557249929844e-05,
      "loss": 0.8713,
      "step": 948
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7934738794763622e-05,
      "loss": 0.8963,
      "step": 949
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7929915360771533e-05,
      "loss": 0.8688,
      "step": 950
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7925086950980137e-05,
      "loss": 0.8685,
      "step": 951
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7920253568419132e-05,
      "loss": 0.8799,
      "step": 952
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.791541521612133e-05,
      "loss": 0.8397,
      "step": 953
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7910571897122656e-05,
      "loss": 0.9235,
      "step": 954
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7905723614462162e-05,
      "loss": 0.8357,
      "step": 955
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.790087037118201e-05,
      "loss": 0.8941,
      "step": 956
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.789601217032747e-05,
      "loss": 0.8774,
      "step": 957
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7891149014946926e-05,
      "loss": 0.8949,
      "step": 958
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.788628090809187e-05,
      "loss": 0.894,
      "step": 959
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7881407852816912e-05,
      "loss": 0.8935,
      "step": 960
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.787652985217975e-05,
      "loss": 0.8991,
      "step": 961
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.787164690924119e-05,
      "loss": 0.8746,
      "step": 962
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.786675902706515e-05,
      "loss": 0.2863,
      "step": 963
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.786186620871863e-05,
      "loss": 0.8811,
      "step": 964
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7856968457271744e-05,
      "loss": 0.2571,
      "step": 965
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7852065775797692e-05,
      "loss": 0.8426,
      "step": 966
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.784715816737276e-05,
      "loss": 0.9078,
      "step": 967
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7842245635076345e-05,
      "loss": 0.8462,
      "step": 968
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.783732818199092e-05,
      "loss": 0.8811,
      "step": 969
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.783240581120205e-05,
      "loss": 0.2649,
      "step": 970
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.782747852579838e-05,
      "loss": 0.8672,
      "step": 971
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7822546328871648e-05,
      "loss": 0.9204,
      "step": 972
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7817609223516665e-05,
      "loss": 0.8751,
      "step": 973
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7812667212831332e-05,
      "loss": 0.8875,
      "step": 974
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7807720299916617e-05,
      "loss": 0.8643,
      "step": 975
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.780276848787657e-05,
      "loss": 0.8366,
      "step": 976
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7797811779818316e-05,
      "loss": 0.9066,
      "step": 977
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7792850178852044e-05,
      "loss": 0.872,
      "step": 978
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7787883688091026e-05,
      "loss": 0.8844,
      "step": 979
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7782912310651595e-05,
      "loss": 0.9211,
      "step": 980
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.777793604965315e-05,
      "loss": 0.8442,
      "step": 981
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.777295490821815e-05,
      "loss": 0.8533,
      "step": 982
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.776796888947213e-05,
      "loss": 0.8995,
      "step": 983
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7762977996543667e-05,
      "loss": 0.852,
      "step": 984
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7757982232564418e-05,
      "loss": 0.864,
      "step": 985
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7752981600669077e-05,
      "loss": 0.9461,
      "step": 986
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7747976103995403e-05,
      "loss": 0.8176,
      "step": 987
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7742965745684202e-05,
      "loss": 0.8488,
      "step": 988
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7737950528879334e-05,
      "loss": 0.8185,
      "step": 989
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7732930456727707e-05,
      "loss": 0.9108,
      "step": 990
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7727905532379276e-05,
      "loss": 0.8758,
      "step": 991
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7722875758987038e-05,
      "loss": 0.8753,
      "step": 992
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.771784113970704e-05,
      "loss": 0.879,
      "step": 993
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7712801677698356e-05,
      "loss": 0.2702,
      "step": 994
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7707757376123112e-05,
      "loss": 0.83,
      "step": 995
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.770270823814647e-05,
      "loss": 0.8406,
      "step": 996
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7697654266936615e-05,
      "loss": 0.8443,
      "step": 997
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7692595465664776e-05,
      "loss": 0.2489,
      "step": 998
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.76875318375052e-05,
      "loss": 0.8647,
      "step": 999
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7682463385635186e-05,
      "loss": 0.8861,
      "step": 1000
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7677390113235038e-05,
      "loss": 0.9298,
      "step": 1001
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.767231202348809e-05,
      "loss": 0.9077,
      "step": 1002
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.76672291195807e-05,
      "loss": 0.8149,
      "step": 1003
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7662141404702256e-05,
      "loss": 0.8343,
      "step": 1004
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.765704888204515e-05,
      "loss": 0.8669,
      "step": 1005
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7651951554804797e-05,
      "loss": 0.8411,
      "step": 1006
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.764684942617963e-05,
      "loss": 0.8526,
      "step": 1007
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7641742499371085e-05,
      "loss": 0.8282,
      "step": 1008
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7636630777583623e-05,
      "loss": 0.8427,
      "step": 1009
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7631514264024697e-05,
      "loss": 0.8694,
      "step": 1010
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7626392961904787e-05,
      "loss": 0.8756,
      "step": 1011
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7621266874437357e-05,
      "loss": 0.2586,
      "step": 1012
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7616136004838887e-05,
      "loss": 0.869,
      "step": 1013
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7611000356328853e-05,
      "loss": 0.8591,
      "step": 1014
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7605859932129728e-05,
      "loss": 0.8911,
      "step": 1015
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.760071473546699e-05,
      "loss": 0.8554,
      "step": 1016
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7595564769569097e-05,
      "loss": 0.8444,
      "step": 1017
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7590410037667512e-05,
      "loss": 0.8708,
      "step": 1018
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.758525054299669e-05,
      "loss": 0.8724,
      "step": 1019
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7580086288794064e-05,
      "loss": 0.8177,
      "step": 1020
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7574917278300057e-05,
      "loss": 0.8757,
      "step": 1021
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7569743514758076e-05,
      "loss": 0.264,
      "step": 1022
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7564565001414523e-05,
      "loss": 0.2743,
      "step": 1023
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7559381741518762e-05,
      "loss": 0.9042,
      "step": 1024
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.755419373832314e-05,
      "loss": 0.8647,
      "step": 1025
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.754900099508299e-05,
      "loss": 0.8856,
      "step": 1026
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.754380351505661e-05,
      "loss": 0.8208,
      "step": 1027
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7538601301505272e-05,
      "loss": 0.8873,
      "step": 1028
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7533394357693222e-05,
      "loss": 0.8826,
      "step": 1029
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7528182686887667e-05,
      "loss": 0.8939,
      "step": 1030
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7522966292358784e-05,
      "loss": 0.8363,
      "step": 1031
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.751774517737972e-05,
      "loss": 0.8856,
      "step": 1032
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7512519345226574e-05,
      "loss": 0.8747,
      "step": 1033
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.750728879917841e-05,
      "loss": 0.8739,
      "step": 1034
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7502053542517245e-05,
      "loss": 0.8145,
      "step": 1035
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.749681357852806e-05,
      "loss": 0.8642,
      "step": 1036
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7491568910498788e-05,
      "loss": 0.8565,
      "step": 1037
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7486319541720302e-05,
      "loss": 0.8519,
      "step": 1038
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.748106547548644e-05,
      "loss": 0.8791,
      "step": 1039
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.747580671509398e-05,
      "loss": 0.9018,
      "step": 1040
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7470543263842644e-05,
      "loss": 0.8478,
      "step": 1041
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.746527512503511e-05,
      "loss": 0.9293,
      "step": 1042
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.746000230197697e-05,
      "loss": 0.8729,
      "step": 1043
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7454724797976787e-05,
      "loss": 0.8706,
      "step": 1044
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.744944261634604e-05,
      "loss": 0.8616,
      "step": 1045
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.744415576039915e-05,
      "loss": 0.855,
      "step": 1046
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7438864233453476e-05,
      "loss": 0.8803,
      "step": 1047
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7433568038829296e-05,
      "loss": 0.2669,
      "step": 1048
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7428267179849828e-05,
      "loss": 0.8625,
      "step": 1049
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7422961659841206e-05,
      "loss": 0.8663,
      "step": 1050
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.74176514821325e-05,
      "loss": 0.8527,
      "step": 1051
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7412336650055697e-05,
      "loss": 0.8781,
      "step": 1052
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.740701716694571e-05,
      "loss": 0.2614,
      "step": 1053
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7401693036140352e-05,
      "loss": 0.916,
      "step": 1054
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.739636426098038e-05,
      "loss": 0.2716,
      "step": 1055
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.739103084480944e-05,
      "loss": 0.8589,
      "step": 1056
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7385692790974108e-05,
      "loss": 0.8533,
      "step": 1057
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7380350102823857e-05,
      "loss": 0.8699,
      "step": 1058
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.737500278371108e-05,
      "loss": 0.2617,
      "step": 1059
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7369650836991064e-05,
      "loss": 0.8959,
      "step": 1060
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.736429426602201e-05,
      "loss": 0.8399,
      "step": 1061
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7358933074165013e-05,
      "loss": 0.8218,
      "step": 1062
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7353567264784074e-05,
      "loss": 0.8598,
      "step": 1063
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7348196841246083e-05,
      "loss": 0.8737,
      "step": 1064
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.734282180692083e-05,
      "loss": 0.906,
      "step": 1065
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7337442165181006e-05,
      "loss": 0.8282,
      "step": 1066
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7332057919402174e-05,
      "loss": 0.8837,
      "step": 1067
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7326669072962805e-05,
      "loss": 0.8689,
      "step": 1068
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7321275629244247e-05,
      "loss": 0.8647,
      "step": 1069
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7315877591630733e-05,
      "loss": 0.8718,
      "step": 1070
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.731047496350938e-05,
      "loss": 0.2661,
      "step": 1071
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.730506774827019e-05,
      "loss": 0.8519,
      "step": 1072
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7299655949306036e-05,
      "loss": 0.8829,
      "step": 1073
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7294239570012666e-05,
      "loss": 0.815,
      "step": 1074
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.728881861378871e-05,
      "loss": 0.8752,
      "step": 1075
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7283393084035665e-05,
      "loss": 0.8466,
      "step": 1076
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.72779629841579e-05,
      "loss": 0.9032,
      "step": 1077
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7272528317562648e-05,
      "loss": 0.8798,
      "step": 1078
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7267089087660015e-05,
      "loss": 0.8681,
      "step": 1079
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.726164529786296e-05,
      "loss": 0.8554,
      "step": 1080
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7256196951587304e-05,
      "loss": 0.8643,
      "step": 1081
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.725074405225174e-05,
      "loss": 0.8698,
      "step": 1082
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7245286603277804e-05,
      "loss": 0.8829,
      "step": 1083
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7239824608089896e-05,
      "loss": 0.8688,
      "step": 1084
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.723435807011526e-05,
      "loss": 0.8146,
      "step": 1085
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7228886992784e-05,
      "loss": 0.8329,
      "step": 1086
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7223411379529062e-05,
      "loss": 0.8169,
      "step": 1087
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7217931233786238e-05,
      "loss": 0.8452,
      "step": 1088
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7212446558994163e-05,
      "loss": 0.899,
      "step": 1089
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7206957358594324e-05,
      "loss": 0.7827,
      "step": 1090
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7201463636031036e-05,
      "loss": 0.8821,
      "step": 1091
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7195965394751452e-05,
      "loss": 0.881,
      "step": 1092
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7190462638205573e-05,
      "loss": 0.2658,
      "step": 1093
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7184955369846216e-05,
      "loss": 0.8385,
      "step": 1094
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7179443593129042e-05,
      "loss": 0.8821,
      "step": 1095
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7173927311512533e-05,
      "loss": 0.8032,
      "step": 1096
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7168406528458e-05,
      "loss": 0.8882,
      "step": 1097
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.716288124742959e-05,
      "loss": 0.8906,
      "step": 1098
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7157351471894247e-05,
      "loss": 0.275,
      "step": 1099
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.715181720532176e-05,
      "loss": 0.825,
      "step": 1100
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.714627845118472e-05,
      "loss": 0.8789,
      "step": 1101
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7140735212958543e-05,
      "loss": 0.8677,
      "step": 1102
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7135187494121453e-05,
      "loss": 0.2789,
      "step": 1103
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.712963529815449e-05,
      "loss": 0.8884,
      "step": 1104
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7124078628541504e-05,
      "loss": 0.8522,
      "step": 1105
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7118517488769143e-05,
      "loss": 0.8289,
      "step": 1106
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7112951882326872e-05,
      "loss": 0.881,
      "step": 1107
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7107381812706947e-05,
      "loss": 0.8834,
      "step": 1108
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7101807283404436e-05,
      "loss": 0.8957,
      "step": 1109
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7096228297917192e-05,
      "loss": 0.8644,
      "step": 1110
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.709064485974588e-05,
      "loss": 0.8642,
      "step": 1111
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7085056972393945e-05,
      "loss": 0.8411,
      "step": 1112
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7079464639367634e-05,
      "loss": 0.8534,
      "step": 1113
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7073867864175975e-05,
      "loss": 0.2708,
      "step": 1114
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.706826665033079e-05,
      "loss": 0.8447,
      "step": 1115
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7062661001346683e-05,
      "loss": 0.956,
      "step": 1116
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.705705092074104e-05,
      "loss": 0.8538,
      "step": 1117
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.705143641203403e-05,
      "loss": 0.8634,
      "step": 1118
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.70458174787486e-05,
      "loss": 0.8474,
      "step": 1119
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7040194124410475e-05,
      "loss": 0.9246,
      "step": 1120
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.703456635254815e-05,
      "loss": 0.8909,
      "step": 1121
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.702893416669289e-05,
      "loss": 0.8381,
      "step": 1122
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7023297570378738e-05,
      "loss": 0.8165,
      "step": 1123
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7017656567142498e-05,
      "loss": 0.8545,
      "step": 1124
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.7012011160523744e-05,
      "loss": 0.2394,
      "step": 1125
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.7006361354064807e-05,
      "loss": 0.85,
      "step": 1126
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.7000707151310785e-05,
      "loss": 0.8372,
      "step": 1127
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.699504855580953e-05,
      "loss": 0.9353,
      "step": 1128
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.698938557111165e-05,
      "loss": 0.8155,
      "step": 1129
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6983718200770512e-05,
      "loss": 0.9229,
      "step": 1130
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6978046448342228e-05,
      "loss": 0.8969,
      "step": 1131
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.697237031738567e-05,
      "loss": 0.88,
      "step": 1132
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6966689811462444e-05,
      "loss": 0.8842,
      "step": 1133
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.696100493413691e-05,
      "loss": 0.8069,
      "step": 1134
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.695531568897617e-05,
      "loss": 0.8457,
      "step": 1135
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6949622079550066e-05,
      "loss": 0.8602,
      "step": 1136
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.694392410943118e-05,
      "loss": 0.8309,
      "step": 1137
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6938221782194822e-05,
      "loss": 0.864,
      "step": 1138
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6932515101419048e-05,
      "loss": 0.897,
      "step": 1139
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6926804070684638e-05,
      "loss": 0.248,
      "step": 1140
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6921088693575104e-05,
      "loss": 0.8988,
      "step": 1141
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6915368973676685e-05,
      "loss": 0.819,
      "step": 1142
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.690964491457834e-05,
      "loss": 0.8831,
      "step": 1143
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6903916519871763e-05,
      "loss": 0.8416,
      "step": 1144
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6898183793151356e-05,
      "loss": 0.8969,
      "step": 1145
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6892446738014244e-05,
      "loss": 0.7942,
      "step": 1146
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6886705358060274e-05,
      "loss": 0.8627,
      "step": 1147
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6880959656891992e-05,
      "loss": 0.871,
      "step": 1148
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.687520963811467e-05,
      "loss": 0.8722,
      "step": 1149
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6869455305336284e-05,
      "loss": 0.8616,
      "step": 1150
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6863696662167513e-05,
      "loss": 0.8529,
      "step": 1151
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.685793371222174e-05,
      "loss": 0.8493,
      "step": 1152
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6852166459115064e-05,
      "loss": 0.8288,
      "step": 1153
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6846394906466266e-05,
      "loss": 0.8861,
      "step": 1154
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6840619057896845e-05,
      "loss": 0.8984,
      "step": 1155
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6834838917030968e-05,
      "loss": 0.8359,
      "step": 1156
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.682905448749552e-05,
      "loss": 0.8812,
      "step": 1157
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.682326577292007e-05,
      "loss": 0.913,
      "step": 1158
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.681747277693687e-05,
      "loss": 0.8644,
      "step": 1159
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6811675503180866e-05,
      "loss": 0.8663,
      "step": 1160
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.680587395528968e-05,
      "loss": 0.8549,
      "step": 1161
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6800068136903622e-05,
      "loss": 0.8454,
      "step": 1162
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.679425805166568e-05,
      "loss": 0.8766,
      "step": 1163
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6788443703221523e-05,
      "loss": 0.2876,
      "step": 1164
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6782625095219486e-05,
      "loss": 0.8928,
      "step": 1165
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6776802231310587e-05,
      "loss": 0.7978,
      "step": 1166
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6770975115148506e-05,
      "loss": 0.8999,
      "step": 1167
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6765143750389596e-05,
      "loss": 0.2645,
      "step": 1168
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6759308140692873e-05,
      "loss": 0.8855,
      "step": 1169
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.675346828972002e-05,
      "loss": 0.8896,
      "step": 1170
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6747624201135384e-05,
      "loss": 0.8035,
      "step": 1171
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6741775878605963e-05,
      "loss": 0.2514,
      "step": 1172
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6735923325801407e-05,
      "loss": 0.8011,
      "step": 1173
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.673006654639404e-05,
      "loss": 0.8039,
      "step": 1174
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6724205544058825e-05,
      "loss": 0.9175,
      "step": 1175
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6718340322473373e-05,
      "loss": 0.8046,
      "step": 1176
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6712470885317946e-05,
      "loss": 0.8342,
      "step": 1177
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6706597236275452e-05,
      "loss": 0.8285,
      "step": 1178
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.670071937903144e-05,
      "loss": 0.7959,
      "step": 1179
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.669483731727411e-05,
      "loss": 0.842,
      "step": 1180
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6688951054694274e-05,
      "loss": 0.807,
      "step": 1181
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6683060594985405e-05,
      "loss": 0.8932,
      "step": 1182
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6677165941843603e-05,
      "loss": 0.8966,
      "step": 1183
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6671267098967594e-05,
      "loss": 0.8924,
      "step": 1184
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.666536407005874e-05,
      "loss": 0.8422,
      "step": 1185
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6659456858821022e-05,
      "loss": 0.8383,
      "step": 1186
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6653545468961047e-05,
      "loss": 0.8774,
      "step": 1187
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6647629904188056e-05,
      "loss": 0.8946,
      "step": 1188
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.664171016821389e-05,
      "loss": 0.8268,
      "step": 1189
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6635786264753025e-05,
      "loss": 0.7925,
      "step": 1190
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6629858197522537e-05,
      "loss": 0.8657,
      "step": 1191
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6623925970242126e-05,
      "loss": 0.9027,
      "step": 1192
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6617989586634097e-05,
      "loss": 0.8352,
      "step": 1193
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6612049050423363e-05,
      "loss": 0.9089,
      "step": 1194
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6606104365337444e-05,
      "loss": 0.8665,
      "step": 1195
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.660015553510647e-05,
      "loss": 0.871,
      "step": 1196
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6594202563463152e-05,
      "loss": 0.8348,
      "step": 1197
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.658824545414282e-05,
      "loss": 0.8024,
      "step": 1198
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6582284210883396e-05,
      "loss": 0.8789,
      "step": 1199
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6576318837425386e-05,
      "loss": 0.9047,
      "step": 1200
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6570349337511903e-05,
      "loss": 0.8809,
      "step": 1201
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.656437571488863e-05,
      "loss": 0.2755,
      "step": 1202
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.655839797330385e-05,
      "loss": 0.8255,
      "step": 1203
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6552416116508435e-05,
      "loss": 0.8244,
      "step": 1204
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.654643014825583e-05,
      "loss": 0.8658,
      "step": 1205
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6540440072302056e-05,
      "loss": 0.8355,
      "step": 1206
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6534445892405722e-05,
      "loss": 0.9336,
      "step": 1207
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.652844761232801e-05,
      "loss": 0.8513,
      "step": 1208
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.652244523583267e-05,
      "loss": 0.8844,
      "step": 1209
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.651643876668602e-05,
      "loss": 0.906,
      "step": 1210
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.651042820865696e-05,
      "loss": 0.8288,
      "step": 1211
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.650441356551694e-05,
      "loss": 0.8485,
      "step": 1212
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.649839484103998e-05,
      "loss": 0.8216,
      "step": 1213
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6492372039002664e-05,
      "loss": 0.826,
      "step": 1214
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.648634516318413e-05,
      "loss": 0.9051,
      "step": 1215
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.648031421736607e-05,
      "loss": 0.8731,
      "step": 1216
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6474279205332738e-05,
      "loss": 0.8695,
      "step": 1217
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6468240130870933e-05,
      "loss": 0.861,
      "step": 1218
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.646219699777e-05,
      "loss": 0.9156,
      "step": 1219
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6456149809821845e-05,
      "loss": 0.8879,
      "step": 1220
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.64500985708209e-05,
      "loss": 0.8235,
      "step": 1221
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6444043284564152e-05,
      "loss": 0.8618,
      "step": 1222
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.643798395485112e-05,
      "loss": 0.8639,
      "step": 1223
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6431920585483866e-05,
      "loss": 0.9006,
      "step": 1224
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6425853180266988e-05,
      "loss": 0.9211,
      "step": 1225
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6419781743007604e-05,
      "loss": 0.8659,
      "step": 1226
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6413706277515375e-05,
      "loss": 0.8833,
      "step": 1227
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6407626787602486e-05,
      "loss": 0.8861,
      "step": 1228
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6401543277083643e-05,
      "loss": 0.8961,
      "step": 1229
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.639545574977608e-05,
      "loss": 0.8357,
      "step": 1230
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.638936420949955e-05,
      "loss": 0.8693,
      "step": 1231
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6383268660076325e-05,
      "loss": 0.8598,
      "step": 1232
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6377169105331185e-05,
      "loss": 0.8304,
      "step": 1233
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6371065549091436e-05,
      "loss": 0.8669,
      "step": 1234
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6364957995186886e-05,
      "loss": 0.7638,
      "step": 1235
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6358846447449853e-05,
      "loss": 0.8305,
      "step": 1236
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.635273090971516e-05,
      "loss": 0.8746,
      "step": 1237
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.634661138582014e-05,
      "loss": 0.2772,
      "step": 1238
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.634048787960462e-05,
      "loss": 0.7673,
      "step": 1239
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6334360394910925e-05,
      "loss": 0.8797,
      "step": 1240
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6328228935583886e-05,
      "loss": 0.2906,
      "step": 1241
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6322093505470816e-05,
      "loss": 0.8461,
      "step": 1242
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6315954108421527e-05,
      "loss": 0.845,
      "step": 1243
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.630981074828832e-05,
      "loss": 0.8924,
      "step": 1244
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.630366342892598e-05,
      "loss": 0.8285,
      "step": 1245
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6297512154191778e-05,
      "loss": 0.7861,
      "step": 1246
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6291356927945466e-05,
      "loss": 0.898,
      "step": 1247
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6285197754049275e-05,
      "loss": 0.8267,
      "step": 1248
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6279034636367912e-05,
      "loss": 0.8459,
      "step": 1249
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.627286757876856e-05,
      "loss": 0.799,
      "step": 1250
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6266696585120884e-05,
      "loss": 0.9174,
      "step": 1251
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6260521659296994e-05,
      "loss": 0.8843,
      "step": 1252
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6254342805171494e-05,
      "loss": 0.8936,
      "step": 1253
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.624816002662144e-05,
      "loss": 0.818,
      "step": 1254
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6241973327526347e-05,
      "loss": 0.8599,
      "step": 1255
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.62357827117682e-05,
      "loss": 0.7875,
      "step": 1256
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6229588183231436e-05,
      "loss": 0.8331,
      "step": 1257
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6223389745802943e-05,
      "loss": 0.8262,
      "step": 1258
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6217187403372074e-05,
      "loss": 0.882,
      "step": 1259
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.621098115983062e-05,
      "loss": 0.8678,
      "step": 1260
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6204771019072826e-05,
      "loss": 0.8622,
      "step": 1261
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.619855698499538e-05,
      "loss": 0.8854,
      "step": 1262
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6192339061497415e-05,
      "loss": 0.9076,
      "step": 1263
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6186117252480502e-05,
      "loss": 0.276,
      "step": 1264
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6179891561848653e-05,
      "loss": 0.9037,
      "step": 1265
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6173661993508312e-05,
      "loss": 0.2533,
      "step": 1266
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.616742855136836e-05,
      "loss": 0.8949,
      "step": 1267
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6161191239340112e-05,
      "loss": 0.8146,
      "step": 1268
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6154950061337295e-05,
      "loss": 0.9023,
      "step": 1269
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6148705021276082e-05,
      "loss": 0.8379,
      "step": 1270
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6142456123075056e-05,
      "loss": 0.8573,
      "step": 1271
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.613620337065523e-05,
      "loss": 0.8634,
      "step": 1272
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.612994676794002e-05,
      "loss": 0.8274,
      "step": 1273
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.612368631885528e-05,
      "loss": 0.8783,
      "step": 1274
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6117422027329267e-05,
      "loss": 0.8297,
      "step": 1275
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6111153897292634e-05,
      "loss": 0.8648,
      "step": 1276
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.610488193267847e-05,
      "loss": 0.9123,
      "step": 1277
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6098606137422254e-05,
      "loss": 0.2755,
      "step": 1278
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.609232651546187e-05,
      "loss": 0.7494,
      "step": 1279
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.60860430707376e-05,
      "loss": 0.8721,
      "step": 1280
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6079755807192138e-05,
      "loss": 0.8074,
      "step": 1281
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.607346472877056e-05,
      "loss": 0.86,
      "step": 1282
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.606716983942034e-05,
      "loss": 0.8708,
      "step": 1283
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.606087114309135e-05,
      "loss": 0.8128,
      "step": 1284
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.605456864373584e-05,
      "loss": 0.9414,
      "step": 1285
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.604826234530845e-05,
      "loss": 0.8162,
      "step": 1286
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.604195225176621e-05,
      "loss": 0.2522,
      "step": 1287
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6035638367068528e-05,
      "loss": 0.8824,
      "step": 1288
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6029320695177183e-05,
      "loss": 0.885,
      "step": 1289
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.602299924005634e-05,
      "loss": 0.8653,
      "step": 1290
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6016674005672536e-05,
      "loss": 0.8853,
      "step": 1291
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6010344995994674e-05,
      "loss": 0.9285,
      "step": 1292
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6004012214994036e-05,
      "loss": 0.2724,
      "step": 1293
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.599767566664426e-05,
      "loss": 0.8586,
      "step": 1294
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5991335354921353e-05,
      "loss": 0.8264,
      "step": 1295
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5984991283803684e-05,
      "loss": 0.7806,
      "step": 1296
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5978643457271977e-05,
      "loss": 0.8791,
      "step": 1297
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5972291879309315e-05,
      "loss": 0.8955,
      "step": 1298
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5965936553901137e-05,
      "loss": 0.8649,
      "step": 1299
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5959577485035232e-05,
      "loss": 0.8375,
      "step": 1300
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5953214676701737e-05,
      "loss": 0.8945,
      "step": 1301
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5946848132893133e-05,
      "loss": 0.831,
      "step": 1302
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5940477857604245e-05,
      "loss": 0.8991,
      "step": 1303
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5934103854832252e-05,
      "loss": 0.9188,
      "step": 1304
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.592772612857665e-05,
      "loss": 0.8202,
      "step": 1305
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5921344682839293e-05,
      "loss": 0.9011,
      "step": 1306
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5914959521624357e-05,
      "loss": 0.8247,
      "step": 1307
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5908570648938345e-05,
      "loss": 0.8521,
      "step": 1308
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5902178068790105e-05,
      "loss": 0.296,
      "step": 1309
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.589578178519079e-05,
      "loss": 0.8843,
      "step": 1310
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5889381802153898e-05,
      "loss": 0.9158,
      "step": 1311
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5882978123695233e-05,
      "loss": 0.883,
      "step": 1312
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5876570753832927e-05,
      "loss": 0.8716,
      "step": 1313
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5870159696587425e-05,
      "loss": 0.8797,
      "step": 1314
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5863744955981485e-05,
      "loss": 0.8276,
      "step": 1315
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5857326536040175e-05,
      "loss": 0.861,
      "step": 1316
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5850904440790872e-05,
      "loss": 0.2683,
      "step": 1317
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5844478674263265e-05,
      "loss": 0.8671,
      "step": 1318
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5838049240489337e-05,
      "loss": 0.2966,
      "step": 1319
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5831616143503383e-05,
      "loss": 0.9106,
      "step": 1320
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5825179387341988e-05,
      "loss": 0.8496,
      "step": 1321
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5818738976044033e-05,
      "loss": 0.8717,
      "step": 1322
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5812294913650697e-05,
      "loss": 0.8934,
      "step": 1323
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.580584720420545e-05,
      "loss": 0.8654,
      "step": 1324
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.579939585175405e-05,
      "loss": 0.8604,
      "step": 1325
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.579294086034454e-05,
      "loss": 0.8377,
      "step": 1326
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5786482234027237e-05,
      "loss": 0.8478,
      "step": 1327
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.578001997685476e-05,
      "loss": 0.885,
      "step": 1328
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5773554092881986e-05,
      "loss": 0.8265,
      "step": 1329
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5767084586166078e-05,
      "loss": 0.8414,
      "step": 1330
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5760611460766468e-05,
      "loss": 0.8742,
      "step": 1331
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5754134720744863e-05,
      "loss": 0.8458,
      "step": 1332
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5747654370165235e-05,
      "loss": 0.8499,
      "step": 1333
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.574117041309382e-05,
      "loss": 0.8625,
      "step": 1334
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5734682853599124e-05,
      "loss": 0.8496,
      "step": 1335
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5728191695751906e-05,
      "loss": 0.9153,
      "step": 1336
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.572169694362518e-05,
      "loss": 0.2803,
      "step": 1337
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5715198601294228e-05,
      "loss": 0.8442,
      "step": 1338
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5708696672836576e-05,
      "loss": 0.9068,
      "step": 1339
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5702191162332e-05,
      "loss": 0.884,
      "step": 1340
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5695682073862527e-05,
      "loss": 0.8658,
      "step": 1341
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.568916941151242e-05,
      "loss": 0.8554,
      "step": 1342
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5682653179368205e-05,
      "loss": 0.8579,
      "step": 1343
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5676133381518623e-05,
      "loss": 0.8281,
      "step": 1344
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5669610022054672e-05,
      "loss": 0.8748,
      "step": 1345
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.566308310506957e-05,
      "loss": 0.846,
      "step": 1346
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5656552634658776e-05,
      "loss": 0.8195,
      "step": 1347
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5650018614919984e-05,
      "loss": 0.7697,
      "step": 1348
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5643481049953102e-05,
      "loss": 0.833,
      "step": 1349
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5636939943860268e-05,
      "loss": 0.9133,
      "step": 1350
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5630395300745843e-05,
      "loss": 0.834,
      "step": 1351
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.562384712471641e-05,
      "loss": 0.9051,
      "step": 1352
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.561729541988076e-05,
      "loss": 0.8513,
      "step": 1353
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5610740190349908e-05,
      "loss": 0.8921,
      "step": 1354
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5604181440237077e-05,
      "loss": 0.8507,
      "step": 1355
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5597619173657695e-05,
      "loss": 0.8759,
      "step": 1356
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5591053394729404e-05,
      "loss": 0.9068,
      "step": 1357
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5584484107572045e-05,
      "loss": 0.8098,
      "step": 1358
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5577911316307658e-05,
      "loss": 0.8213,
      "step": 1359
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.557133502506049e-05,
      "loss": 0.8545,
      "step": 1360
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5564755237956973e-05,
      "loss": 0.8323,
      "step": 1361
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5558171959125747e-05,
      "loss": 0.8737,
      "step": 1362
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5551585192697628e-05,
      "loss": 0.8792,
      "step": 1363
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5544994942805632e-05,
      "loss": 0.8459,
      "step": 1364
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.553840121358495e-05,
      "loss": 0.8585,
      "step": 1365
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5531804009172964e-05,
      "loss": 0.8205,
      "step": 1366
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.552520333370924e-05,
      "loss": 0.8303,
      "step": 1367
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5518599191335505e-05,
      "loss": 0.8829,
      "step": 1368
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.551199158619569e-05,
      "loss": 0.82,
      "step": 1369
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5505380522435865e-05,
      "loss": 0.8564,
      "step": 1370
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5498766004204294e-05,
      "loss": 0.881,
      "step": 1371
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.54921480356514e-05,
      "loss": 0.8494,
      "step": 1372
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5485526620929778e-05,
      "loss": 0.8533,
      "step": 1373
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5478901764194177e-05,
      "loss": 0.8442,
      "step": 1374
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5472273469601504e-05,
      "loss": 0.2716,
      "step": 1375
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.546564174131083e-05,
      "loss": 0.8941,
      "step": 1376
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.545900658348338e-05,
      "loss": 0.9167,
      "step": 1377
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5452368000282533e-05,
      "loss": 0.8547,
      "step": 1378
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.544572599587381e-05,
      "loss": 0.8786,
      "step": 1379
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5439080574424878e-05,
      "loss": 0.8714,
      "step": 1380
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5432431740105554e-05,
      "loss": 0.8495,
      "step": 1381
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5425779497087804e-05,
      "loss": 0.8427,
      "step": 1382
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.541912384954571e-05,
      "loss": 0.7365,
      "step": 1383
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5412464801655512e-05,
      "loss": 0.8505,
      "step": 1384
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.540580235759558e-05,
      "loss": 0.9225,
      "step": 1385
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.53991365215464e-05,
      "loss": 0.8651,
      "step": 1386
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5392467297690605e-05,
      "loss": 0.8269,
      "step": 1387
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.538579469021294e-05,
      "loss": 0.7819,
      "step": 1388
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5379118703300285e-05,
      "loss": 0.8873,
      "step": 1389
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.537243934114163e-05,
      "loss": 0.8152,
      "step": 1390
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.536575660792809e-05,
      "loss": 0.8328,
      "step": 1391
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.535907050785289e-05,
      "loss": 0.8623,
      "step": 1392
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.535238104511137e-05,
      "loss": 0.9442,
      "step": 1393
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5345688223900984e-05,
      "loss": 0.8688,
      "step": 1394
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.533899204842128e-05,
      "loss": 0.2701,
      "step": 1395
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5332292522873933e-05,
      "loss": 0.8783,
      "step": 1396
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5325589651462704e-05,
      "loss": 0.8797,
      "step": 1397
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.531888343839345e-05,
      "loss": 0.8294,
      "step": 1398
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5312173887874137e-05,
      "loss": 0.2684,
      "step": 1399
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5305461004114813e-05,
      "loss": 0.845,
      "step": 1400
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5298744791327632e-05,
      "loss": 0.8276,
      "step": 1401
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5292025253726826e-05,
      "loss": 0.8585,
      "step": 1402
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5285302395528716e-05,
      "loss": 0.8688,
      "step": 1403
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.527857622095171e-05,
      "loss": 0.8667,
      "step": 1404
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5271846734216284e-05,
      "loss": 0.8048,
      "step": 1405
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5265113939545014e-05,
      "loss": 0.8686,
      "step": 1406
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5258377841162534e-05,
      "loss": 0.8206,
      "step": 1407
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5251638443295557e-05,
      "loss": 0.2623,
      "step": 1408
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5244895750172868e-05,
      "loss": 0.788,
      "step": 1409
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5238149766025311e-05,
      "loss": 0.8513,
      "step": 1410
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.523140049508581e-05,
      "loss": 0.8878,
      "step": 1411
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.522464794158934e-05,
      "loss": 0.8278,
      "step": 1412
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5217892109772936e-05,
      "loss": 0.8132,
      "step": 1413
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5211133003875697e-05,
      "loss": 0.8416,
      "step": 1414
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5204370628138769e-05,
      "loss": 0.8155,
      "step": 1415
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5197604986805356e-05,
      "loss": 0.2641,
      "step": 1416
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5190836084120706e-05,
      "loss": 0.8674,
      "step": 1417
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5184063924332115e-05,
      "loss": 0.9092,
      "step": 1418
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5177288511688927e-05,
      "loss": 0.8192,
      "step": 1419
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5170509850442519e-05,
      "loss": 0.851,
      "step": 1420
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5163727944846312e-05,
      "loss": 0.8583,
      "step": 1421
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5156942799155759e-05,
      "loss": 0.892,
      "step": 1422
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5150154417628353e-05,
      "loss": 0.8866,
      "step": 1423
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5143362804523609e-05,
      "loss": 0.8071,
      "step": 1424
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5136567964103077e-05,
      "loss": 0.8925,
      "step": 1425
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5129769900630326e-05,
      "loss": 0.8453,
      "step": 1426
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.512296861837095e-05,
      "loss": 0.8554,
      "step": 1427
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5116164121592566e-05,
      "loss": 0.86,
      "step": 1428
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.51093564145648e-05,
      "loss": 0.8546,
      "step": 1429
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5102545501559302e-05,
      "loss": 0.8299,
      "step": 1430
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5095731386849725e-05,
      "loss": 0.9058,
      "step": 1431
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5088914074711735e-05,
      "loss": 0.844,
      "step": 1432
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5082093569423006e-05,
      "loss": 0.8695,
      "step": 1433
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5075269875263212e-05,
      "loss": 0.8555,
      "step": 1434
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.506844299651403e-05,
      "loss": 0.8442,
      "step": 1435
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5061612937459136e-05,
      "loss": 0.2431,
      "step": 1436
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5054779702384199e-05,
      "loss": 0.8225,
      "step": 1437
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5047943295576882e-05,
      "loss": 0.8498,
      "step": 1438
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5041103721326838e-05,
      "loss": 0.8486,
      "step": 1439
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5034260983925708e-05,
      "loss": 0.8477,
      "step": 1440
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.502741508766712e-05,
      "loss": 0.8384,
      "step": 1441
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5020566036846682e-05,
      "loss": 0.8619,
      "step": 1442
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5013713835761978e-05,
      "loss": 0.8929,
      "step": 1443
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.500685848871257e-05,
      "loss": 0.8191,
      "step": 1444
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 0.8349,
      "step": 1445
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4993138373927778e-05,
      "loss": 0.8322,
      "step": 1446
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4986273614801379e-05,
      "loss": 0.8003,
      "step": 1447
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4979405726928247e-05,
      "loss": 0.7981,
      "step": 1448
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4972534714617792e-05,
      "loss": 0.814,
      "step": 1449
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4965660582181374e-05,
      "loss": 0.8095,
      "step": 1450
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4958783333932328e-05,
      "loss": 0.8332,
      "step": 1451
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4951902974185924e-05,
      "loss": 0.8742,
      "step": 1452
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4945019507259401e-05,
      "loss": 0.9018,
      "step": 1453
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.493813293747194e-05,
      "loss": 0.8215,
      "step": 1454
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4931243269144673e-05,
      "loss": 0.2738,
      "step": 1455
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4924350506600669e-05,
      "loss": 0.2557,
      "step": 1456
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4917454654164944e-05,
      "loss": 0.8054,
      "step": 1457
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4910555716164454e-05,
      "loss": 0.8524,
      "step": 1458
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4903653696928086e-05,
      "loss": 0.8283,
      "step": 1459
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4896748600786665e-05,
      "loss": 0.25,
      "step": 1460
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4889840432072947e-05,
      "loss": 0.8192,
      "step": 1461
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4882929195121609e-05,
      "loss": 0.8908,
      "step": 1462
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4876014894269264e-05,
      "loss": 0.8599,
      "step": 1463
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4869097533854438e-05,
      "loss": 0.8713,
      "step": 1464
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4862177118217578e-05,
      "loss": 0.2489,
      "step": 1465
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4855253651701057e-05,
      "loss": 0.8445,
      "step": 1466
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4848327138649153e-05,
      "loss": 0.8568,
      "step": 1467
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4841397583408057e-05,
      "loss": 0.8331,
      "step": 1468
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.483446499032587e-05,
      "loss": 0.9057,
      "step": 1469
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4827529363752605e-05,
      "loss": 0.85,
      "step": 1470
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4820590708040166e-05,
      "loss": 0.8874,
      "step": 1471
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4813649027542368e-05,
      "loss": 0.8535,
      "step": 1472
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.480670432661492e-05,
      "loss": 0.8264,
      "step": 1473
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.479975660961543e-05,
      "loss": 0.2432,
      "step": 1474
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4792805880903391e-05,
      "loss": 0.8379,
      "step": 1475
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4785852144840193e-05,
      "loss": 0.8473,
      "step": 1476
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4778895405789109e-05,
      "loss": 0.864,
      "step": 1477
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4771935668115297e-05,
      "loss": 0.8283,
      "step": 1478
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4764972936185798e-05,
      "loss": 0.8874,
      "step": 1479
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4758007214369532e-05,
      "loss": 0.8542,
      "step": 1480
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4751038507037289e-05,
      "loss": 0.8583,
      "step": 1481
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4744066818561744e-05,
      "loss": 0.8672,
      "step": 1482
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4737092153317432e-05,
      "loss": 0.8273,
      "step": 1483
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.473011451568076e-05,
      "loss": 0.7893,
      "step": 1484
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4723133910029998e-05,
      "loss": 0.8238,
      "step": 1485
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4716150340745286e-05,
      "loss": 0.2817,
      "step": 1486
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4709163812208613e-05,
      "loss": 0.8803,
      "step": 1487
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4702174328803828e-05,
      "loss": 0.8722,
      "step": 1488
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4695181894916639e-05,
      "loss": 0.8538,
      "step": 1489
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4688186514934601e-05,
      "loss": 0.8455,
      "step": 1490
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4681188193247118e-05,
      "loss": 0.8113,
      "step": 1491
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4674186934245442e-05,
      "loss": 0.8739,
      "step": 1492
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4667182742322665e-05,
      "loss": 0.8173,
      "step": 1493
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4660175621873721e-05,
      "loss": 0.9074,
      "step": 1494
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4653165577295383e-05,
      "loss": 0.8217,
      "step": 1495
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4646152612986257e-05,
      "loss": 0.8198,
      "step": 1496
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4639136733346777e-05,
      "loss": 0.8549,
      "step": 1497
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4632117942779214e-05,
      "loss": 0.8707,
      "step": 1498
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4625096245687665e-05,
      "loss": 0.8522,
      "step": 1499
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.461807164647804e-05,
      "loss": 0.8554,
      "step": 1500
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4611044149558087e-05,
      "loss": 0.8565,
      "step": 1501
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4604013759337354e-05,
      "loss": 0.8461,
      "step": 1502
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4596980480227223e-05,
      "loss": 0.8043,
      "step": 1503
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.458994431664087e-05,
      "loss": 0.8363,
      "step": 1504
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4582905272993295e-05,
      "loss": 0.8424,
      "step": 1505
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4575863353701303e-05,
      "loss": 0.8535,
      "step": 1506
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4568818563183493e-05,
      "loss": 0.7604,
      "step": 1507
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4561770905860281e-05,
      "loss": 0.8227,
      "step": 1508
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4554720386153872e-05,
      "loss": 0.8716,
      "step": 1509
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4547667008488264e-05,
      "loss": 0.847,
      "step": 1510
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.454061077728926e-05,
      "loss": 0.8294,
      "step": 1511
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4533551696984446e-05,
      "loss": 0.864,
      "step": 1512
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4526489772003194e-05,
      "loss": 0.7709,
      "step": 1513
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4519425006776664e-05,
      "loss": 0.8359,
      "step": 1514
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4512357405737798e-05,
      "loss": 0.7997,
      "step": 1515
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.450528697332132e-05,
      "loss": 0.265,
      "step": 1516
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4498213713963724e-05,
      "loss": 0.8072,
      "step": 1517
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4491137632103289e-05,
      "loss": 0.3179,
      "step": 1518
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.448405873218005e-05,
      "loss": 0.8547,
      "step": 1519
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4476977018635818e-05,
      "loss": 0.8275,
      "step": 1520
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4469892495914174e-05,
      "loss": 0.8067,
      "step": 1521
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4462805168460454e-05,
      "loss": 0.8533,
      "step": 1522
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4455715040721758e-05,
      "loss": 0.8505,
      "step": 1523
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4448622117146945e-05,
      "loss": 0.8098,
      "step": 1524
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.444152640218662e-05,
      "loss": 0.8601,
      "step": 1525
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4434427900293148e-05,
      "loss": 0.8214,
      "step": 1526
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4427326615920642e-05,
      "loss": 0.8753,
      "step": 1527
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4420222553524953e-05,
      "loss": 0.7989,
      "step": 1528
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4413115717563689e-05,
      "loss": 0.8258,
      "step": 1529
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.440600611249618e-05,
      "loss": 0.8054,
      "step": 1530
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4398893742783508e-05,
      "loss": 0.8145,
      "step": 1531
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4391778612888492e-05,
      "loss": 0.8098,
      "step": 1532
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4384660727275664e-05,
      "loss": 0.7706,
      "step": 1533
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4377540090411304e-05,
      "loss": 0.8802,
      "step": 1534
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4370416706763415e-05,
      "loss": 0.8197,
      "step": 1535
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4363290580801713e-05,
      "loss": 0.8291,
      "step": 1536
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4356161716997647e-05,
      "loss": 0.8207,
      "step": 1537
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4349030119824379e-05,
      "loss": 0.8458,
      "step": 1538
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4341895793756784e-05,
      "loss": 0.9009,
      "step": 1539
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.433475874327145e-05,
      "loss": 0.903,
      "step": 1540
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.432761897284668e-05,
      "loss": 0.8468,
      "step": 1541
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4320476486962476e-05,
      "loss": 0.8218,
      "step": 1542
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4313331290100547e-05,
      "loss": 0.7766,
      "step": 1543
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.430618338674431e-05,
      "loss": 0.7735,
      "step": 1544
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4299032781378865e-05,
      "loss": 0.9459,
      "step": 1545
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4291879478491025e-05,
      "loss": 0.8549,
      "step": 1546
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4284723482569277e-05,
      "loss": 0.8599,
      "step": 1547
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4277564798103816e-05,
      "loss": 0.8114,
      "step": 1548
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4270403429586508e-05,
      "loss": 0.8643,
      "step": 1549
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4263239381510916e-05,
      "loss": 0.8313,
      "step": 1550
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.425607265837228e-05,
      "loss": 0.8375,
      "step": 1551
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4248903264667515e-05,
      "loss": 0.83,
      "step": 1552
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4241731204895213e-05,
      "loss": 0.7964,
      "step": 1553
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4234556483555647e-05,
      "loss": 0.8174,
      "step": 1554
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4227379105150742e-05,
      "loss": 0.7308,
      "step": 1555
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4220199074184112e-05,
      "loss": 0.7683,
      "step": 1556
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4213016395161017e-05,
      "loss": 0.8271,
      "step": 1557
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4205831072588391e-05,
      "loss": 0.8343,
      "step": 1558
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4198643110974822e-05,
      "loss": 0.2549,
      "step": 1559
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4191452514830552e-05,
      "loss": 0.8681,
      "step": 1560
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4184259288667484e-05,
      "loss": 0.8345,
      "step": 1561
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4177063436999157e-05,
      "loss": 0.792,
      "step": 1562
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.416986496434077e-05,
      "loss": 0.8237,
      "step": 1563
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4162663875209166e-05,
      "loss": 0.9091,
      "step": 1564
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4155460174122824e-05,
      "loss": 0.7793,
      "step": 1565
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4148253865601857e-05,
      "loss": 0.8608,
      "step": 1566
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4141044954168034e-05,
      "loss": 0.8468,
      "step": 1567
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4133833444344733e-05,
      "loss": 0.8016,
      "step": 1568
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4126619340656981e-05,
      "loss": 0.8368,
      "step": 1569
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4119402647631426e-05,
      "loss": 0.2538,
      "step": 1570
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4112183369796335e-05,
      "loss": 0.9261,
      "step": 1571
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4104961511681605e-05,
      "loss": 0.8417,
      "step": 1572
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4097737077818752e-05,
      "loss": 0.8536,
      "step": 1573
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4090510072740902e-05,
      "loss": 0.7752,
      "step": 1574
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4083280500982797e-05,
      "loss": 0.8118,
      "step": 1575
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4076048367080793e-05,
      "loss": 0.8081,
      "step": 1576
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4068813675572852e-05,
      "loss": 0.8911,
      "step": 1577
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4061576430998536e-05,
      "loss": 0.8256,
      "step": 1578
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4054336637899015e-05,
      "loss": 0.2707,
      "step": 1579
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4047094300817056e-05,
      "loss": 0.8493,
      "step": 1580
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4039849424297023e-05,
      "loss": 0.8391,
      "step": 1581
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4032602012884871e-05,
      "loss": 0.8783,
      "step": 1582
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4025352071128147e-05,
      "loss": 0.8299,
      "step": 1583
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4018099603575988e-05,
      "loss": 0.8052,
      "step": 1584
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.401084461477911e-05,
      "loss": 0.2475,
      "step": 1585
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4003587109289817e-05,
      "loss": 0.8271,
      "step": 1586
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3996327091661996e-05,
      "loss": 0.8608,
      "step": 1587
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3989064566451094e-05,
      "loss": 0.8085,
      "step": 1588
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.398179953821415e-05,
      "loss": 0.8452,
      "step": 1589
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3974532011509759e-05,
      "loss": 0.803,
      "step": 1590
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3967261990898093e-05,
      "loss": 0.2575,
      "step": 1591
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3959989480940886e-05,
      "loss": 0.7812,
      "step": 1592
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3952714486201436e-05,
      "loss": 0.8191,
      "step": 1593
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3945437011244591e-05,
      "loss": 0.2764,
      "step": 1594
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3938157060636768e-05,
      "loss": 0.8477,
      "step": 1595
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.393087463894593e-05,
      "loss": 0.8007,
      "step": 1596
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3923589750741595e-05,
      "loss": 0.9021,
      "step": 1597
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3916302400594823e-05,
      "loss": 0.8551,
      "step": 1598
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3909012593078224e-05,
      "loss": 0.8807,
      "step": 1599
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.390172033276594e-05,
      "loss": 0.832,
      "step": 1600
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.389442562423367e-05,
      "loss": 0.8634,
      "step": 1601
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3887128472058634e-05,
      "loss": 0.8459,
      "step": 1602
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3879828880819588e-05,
      "loss": 0.8574,
      "step": 1603
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3872526855096824e-05,
      "loss": 0.8725,
      "step": 1604
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3865222399472155e-05,
      "loss": 0.7995,
      "step": 1605
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3857915518528931e-05,
      "loss": 0.8569,
      "step": 1606
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3850606216852e-05,
      "loss": 0.8438,
      "step": 1607
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3843294499027757e-05,
      "loss": 0.806,
      "step": 1608
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.383598036964409e-05,
      "loss": 0.8104,
      "step": 1609
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3828663833290416e-05,
      "loss": 0.8047,
      "step": 1610
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3821344894557653e-05,
      "loss": 0.2665,
      "step": 1611
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3814023558038235e-05,
      "loss": 0.8786,
      "step": 1612
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3806699828326084e-05,
      "loss": 0.7974,
      "step": 1613
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3799373710016645e-05,
      "loss": 0.8438,
      "step": 1614
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.379204520770685e-05,
      "loss": 0.783,
      "step": 1615
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3784714325995126e-05,
      "loss": 0.859,
      "step": 1616
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3777381069481397e-05,
      "loss": 0.7597,
      "step": 1617
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3770045442767073e-05,
      "loss": 0.8327,
      "step": 1618
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3762707450455058e-05,
      "loss": 0.8885,
      "step": 1619
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3755367097149733e-05,
      "loss": 0.8567,
      "step": 1620
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3748024387456965e-05,
      "loss": 0.8772,
      "step": 1621
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3740679325984099e-05,
      "loss": 0.8208,
      "step": 1622
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3733331917339952e-05,
      "loss": 0.8439,
      "step": 1623
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3725982166134819e-05,
      "loss": 0.7983,
      "step": 1624
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3718630076980461e-05,
      "loss": 0.79,
      "step": 1625
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.371127565449011e-05,
      "loss": 0.2657,
      "step": 1626
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.370391890327846e-05,
      "loss": 0.8655,
      "step": 1627
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3696559827961661e-05,
      "loss": 0.7975,
      "step": 1628
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3689198433157334e-05,
      "loss": 0.7883,
      "step": 1629
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.368183472348454e-05,
      "loss": 0.2726,
      "step": 1630
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3674468703563801e-05,
      "loss": 0.8767,
      "step": 1631
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3667100378017093e-05,
      "loss": 0.8286,
      "step": 1632
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3659729751467829e-05,
      "loss": 0.8572,
      "step": 1633
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3652356828540874e-05,
      "loss": 0.845,
      "step": 1634
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3644981613862524e-05,
      "loss": 0.8701,
      "step": 1635
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3637604112060527e-05,
      "loss": 0.8369,
      "step": 1636
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3630224327764052e-05,
      "loss": 0.7513,
      "step": 1637
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.362284226560371e-05,
      "loss": 0.8495,
      "step": 1638
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3615457930211538e-05,
      "loss": 0.8447,
      "step": 1639
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3608071326220992e-05,
      "loss": 0.8546,
      "step": 1640
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3600682458266972e-05,
      "loss": 0.8402,
      "step": 1641
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3593291330985772e-05,
      "loss": 0.8183,
      "step": 1642
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3585897949015126e-05,
      "loss": 0.8266,
      "step": 1643
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3578502316994169e-05,
      "loss": 0.9013,
      "step": 1644
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3571104439563456e-05,
      "loss": 0.853,
      "step": 1645
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3563704321364943e-05,
      "loss": 0.8159,
      "step": 1646
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3556301967041999e-05,
      "loss": 0.8079,
      "step": 1647
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3548897381239394e-05,
      "loss": 0.82,
      "step": 1648
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3541490568603295e-05,
      "loss": 0.8115,
      "step": 1649
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3534081533781272e-05,
      "loss": 0.8373,
      "step": 1650
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3526670281422287e-05,
      "loss": 0.7986,
      "step": 1651
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.351925681617669e-05,
      "loss": 0.2789,
      "step": 1652
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3511841142696222e-05,
      "loss": 0.8074,
      "step": 1653
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3504423265634015e-05,
      "loss": 0.8065,
      "step": 1654
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3497003189644574e-05,
      "loss": 0.862,
      "step": 1655
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3489580919383788e-05,
      "loss": 0.2774,
      "step": 1656
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3482156459508925e-05,
      "loss": 0.8453,
      "step": 1657
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3474729814678626e-05,
      "loss": 0.8412,
      "step": 1658
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.34673009895529e-05,
      "loss": 0.8393,
      "step": 1659
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.345986998879313e-05,
      "loss": 0.8229,
      "step": 1660
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3452436817062057e-05,
      "loss": 0.8104,
      "step": 1661
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.344500147902379e-05,
      "loss": 0.8142,
      "step": 1662
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3437563979343788e-05,
      "loss": 0.2829,
      "step": 1663
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3430124322688885e-05,
      "loss": 0.8338,
      "step": 1664
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3422682513727246e-05,
      "loss": 0.8538,
      "step": 1665
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3415238557128403e-05,
      "loss": 0.8602,
      "step": 1666
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3407792457563231e-05,
      "loss": 0.8612,
      "step": 1667
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3400344219703943e-05,
      "loss": 0.842,
      "step": 1668
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.33928938482241e-05,
      "loss": 0.8458,
      "step": 1669
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3385441347798607e-05,
      "loss": 0.8273,
      "step": 1670
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3377986723103694e-05,
      "loss": 0.8457,
      "step": 1671
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3370529978816928e-05,
      "loss": 0.8062,
      "step": 1672
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.336307111961721e-05,
      "loss": 0.8591,
      "step": 1673
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3355610150184763e-05,
      "loss": 0.851,
      "step": 1674
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3348147075201137e-05,
      "loss": 0.8808,
      "step": 1675
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3340681899349197e-05,
      "loss": 0.8664,
      "step": 1676
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.333321462731314e-05,
      "loss": 0.8398,
      "step": 1677
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3325745263778465e-05,
      "loss": 0.8461,
      "step": 1678
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3318273813431986e-05,
      "loss": 0.8344,
      "step": 1679
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3310800280961831e-05,
      "loss": 0.7957,
      "step": 1680
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3303324671057432e-05,
      "loss": 0.8458,
      "step": 1681
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3295846988409522e-05,
      "loss": 0.2801,
      "step": 1682
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.328836723771014e-05,
      "loss": 0.8201,
      "step": 1683
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3280885423652616e-05,
      "loss": 0.871,
      "step": 1684
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3273401550931581e-05,
      "loss": 0.8214,
      "step": 1685
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3265915624242957e-05,
      "loss": 0.9009,
      "step": 1686
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.325842764828395e-05,
      "loss": 0.8367,
      "step": 1687
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3250937627753056e-05,
      "loss": 0.8003,
      "step": 1688
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3243445567350048e-05,
      "loss": 0.8962,
      "step": 1689
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3235951471775992e-05,
      "loss": 0.7773,
      "step": 1690
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3228455345733215e-05,
      "loss": 0.8623,
      "step": 1691
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.322095719392533e-05,
      "loss": 0.8242,
      "step": 1692
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3213457021057216e-05,
      "loss": 0.831,
      "step": 1693
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3205954831835017e-05,
      "loss": 0.8899,
      "step": 1694
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3198450630966153e-05,
      "loss": 0.7841,
      "step": 1695
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3190944423159295e-05,
      "loss": 0.8157,
      "step": 1696
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3183436213124379e-05,
      "loss": 0.8543,
      "step": 1697
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3175926005572597e-05,
      "loss": 0.7898,
      "step": 1698
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3168413805216389e-05,
      "loss": 0.8418,
      "step": 1699
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3160899616769457e-05,
      "loss": 0.7764,
      "step": 1700
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3153383444946736e-05,
      "loss": 0.8301,
      "step": 1701
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3145865294464417e-05,
      "loss": 0.909,
      "step": 1702
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.313834517003993e-05,
      "loss": 0.7752,
      "step": 1703
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3130823076391939e-05,
      "loss": 0.7922,
      "step": 1704
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3123299018240345e-05,
      "loss": 0.8791,
      "step": 1705
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3115773000306287e-05,
      "loss": 0.8216,
      "step": 1706
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.310824502731213e-05,
      "loss": 0.8577,
      "step": 1707
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3100715103981462e-05,
      "loss": 0.8471,
      "step": 1708
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3093183235039102e-05,
      "loss": 0.8139,
      "step": 1709
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3085649425211085e-05,
      "loss": 0.832,
      "step": 1710
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3078113679224665e-05,
      "loss": 0.8138,
      "step": 1711
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3070576001808315e-05,
      "loss": 0.8201,
      "step": 1712
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3063036397691708e-05,
      "loss": 0.8487,
      "step": 1713
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.305549487160574e-05,
      "loss": 0.8102,
      "step": 1714
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3047951428282506e-05,
      "loss": 0.8763,
      "step": 1715
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3040406072455303e-05,
      "loss": 0.8387,
      "step": 1716
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3032858808858633e-05,
      "loss": 0.8301,
      "step": 1717
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.302530964222819e-05,
      "loss": 0.8345,
      "step": 1718
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3017758577300863e-05,
      "loss": 0.8234,
      "step": 1719
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3010205618814736e-05,
      "loss": 0.8667,
      "step": 1720
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3002650771509076e-05,
      "loss": 0.8083,
      "step": 1721
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2995094040124341e-05,
      "loss": 0.86,
      "step": 1722
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2987535429402165e-05,
      "loss": 0.855,
      "step": 1723
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2979974944085366e-05,
      "loss": 0.865,
      "step": 1724
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2972412588917931e-05,
      "loss": 0.8574,
      "step": 1725
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2964848368645035e-05,
      "loss": 0.8586,
      "step": 1726
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2957282288013004e-05,
      "loss": 0.8695,
      "step": 1727
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2949714351769346e-05,
      "loss": 0.7916,
      "step": 1728
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2942144564662733e-05,
      "loss": 0.7962,
      "step": 1729
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2934572931442985e-05,
      "loss": 0.8004,
      "step": 1730
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2926999456861099e-05,
      "loss": 0.8227,
      "step": 1731
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2919424145669205e-05,
      "loss": 0.8136,
      "step": 1732
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2911847002620612e-05,
      "loss": 0.8263,
      "step": 1733
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2904268032469753e-05,
      "loss": 0.8725,
      "step": 1734
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.289668723997223e-05,
      "loss": 0.8231,
      "step": 1735
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2889104629884768e-05,
      "loss": 0.8549,
      "step": 1736
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2881520206965245e-05,
      "loss": 0.8339,
      "step": 1737
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2873933975972677e-05,
      "loss": 0.8358,
      "step": 1738
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2866345941667204e-05,
      "loss": 0.8555,
      "step": 1739
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.285875610881011e-05,
      "loss": 0.8217,
      "step": 1740
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2851164482163803e-05,
      "loss": 0.8638,
      "step": 1741
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2843571066491811e-05,
      "loss": 0.796,
      "step": 1742
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2835975866558793e-05,
      "loss": 0.8838,
      "step": 1743
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2828378887130522e-05,
      "loss": 0.8713,
      "step": 1744
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.282078013297389e-05,
      "loss": 0.8296,
      "step": 1745
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2813179608856898e-05,
      "loss": 0.826,
      "step": 1746
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.280557731954867e-05,
      "loss": 0.7952,
      "step": 1747
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2797973269819424e-05,
      "loss": 0.839,
      "step": 1748
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2790367464440484e-05,
      "loss": 0.8542,
      "step": 1749
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2782759908184288e-05,
      "loss": 0.8799,
      "step": 1750
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2775150605824357e-05,
      "loss": 0.7969,
      "step": 1751
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2767539562135319e-05,
      "loss": 0.886,
      "step": 1752
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2759926781892884e-05,
      "loss": 0.8758,
      "step": 1753
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2752312269873868e-05,
      "loss": 0.7667,
      "step": 1754
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2744696030856154e-05,
      "loss": 0.8235,
      "step": 1755
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2737078069618722e-05,
      "loss": 0.88,
      "step": 1756
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2729458390941635e-05,
      "loss": 0.8328,
      "step": 1757
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2721836999606016e-05,
      "loss": 0.8651,
      "step": 1758
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.271421390039408e-05,
      "loss": 0.2829,
      "step": 1759
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2706589098089108e-05,
      "loss": 0.7958,
      "step": 1760
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2698962597475446e-05,
      "loss": 0.8305,
      "step": 1761
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2691334403338517e-05,
      "loss": 0.8716,
      "step": 1762
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.268370452046479e-05,
      "loss": 0.862,
      "step": 1763
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2676072953641807e-05,
      "loss": 0.8362,
      "step": 1764
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.266843970765816e-05,
      "loss": 0.8975,
      "step": 1765
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2660804787303497e-05,
      "loss": 0.8207,
      "step": 1766
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.265316819736852e-05,
      "loss": 0.2527,
      "step": 1767
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2645529942644968e-05,
      "loss": 0.9022,
      "step": 1768
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2637890027925637e-05,
      "loss": 0.807,
      "step": 1769
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2630248458004355e-05,
      "loss": 0.8706,
      "step": 1770
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2622605237675995e-05,
      "loss": 0.7978,
      "step": 1771
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2614960371736464e-05,
      "loss": 0.7551,
      "step": 1772
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2607313864982698e-05,
      "loss": 0.8379,
      "step": 1773
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2599665722212668e-05,
      "loss": 0.8286,
      "step": 1774
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.259201594822537e-05,
      "loss": 0.827,
      "step": 1775
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2584364547820818e-05,
      "loss": 0.7848,
      "step": 1776
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2576711525800052e-05,
      "loss": 0.8177,
      "step": 1777
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2569056886965134e-05,
      "loss": 0.8144,
      "step": 1778
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2561400636119126e-05,
      "loss": 0.7857,
      "step": 1779
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2553742778066118e-05,
      "loss": 0.8666,
      "step": 1780
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2546083317611199e-05,
      "loss": 0.8385,
      "step": 1781
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2538422259560464e-05,
      "loss": 0.828,
      "step": 1782
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2530759608721012e-05,
      "loss": 0.7821,
      "step": 1783
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2523095369900942e-05,
      "loss": 0.8058,
      "step": 1784
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2515429547909349e-05,
      "loss": 0.8122,
      "step": 1785
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2507762147556321e-05,
      "loss": 0.7745,
      "step": 1786
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2500093173652933e-05,
      "loss": 0.8648,
      "step": 1787
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2492422631011261e-05,
      "loss": 0.2692,
      "step": 1788
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2484750524444342e-05,
      "loss": 0.267,
      "step": 1789
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.247707685876622e-05,
      "loss": 0.273,
      "step": 1790
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2469401638791892e-05,
      "loss": 0.825,
      "step": 1791
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2461724869337353e-05,
      "loss": 0.8243,
      "step": 1792
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2454046555219557e-05,
      "loss": 0.8904,
      "step": 1793
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.244636670125643e-05,
      "loss": 0.8942,
      "step": 1794
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2438685312266867e-05,
      "loss": 0.8345,
      "step": 1795
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2431002393070722e-05,
      "loss": 0.2812,
      "step": 1796
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2423317948488814e-05,
      "loss": 0.8918,
      "step": 1797
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2415631983342908e-05,
      "loss": 0.8223,
      "step": 1798
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2407944502455742e-05,
      "loss": 0.7873,
      "step": 1799
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2400255510650989e-05,
      "loss": 0.7845,
      "step": 1800
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2392565012753275e-05,
      "loss": 0.2445,
      "step": 1801
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2384873013588177e-05,
      "loss": 0.889,
      "step": 1802
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.23771795179822e-05,
      "loss": 0.2589,
      "step": 1803
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2369484530762803e-05,
      "loss": 0.8279,
      "step": 1804
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2361788056758373e-05,
      "loss": 0.8494,
      "step": 1805
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2354090100798229e-05,
      "loss": 0.7905,
      "step": 1806
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2346390667712625e-05,
      "loss": 0.847,
      "step": 1807
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2338689762332738e-05,
      "loss": 0.7736,
      "step": 1808
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2330987389490671e-05,
      "loss": 0.8724,
      "step": 1809
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2323283554019446e-05,
      "loss": 0.7629,
      "step": 1810
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2315578260753004e-05,
      "loss": 0.8509,
      "step": 1811
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2307871514526196e-05,
      "loss": 0.8227,
      "step": 1812
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2300163320174798e-05,
      "loss": 0.794,
      "step": 1813
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2292453682535479e-05,
      "loss": 0.8003,
      "step": 1814
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2284742606445818e-05,
      "loss": 0.7734,
      "step": 1815
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2277030096744307e-05,
      "loss": 0.8413,
      "step": 1816
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2269316158270324e-05,
      "loss": 0.2507,
      "step": 1817
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2261600795864146e-05,
      "loss": 0.822,
      "step": 1818
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2253884014366955e-05,
      "loss": 0.8,
      "step": 1819
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2246165818620805e-05,
      "loss": 0.7542,
      "step": 1820
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2238446213468656e-05,
      "loss": 0.777,
      "step": 1821
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2230725203754336e-05,
      "loss": 0.7912,
      "step": 1822
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.222300279432257e-05,
      "loss": 0.8595,
      "step": 1823
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2215278990018944e-05,
      "loss": 0.891,
      "step": 1824
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2207553795689935e-05,
      "loss": 0.8418,
      "step": 1825
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2199827216182885e-05,
      "loss": 0.8327,
      "step": 1826
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2192099256345999e-05,
      "loss": 0.8049,
      "step": 1827
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2184369921028365e-05,
      "loss": 0.8043,
      "step": 1828
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2176639215079917e-05,
      "loss": 0.8507,
      "step": 1829
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2168907143351457e-05,
      "loss": 0.8646,
      "step": 1830
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2161173710694647e-05,
      "loss": 0.874,
      "step": 1831
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2153438921961994e-05,
      "loss": 0.8033,
      "step": 1832
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2145702782006863e-05,
      "loss": 0.8559,
      "step": 1833
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2137965295683465e-05,
      "loss": 0.8524,
      "step": 1834
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2130226467846855e-05,
      "loss": 0.805,
      "step": 1835
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.212248630335293e-05,
      "loss": 0.8502,
      "step": 1836
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2114744807058425e-05,
      "loss": 0.8486,
      "step": 1837
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2107001983820913e-05,
      "loss": 0.8514,
      "step": 1838
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.20992578384988e-05,
      "loss": 0.8155,
      "step": 1839
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2091512375951317e-05,
      "loss": 0.8219,
      "step": 1840
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2083765601038523e-05,
      "loss": 0.8947,
      "step": 1841
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2076017518621308e-05,
      "loss": 0.7694,
      "step": 1842
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.206826813356137e-05,
      "loss": 0.8603,
      "step": 1843
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2060517450721231e-05,
      "loss": 0.858,
      "step": 1844
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2052765474964232e-05,
      "loss": 0.804,
      "step": 1845
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2045012211154514e-05,
      "loss": 0.8011,
      "step": 1846
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2037257664157043e-05,
      "loss": 0.8148,
      "step": 1847
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2029501838837563e-05,
      "loss": 0.8822,
      "step": 1848
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2021744740062648e-05,
      "loss": 0.8225,
      "step": 1849
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.201398637269966e-05,
      "loss": 0.2646,
      "step": 1850
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.200622674161675e-05,
      "loss": 0.8428,
      "step": 1851
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1998465851682877e-05,
      "loss": 0.8047,
      "step": 1852
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1990703707767775e-05,
      "loss": 0.8118,
      "step": 1853
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1982940314741976e-05,
      "loss": 0.2865,
      "step": 1854
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1975175677476785e-05,
      "loss": 0.807,
      "step": 1855
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1967409800844301e-05,
      "loss": 0.8481,
      "step": 1856
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.195964268971739e-05,
      "loss": 0.8241,
      "step": 1857
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1951874348969697e-05,
      "loss": 0.7754,
      "step": 1858
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.194410478347564e-05,
      "loss": 0.2637,
      "step": 1859
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1936333998110394e-05,
      "loss": 0.7995,
      "step": 1860
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1928561997749922e-05,
      "loss": 0.8033,
      "step": 1861
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1920788787270927e-05,
      "loss": 0.8332,
      "step": 1862
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1913014371550881e-05,
      "loss": 0.8382,
      "step": 1863
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1905238755468018e-05,
      "loss": 0.7868,
      "step": 1864
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1897461943901311e-05,
      "loss": 0.8131,
      "step": 1865
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1889683941730498e-05,
      "loss": 0.2688,
      "step": 1866
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1881904753836051e-05,
      "loss": 0.7824,
      "step": 1867
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1874124385099197e-05,
      "loss": 0.7877,
      "step": 1868
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1866342840401892e-05,
      "loss": 0.8005,
      "step": 1869
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1858560124626845e-05,
      "loss": 0.7974,
      "step": 1870
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1850776242657488e-05,
      "loss": 0.8298,
      "step": 1871
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1842991199377988e-05,
      "loss": 0.8708,
      "step": 1872
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1835204999673242e-05,
      "loss": 0.8439,
      "step": 1873
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1827417648428872e-05,
      "loss": 0.7673,
      "step": 1874
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1819629150531218e-05,
      "loss": 0.8511,
      "step": 1875
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.181183951086735e-05,
      "loss": 0.8787,
      "step": 1876
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1804048734325046e-05,
      "loss": 0.8259,
      "step": 1877
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1796256825792802e-05,
      "loss": 0.8276,
      "step": 1878
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1788463790159816e-05,
      "loss": 0.8287,
      "step": 1879
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1780669632316005e-05,
      "loss": 0.821,
      "step": 1880
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.177287435715198e-05,
      "loss": 0.8593,
      "step": 1881
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1765077969559057e-05,
      "loss": 0.7704,
      "step": 1882
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1757280474429256e-05,
      "loss": 0.8083,
      "step": 1883
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.174948187665528e-05,
      "loss": 0.2916,
      "step": 1884
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1741682181130536e-05,
      "loss": 0.7982,
      "step": 1885
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1733881392749107e-05,
      "loss": 0.8204,
      "step": 1886
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1726079516405778e-05,
      "loss": 0.8145,
      "step": 1887
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1718276556995998e-05,
      "loss": 0.8445,
      "step": 1888
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.171047251941591e-05,
      "loss": 0.7965,
      "step": 1889
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1702667408562328e-05,
      "loss": 0.8175,
      "step": 1890
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.169486122933274e-05,
      "loss": 0.8186,
      "step": 1891
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1687053986625305e-05,
      "loss": 0.7967,
      "step": 1892
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1679245685338847e-05,
      "loss": 0.8543,
      "step": 1893
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1671436330372855e-05,
      "loss": 0.8805,
      "step": 1894
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1663625926627483e-05,
      "loss": 0.7207,
      "step": 1895
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1655814479003538e-05,
      "loss": 0.8271,
      "step": 1896
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1648001992402485e-05,
      "loss": 0.811,
      "step": 1897
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.164018847172644e-05,
      "loss": 0.8938,
      "step": 1898
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1632373921878168e-05,
      "loss": 0.7514,
      "step": 1899
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1624558347761076e-05,
      "loss": 0.8516,
      "step": 1900
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1616741754279223e-05,
      "loss": 0.8752,
      "step": 1901
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1608924146337304e-05,
      "loss": 0.2692,
      "step": 1902
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.160110552884064e-05,
      "loss": 0.7832,
      "step": 1903
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.15932859066952e-05,
      "loss": 0.8299,
      "step": 1904
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1585465284807575e-05,
      "loss": 0.7342,
      "step": 1905
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1577643668084987e-05,
      "loss": 0.8661,
      "step": 1906
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.156982106143528e-05,
      "loss": 0.8357,
      "step": 1907
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1561997469766921e-05,
      "loss": 0.8315,
      "step": 1908
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1554172897988996e-05,
      "loss": 0.804,
      "step": 1909
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1546347351011197e-05,
      "loss": 0.821,
      "step": 1910
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1538520833743845e-05,
      "loss": 0.7958,
      "step": 1911
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1530693351097852e-05,
      "loss": 0.8028,
      "step": 1912
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1522864907984748e-05,
      "loss": 0.8086,
      "step": 1913
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1515035509316659e-05,
      "loss": 0.8257,
      "step": 1914
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1507205160006314e-05,
      "loss": 0.7884,
      "step": 1915
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1499373864967038e-05,
      "loss": 0.8006,
      "step": 1916
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1491541629112746e-05,
      "loss": 0.8391,
      "step": 1917
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1483708457357952e-05,
      "loss": 0.8448,
      "step": 1918
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1475874354617742e-05,
      "loss": 0.8078,
      "step": 1919
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1468039325807802e-05,
      "loss": 0.8479,
      "step": 1920
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1460203375844396e-05,
      "loss": 0.7904,
      "step": 1921
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1452366509644355e-05,
      "loss": 0.8211,
      "step": 1922
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1444528732125098e-05,
      "loss": 0.826,
      "step": 1923
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1436690048204605e-05,
      "loss": 0.7664,
      "step": 1924
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1428850462801438e-05,
      "loss": 0.8531,
      "step": 1925
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.142100998083471e-05,
      "loss": 0.8183,
      "step": 1926
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1413168607224103e-05,
      "loss": 0.8327,
      "step": 1927
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1405326346889867e-05,
      "loss": 0.2562,
      "step": 1928
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.139748320475279e-05,
      "loss": 0.7765,
      "step": 1929
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1389639185734232e-05,
      "loss": 0.7828,
      "step": 1930
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1381794294756089e-05,
      "loss": 0.8594,
      "step": 1931
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.137394853674081e-05,
      "loss": 0.8331,
      "step": 1932
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1366101916611392e-05,
      "loss": 0.7946,
      "step": 1933
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1358254439291368e-05,
      "loss": 0.8211,
      "step": 1934
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1350406109704806e-05,
      "loss": 0.8067,
      "step": 1935
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1342556932776316e-05,
      "loss": 0.7908,
      "step": 1936
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.133470691343104e-05,
      "loss": 0.7946,
      "step": 1937
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1326856056594633e-05,
      "loss": 0.8324,
      "step": 1938
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1319004367193296e-05,
      "loss": 0.8447,
      "step": 1939
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1311151850153744e-05,
      "loss": 0.8448,
      "step": 1940
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1303298510403206e-05,
      "loss": 0.824,
      "step": 1941
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1295444352869437e-05,
      "loss": 0.8325,
      "step": 1942
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1287589382480699e-05,
      "loss": 0.8025,
      "step": 1943
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1279733604165765e-05,
      "loss": 0.7784,
      "step": 1944
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1271877022853915e-05,
      "loss": 0.8415,
      "step": 1945
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1264019643474932e-05,
      "loss": 0.792,
      "step": 1946
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1256161470959107e-05,
      "loss": 0.7756,
      "step": 1947
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1248302510237215e-05,
      "loss": 0.896,
      "step": 1948
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.124044276624054e-05,
      "loss": 0.853,
      "step": 1949
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1232582243900842e-05,
      "loss": 0.8334,
      "step": 1950
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.122472094815039e-05,
      "loss": 0.8166,
      "step": 1951
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1216858883921914e-05,
      "loss": 0.851,
      "step": 1952
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1208996056148646e-05,
      "loss": 0.8233,
      "step": 1953
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1201132469764292e-05,
      "loss": 0.7979,
      "step": 1954
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1193268129703025e-05,
      "loss": 0.8275,
      "step": 1955
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1185403040899502e-05,
      "loss": 0.7905,
      "step": 1956
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1177537208288845e-05,
      "loss": 0.8284,
      "step": 1957
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1169670636806643e-05,
      "loss": 0.8818,
      "step": 1958
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1161803331388942e-05,
      "loss": 0.8376,
      "step": 1959
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.115393529697227e-05,
      "loss": 0.2833,
      "step": 1960
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1146066538493585e-05,
      "loss": 0.846,
      "step": 1961
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1138197060890315e-05,
      "loss": 0.8326,
      "step": 1962
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1130326869100338e-05,
      "loss": 0.8646,
      "step": 1963
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1122455968061978e-05,
      "loss": 0.8993,
      "step": 1964
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1114584362714006e-05,
      "loss": 0.7768,
      "step": 1965
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1106712057995632e-05,
      "loss": 0.8155,
      "step": 1966
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1098839058846504e-05,
      "loss": 0.8528,
      "step": 1967
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1090965370206712e-05,
      "loss": 0.7985,
      "step": 1968
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1083090997016771e-05,
      "loss": 0.7848,
      "step": 1969
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1075215944217635e-05,
      "loss": 0.2534,
      "step": 1970
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1067340216750669e-05,
      "loss": 0.8003,
      "step": 1971
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1059463819557679e-05,
      "loss": 0.8405,
      "step": 1972
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.105158675758088e-05,
      "loss": 0.8447,
      "step": 1973
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1043709035762909e-05,
      "loss": 0.8467,
      "step": 1974
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1035830659046813e-05,
      "loss": 0.745,
      "step": 1975
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1027951632376053e-05,
      "loss": 0.8615,
      "step": 1976
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1020071960694498e-05,
      "loss": 0.8139,
      "step": 1977
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.101219164894642e-05,
      "loss": 0.8652,
      "step": 1978
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1004310702076495e-05,
      "loss": 0.8075,
      "step": 1979
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0996429125029793e-05,
      "loss": 0.8375,
      "step": 1980
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0988546922751785e-05,
      "loss": 0.8215,
      "step": 1981
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0980664100188332e-05,
      "loss": 0.8097,
      "step": 1982
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0972780662285683e-05,
      "loss": 0.865,
      "step": 1983
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.096489661399047e-05,
      "loss": 0.8238,
      "step": 1984
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.095701196024972e-05,
      "loss": 0.7598,
      "step": 1985
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0949126706010823e-05,
      "loss": 0.8102,
      "step": 1986
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0941240856221563e-05,
      "loss": 0.8292,
      "step": 1987
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0933354415830082e-05,
      "loss": 0.8218,
      "step": 1988
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0925467389784905e-05,
      "loss": 0.7969,
      "step": 1989
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0917579783034917e-05,
      "loss": 0.8225,
      "step": 1990
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0909691600529374e-05,
      "loss": 0.8495,
      "step": 1991
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0901802847217883e-05,
      "loss": 0.7989,
      "step": 1992
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0893913528050419e-05,
      "loss": 0.2718,
      "step": 1993
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0886023647977311e-05,
      "loss": 0.7605,
      "step": 1994
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.087813321194923e-05,
      "loss": 0.8461,
      "step": 1995
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0870242224917207e-05,
      "loss": 0.8015,
      "step": 1996
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0862350691832614e-05,
      "loss": 0.8076,
      "step": 1997
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.085445861764717e-05,
      "loss": 0.8034,
      "step": 1998
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.084656600731293e-05,
      "loss": 0.7723,
      "step": 1999
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0838672865782276e-05,
      "loss": 0.7687,
      "step": 2000
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0830779198007943e-05,
      "loss": 0.2921,
      "step": 2001
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0822885008942982e-05,
      "loss": 0.7955,
      "step": 2002
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0814990303540777e-05,
      "loss": 0.8152,
      "step": 2003
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0807095086755032e-05,
      "loss": 0.8553,
      "step": 2004
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0799199363539777e-05,
      "loss": 0.8583,
      "step": 2005
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0791303138849352e-05,
      "loss": 0.8163,
      "step": 2006
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0783406417638418e-05,
      "loss": 0.7651,
      "step": 2007
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0775509204861947e-05,
      "loss": 0.8287,
      "step": 2008
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0767611505475217e-05,
      "loss": 0.7752,
      "step": 2009
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0759713324433814e-05,
      "loss": 0.7991,
      "step": 2010
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0751814666693625e-05,
      "loss": 0.7844,
      "step": 2011
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0743915537210836e-05,
      "loss": 0.8641,
      "step": 2012
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0736015940941928e-05,
      "loss": 0.8473,
      "step": 2013
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0728115882843675e-05,
      "loss": 0.8148,
      "step": 2014
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0720215367873145e-05,
      "loss": 0.7678,
      "step": 2015
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0712314400987687e-05,
      "loss": 0.8295,
      "step": 2016
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0704412987144936e-05,
      "loss": 0.8229,
      "step": 2017
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0696511131302809e-05,
      "loss": 0.8063,
      "step": 2018
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0688608838419496e-05,
      "loss": 0.8602,
      "step": 2019
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0680706113453469e-05,
      "loss": 0.7971,
      "step": 2020
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.067280296136346e-05,
      "loss": 0.2733,
      "step": 2021
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0664899387108478e-05,
      "loss": 0.789,
      "step": 2022
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0656995395647793e-05,
      "loss": 0.822,
      "step": 2023
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0649090991940939e-05,
      "loss": 0.8089,
      "step": 2024
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0641186180947709e-05,
      "loss": 0.8155,
      "step": 2025
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0633280967628146e-05,
      "loss": 0.8223,
      "step": 2026
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.062537535694255e-05,
      "loss": 0.8055,
      "step": 2027
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0617469353851478e-05,
      "loss": 0.8379,
      "step": 2028
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.060956296331571e-05,
      "loss": 0.7989,
      "step": 2029
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0601656190296299e-05,
      "loss": 0.8015,
      "step": 2030
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0593749039754513e-05,
      "loss": 0.7661,
      "step": 2031
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0585841516651871e-05,
      "loss": 0.7438,
      "step": 2032
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.057793362595012e-05,
      "loss": 0.7513,
      "step": 2033
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0570025372611246e-05,
      "loss": 0.8263,
      "step": 2034
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0562116761597446e-05,
      "loss": 0.7644,
      "step": 2035
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0554207797871157e-05,
      "loss": 0.816,
      "step": 2036
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0546298486395033e-05,
      "loss": 0.829,
      "step": 2037
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.053838883213194e-05,
      "loss": 0.831,
      "step": 2038
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0530478840044971e-05,
      "loss": 0.7986,
      "step": 2039
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0522568515097416e-05,
      "loss": 0.7754,
      "step": 2040
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0514657862252788e-05,
      "loss": 0.769,
      "step": 2041
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0506746886474797e-05,
      "loss": 0.7885,
      "step": 2042
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0498835592727357e-05,
      "loss": 0.8849,
      "step": 2043
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0490923985974588e-05,
      "loss": 0.8145,
      "step": 2044
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0483012071180795e-05,
      "loss": 0.813,
      "step": 2045
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0475099853310486e-05,
      "loss": 0.8137,
      "step": 2046
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0467187337328355e-05,
      "loss": 0.7708,
      "step": 2047
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0459274528199284e-05,
      "loss": 0.2609,
      "step": 2048
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0451361430888336e-05,
      "loss": 0.8904,
      "step": 2049
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0443448050360763e-05,
      "loss": 0.8379,
      "step": 2050
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0435534391581983e-05,
      "loss": 0.7844,
      "step": 2051
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0427620459517598e-05,
      "loss": 0.8314,
      "step": 2052
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0419706259133378e-05,
      "loss": 0.23,
      "step": 2053
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0411791795395259e-05,
      "loss": 0.7997,
      "step": 2054
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0403877073269349e-05,
      "loss": 0.8433,
      "step": 2055
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0395962097721913e-05,
      "loss": 0.7242,
      "step": 2056
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0388046873719371e-05,
      "loss": 0.7448,
      "step": 2057
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0380131406228313e-05,
      "loss": 0.8302,
      "step": 2058
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0372215700215462e-05,
      "loss": 0.8143,
      "step": 2059
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0364299760647713e-05,
      "loss": 0.8115,
      "step": 2060
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0356383592492086e-05,
      "loss": 0.8178,
      "step": 2061
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0348467200715757e-05,
      "loss": 0.8505,
      "step": 2062
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0340550590286045e-05,
      "loss": 0.8136,
      "step": 2063
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0332633766170392e-05,
      "loss": 0.2781,
      "step": 2064
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.032471673333639e-05,
      "loss": 0.8096,
      "step": 2065
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0316799496751752e-05,
      "loss": 0.8225,
      "step": 2066
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0308882061384325e-05,
      "loss": 0.8057,
      "step": 2067
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.030096443220207e-05,
      "loss": 0.8367,
      "step": 2068
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0293046614173084e-05,
      "loss": 0.8465,
      "step": 2069
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0285128612265571e-05,
      "loss": 0.7817,
      "step": 2070
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.027721043144786e-05,
      "loss": 0.8187,
      "step": 2071
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0269292076688384e-05,
      "loss": 0.8553,
      "step": 2072
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0261373552955689e-05,
      "loss": 0.7781,
      "step": 2073
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0253454865218425e-05,
      "loss": 0.8583,
      "step": 2074
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0245536018445346e-05,
      "loss": 0.7903,
      "step": 2075
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0237617017605309e-05,
      "loss": 0.8709,
      "step": 2076
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0229697867667263e-05,
      "loss": 0.2604,
      "step": 2077
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0221778573600252e-05,
      "loss": 0.7926,
      "step": 2078
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0213859140373412e-05,
      "loss": 0.8032,
      "step": 2079
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0205939572955962e-05,
      "loss": 0.8299,
      "step": 2080
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0198019876317214e-05,
      "loss": 0.8368,
      "step": 2081
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.019010005542655e-05,
      "loss": 0.8141,
      "step": 2082
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0182180115253434e-05,
      "loss": 0.9044,
      "step": 2083
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0174260060767417e-05,
      "loss": 0.8334,
      "step": 2084
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0166339896938099e-05,
      "loss": 0.7771,
      "step": 2085
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0158419628735163e-05,
      "loss": 0.8433,
      "step": 2086
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0150499261128356e-05,
      "loss": 0.7758,
      "step": 2087
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.014257879908749e-05,
      "loss": 0.7963,
      "step": 2088
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.013465824758243e-05,
      "loss": 0.2381,
      "step": 2089
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0126737611583098e-05,
      "loss": 0.802,
      "step": 2090
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0118816896059474e-05,
      "loss": 0.7902,
      "step": 2091
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.011089610598158e-05,
      "loss": 0.8117,
      "step": 2092
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0102975246319495e-05,
      "loss": 0.2881,
      "step": 2093
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0095054322043331e-05,
      "loss": 0.8295,
      "step": 2094
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.008713333812325e-05,
      "loss": 0.8189,
      "step": 2095
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0079212299529445e-05,
      "loss": 0.8005,
      "step": 2096
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0071291211232143e-05,
      "loss": 0.7811,
      "step": 2097
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0063370078201611e-05,
      "loss": 0.8162,
      "step": 2098
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.005544890540813e-05,
      "loss": 0.8477,
      "step": 2099
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0047527697822017e-05,
      "loss": 0.8058,
      "step": 2100
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0039606460413611e-05,
      "loss": 0.7899,
      "step": 2101
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0031685198153261e-05,
      "loss": 0.8408,
      "step": 2102
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0023763916011337e-05,
      "loss": 0.781,
      "step": 2103
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0015842618958225e-05,
      "loss": 0.7916,
      "step": 2104
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0007921311964315e-05,
      "loss": 0.7914,
      "step": 2105
    },
    {
      "epoch": 0.52,
      "learning_rate": 1e-05,
      "loss": 0.8465,
      "step": 2106
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.992078688035688e-06,
      "loss": 0.8094,
      "step": 2107
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.984157381041779e-06,
      "loss": 0.8338,
      "step": 2108
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.976236083988664e-06,
      "loss": 0.8761,
      "step": 2109
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.968314801846744e-06,
      "loss": 0.7937,
      "step": 2110
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.960393539586392e-06,
      "loss": 0.2751,
      "step": 2111
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.952472302177986e-06,
      "loss": 0.8141,
      "step": 2112
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.944551094591873e-06,
      "loss": 0.8318,
      "step": 2113
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.936629921798392e-06,
      "loss": 0.8203,
      "step": 2114
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.928708788767859e-06,
      "loss": 0.8326,
      "step": 2115
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.920787700470558e-06,
      "loss": 0.8461,
      "step": 2116
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.912866661876752e-06,
      "loss": 0.8702,
      "step": 2117
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.90494567795667e-06,
      "loss": 0.8105,
      "step": 2118
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.897024753680509e-06,
      "loss": 0.8027,
      "step": 2119
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.88910389401842e-06,
      "loss": 0.8146,
      "step": 2120
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.881183103940527e-06,
      "loss": 0.8705,
      "step": 2121
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.873262388416904e-06,
      "loss": 0.8483,
      "step": 2122
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.86534175241757e-06,
      "loss": 0.8634,
      "step": 2123
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.85742120091251e-06,
      "loss": 0.7724,
      "step": 2124
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.849500738871642e-06,
      "loss": 0.7221,
      "step": 2125
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.84158037126484e-06,
      "loss": 0.8511,
      "step": 2126
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.833660103061903e-06,
      "loss": 0.8432,
      "step": 2127
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.825739939232585e-06,
      "loss": 0.7861,
      "step": 2128
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.817819884746567e-06,
      "loss": 0.8153,
      "step": 2129
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.809899944573456e-06,
      "loss": 0.8743,
      "step": 2130
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.801980123682791e-06,
      "loss": 0.8025,
      "step": 2131
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.794060427044043e-06,
      "loss": 0.8588,
      "step": 2132
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.786140859626593e-06,
      "loss": 0.7444,
      "step": 2133
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.778221426399753e-06,
      "loss": 0.7671,
      "step": 2134
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.770302132332742e-06,
      "loss": 0.819,
      "step": 2135
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.762382982394696e-06,
      "loss": 0.7884,
      "step": 2136
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.754463981554658e-06,
      "loss": 0.8921,
      "step": 2137
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.74654513478158e-06,
      "loss": 0.865,
      "step": 2138
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.738626447044318e-06,
      "loss": 0.7982,
      "step": 2139
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.730707923311619e-06,
      "loss": 0.7789,
      "step": 2140
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.722789568552144e-06,
      "loss": 0.8545,
      "step": 2141
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.71487138773443e-06,
      "loss": 0.7881,
      "step": 2142
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.706953385826921e-06,
      "loss": 0.837,
      "step": 2143
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.699035567797934e-06,
      "loss": 0.8096,
      "step": 2144
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.691117938615679e-06,
      "loss": 0.8088,
      "step": 2145
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.68320050324825e-06,
      "loss": 0.7668,
      "step": 2146
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.675283266663611e-06,
      "loss": 0.8426,
      "step": 2147
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.66736623382961e-06,
      "loss": 0.8032,
      "step": 2148
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.659449409713958e-06,
      "loss": 0.8728,
      "step": 2149
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.651532799284246e-06,
      "loss": 0.7999,
      "step": 2150
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.643616407507918e-06,
      "loss": 0.3056,
      "step": 2151
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.63570023935229e-06,
      "loss": 0.7997,
      "step": 2152
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.627784299784539e-06,
      "loss": 0.7667,
      "step": 2153
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.61986859377169e-06,
      "loss": 0.8484,
      "step": 2154
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.61195312628063e-06,
      "loss": 0.8208,
      "step": 2155
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.60403790227809e-06,
      "loss": 0.7543,
      "step": 2156
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.596122926730655e-06,
      "loss": 0.7926,
      "step": 2157
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.588208204604743e-06,
      "loss": 0.816,
      "step": 2158
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.580293740866625e-06,
      "loss": 0.8188,
      "step": 2159
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.572379540482406e-06,
      "loss": 0.7891,
      "step": 2160
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.564465608418019e-06,
      "loss": 0.8167,
      "step": 2161
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.55655194963924e-06,
      "loss": 0.8016,
      "step": 2162
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.548638569111665e-06,
      "loss": 0.7973,
      "step": 2163
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.540725471800718e-06,
      "loss": 0.8003,
      "step": 2164
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.532812662671646e-06,
      "loss": 0.8149,
      "step": 2165
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.524900146689512e-06,
      "loss": 0.7322,
      "step": 2166
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.516987928819206e-06,
      "loss": 0.8126,
      "step": 2167
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.509076014025412e-06,
      "loss": 0.2841,
      "step": 2168
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.501164407272643e-06,
      "loss": 0.835,
      "step": 2169
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.493253113525204e-06,
      "loss": 0.2624,
      "step": 2170
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.485342137747214e-06,
      "loss": 0.8313,
      "step": 2171
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.477431484902584e-06,
      "loss": 0.7531,
      "step": 2172
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.46952115995503e-06,
      "loss": 0.8134,
      "step": 2173
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.461611167868062e-06,
      "loss": 0.7651,
      "step": 2174
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.453701513604972e-06,
      "loss": 0.8596,
      "step": 2175
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.445792202128846e-06,
      "loss": 0.8323,
      "step": 2176
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.43788323840256e-06,
      "loss": 0.8139,
      "step": 2177
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.42997462738876e-06,
      "loss": 0.847,
      "step": 2178
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.422066374049883e-06,
      "loss": 0.8074,
      "step": 2179
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.414158483348134e-06,
      "loss": 0.8295,
      "step": 2180
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.406250960245494e-06,
      "loss": 0.7884,
      "step": 2181
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.398343809703706e-06,
      "loss": 0.8113,
      "step": 2182
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.39043703668429e-06,
      "loss": 0.9113,
      "step": 2183
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.382530646148529e-06,
      "loss": 0.8001,
      "step": 2184
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.374624643057451e-06,
      "loss": 0.8456,
      "step": 2185
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.366719032371858e-06,
      "loss": 0.7798,
      "step": 2186
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.358813819052294e-06,
      "loss": 0.7689,
      "step": 2187
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.350909008059065e-06,
      "loss": 0.8435,
      "step": 2188
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.343004604352209e-06,
      "loss": 0.7733,
      "step": 2189
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.335100612891524e-06,
      "loss": 0.7454,
      "step": 2190
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.327197038636543e-06,
      "loss": 0.8431,
      "step": 2191
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.319293886546534e-06,
      "loss": 0.299,
      "step": 2192
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.311391161580507e-06,
      "loss": 0.8345,
      "step": 2193
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.303488868697194e-06,
      "loss": 0.8166,
      "step": 2194
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.295587012855066e-06,
      "loss": 0.7875,
      "step": 2195
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.287685599012315e-06,
      "loss": 0.8193,
      "step": 2196
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.279784632126857e-06,
      "loss": 0.235,
      "step": 2197
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.271884117156327e-06,
      "loss": 0.2887,
      "step": 2198
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.263984059058074e-06,
      "loss": 0.8439,
      "step": 2199
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.256084462789167e-06,
      "loss": 0.2775,
      "step": 2200
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.248185333306377e-06,
      "loss": 0.9058,
      "step": 2201
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.24028667556619e-06,
      "loss": 0.3031,
      "step": 2202
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.232388494524786e-06,
      "loss": 0.877,
      "step": 2203
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.224490795138055e-06,
      "loss": 0.8021,
      "step": 2204
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.216593582361585e-06,
      "loss": 0.8361,
      "step": 2205
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.208696861150652e-06,
      "loss": 0.812,
      "step": 2206
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.200800636460226e-06,
      "loss": 0.7917,
      "step": 2207
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.192904913244968e-06,
      "loss": 0.8131,
      "step": 2208
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.185009696459225e-06,
      "loss": 0.8613,
      "step": 2209
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.177114991057018e-06,
      "loss": 0.7893,
      "step": 2210
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.169220801992057e-06,
      "loss": 0.8168,
      "step": 2211
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.161327134217724e-06,
      "loss": 0.7967,
      "step": 2212
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.153433992687072e-06,
      "loss": 0.8056,
      "step": 2213
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.14554138235283e-06,
      "loss": 0.7877,
      "step": 2214
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.137649308167384e-06,
      "loss": 0.7774,
      "step": 2215
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.129757775082795e-06,
      "loss": 0.743,
      "step": 2216
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.121866788050772e-06,
      "loss": 0.8692,
      "step": 2217
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.11397635202269e-06,
      "loss": 0.7844,
      "step": 2218
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.106086471949583e-06,
      "loss": 0.8284,
      "step": 2219
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.098197152782122e-06,
      "loss": 0.7747,
      "step": 2220
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.090308399470632e-06,
      "loss": 0.8056,
      "step": 2221
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.082420216965087e-06,
      "loss": 0.7841,
      "step": 2222
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.074532610215099e-06,
      "loss": 0.7826,
      "step": 2223
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.066645584169925e-06,
      "loss": 0.8595,
      "step": 2224
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.058759143778444e-06,
      "loss": 0.8774,
      "step": 2225
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.050873293989182e-06,
      "loss": 0.8066,
      "step": 2226
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.042988039750287e-06,
      "loss": 0.7789,
      "step": 2227
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.035103386009532e-06,
      "loss": 0.8076,
      "step": 2228
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.027219337714324e-06,
      "loss": 0.8785,
      "step": 2229
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.019335899811671e-06,
      "loss": 0.7969,
      "step": 2230
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.011453077248217e-06,
      "loss": 0.8452,
      "step": 2231
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.003570874970209e-06,
      "loss": 0.844,
      "step": 2232
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.99568929792351e-06,
      "loss": 0.8033,
      "step": 2233
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.987808351053582e-06,
      "loss": 0.7796,
      "step": 2234
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.979928039305504e-06,
      "loss": 0.7474,
      "step": 2235
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.97204836762395e-06,
      "loss": 0.7284,
      "step": 2236
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.96416934095319e-06,
      "loss": 0.7889,
      "step": 2237
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.956290964237095e-06,
      "loss": 0.7812,
      "step": 2238
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.948413242419123e-06,
      "loss": 0.7613,
      "step": 2239
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.940536180442325e-06,
      "loss": 0.7903,
      "step": 2240
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.932659783249333e-06,
      "loss": 0.754,
      "step": 2241
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.924784055782369e-06,
      "loss": 0.788,
      "step": 2242
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.916909002983232e-06,
      "loss": 0.8361,
      "step": 2243
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.909034629793291e-06,
      "loss": 0.7728,
      "step": 2244
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.9011609411535e-06,
      "loss": 0.7856,
      "step": 2245
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.893287942004371e-06,
      "loss": 0.8404,
      "step": 2246
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.885415637285998e-06,
      "loss": 0.7515,
      "step": 2247
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.877544031938024e-06,
      "loss": 0.7918,
      "step": 2248
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.869673130899664e-06,
      "loss": 0.7715,
      "step": 2249
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.861802939109688e-06,
      "loss": 0.8536,
      "step": 2250
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.853933461506418e-06,
      "loss": 0.7707,
      "step": 2251
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.846064703027734e-06,
      "loss": 0.7994,
      "step": 2252
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.838196668611058e-06,
      "loss": 0.8377,
      "step": 2253
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.830329363193362e-06,
      "loss": 0.7502,
      "step": 2254
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.822462791711157e-06,
      "loss": 0.7758,
      "step": 2255
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.814596959100498e-06,
      "loss": 0.8026,
      "step": 2256
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.806731870296975e-06,
      "loss": 0.8696,
      "step": 2257
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.79886753023571e-06,
      "loss": 0.8614,
      "step": 2258
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.791003943851353e-06,
      "loss": 0.2629,
      "step": 2259
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.783141116078086e-06,
      "loss": 0.7824,
      "step": 2260
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.775279051849614e-06,
      "loss": 0.2656,
      "step": 2261
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.767417756099158e-06,
      "loss": 0.7843,
      "step": 2262
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.759557233759461e-06,
      "loss": 0.8098,
      "step": 2263
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.751697489762788e-06,
      "loss": 0.8294,
      "step": 2264
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.743838529040898e-06,
      "loss": 0.2677,
      "step": 2265
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.73598035652507e-06,
      "loss": 0.7962,
      "step": 2266
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.728122977146089e-06,
      "loss": 0.8121,
      "step": 2267
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.72026639583424e-06,
      "loss": 0.8232,
      "step": 2268
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.712410617519306e-06,
      "loss": 0.8361,
      "step": 2269
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.704555647130568e-06,
      "loss": 0.2653,
      "step": 2270
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.696701489596797e-06,
      "loss": 0.8704,
      "step": 2271
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.688848149846261e-06,
      "loss": 0.8534,
      "step": 2272
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.680995632806707e-06,
      "loss": 0.7854,
      "step": 2273
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.673143943405372e-06,
      "loss": 0.8364,
      "step": 2274
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.665293086568967e-06,
      "loss": 0.8077,
      "step": 2275
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.657443067223688e-06,
      "loss": 0.7771,
      "step": 2276
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.649593890295196e-06,
      "loss": 0.799,
      "step": 2277
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.641745560708637e-06,
      "loss": 0.8352,
      "step": 2278
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.63389808338861e-06,
      "loss": 0.8459,
      "step": 2279
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.626051463259191e-06,
      "loss": 0.7587,
      "step": 2280
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.618205705243916e-06,
      "loss": 0.8336,
      "step": 2281
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.610360814265771e-06,
      "loss": 0.8017,
      "step": 2282
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.602516795247213e-06,
      "loss": 0.7797,
      "step": 2283
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.594673653110136e-06,
      "loss": 0.8221,
      "step": 2284
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.5868313927759e-06,
      "loss": 0.2694,
      "step": 2285
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.578990019165294e-06,
      "loss": 0.8104,
      "step": 2286
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.571149537198565e-06,
      "loss": 0.8873,
      "step": 2287
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.563309951795398e-06,
      "loss": 0.7851,
      "step": 2288
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.555471267874905e-06,
      "loss": 0.8599,
      "step": 2289
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.547633490355648e-06,
      "loss": 0.7942,
      "step": 2290
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.539796624155607e-06,
      "loss": 0.8296,
      "step": 2291
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.5319606741922e-06,
      "loss": 0.8168,
      "step": 2292
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.52412564538226e-06,
      "loss": 0.8167,
      "step": 2293
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.516291542642053e-06,
      "loss": 0.2704,
      "step": 2294
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.508458370887255e-06,
      "loss": 0.7766,
      "step": 2295
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.500626135032965e-06,
      "loss": 0.7933,
      "step": 2296
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.492794839993688e-06,
      "loss": 0.2596,
      "step": 2297
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.484964490683343e-06,
      "loss": 0.8822,
      "step": 2298
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.477135092015255e-06,
      "loss": 0.7909,
      "step": 2299
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.46930664890215e-06,
      "loss": 0.2426,
      "step": 2300
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.461479166256157e-06,
      "loss": 0.8119,
      "step": 2301
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.453652648988803e-06,
      "loss": 0.7669,
      "step": 2302
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.445827102011005e-06,
      "loss": 0.8568,
      "step": 2303
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.438002530233079e-06,
      "loss": 0.884,
      "step": 2304
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.43017893856472e-06,
      "loss": 0.8202,
      "step": 2305
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.422356331915014e-06,
      "loss": 0.8043,
      "step": 2306
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.414534715192425e-06,
      "loss": 0.7978,
      "step": 2307
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.4067140933048e-06,
      "loss": 0.8119,
      "step": 2308
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.398894471159362e-06,
      "loss": 0.8309,
      "step": 2309
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.391075853662701e-06,
      "loss": 0.7494,
      "step": 2310
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.383258245720778e-06,
      "loss": 0.7441,
      "step": 2311
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.375441652238927e-06,
      "loss": 0.8009,
      "step": 2312
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.367626078121837e-06,
      "loss": 0.7853,
      "step": 2313
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.359811528273567e-06,
      "loss": 0.8428,
      "step": 2314
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.35199800759752e-06,
      "loss": 0.797,
      "step": 2315
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.344185520996467e-06,
      "loss": 0.8515,
      "step": 2316
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.336374073372522e-06,
      "loss": 0.8502,
      "step": 2317
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.32856366962715e-06,
      "loss": 0.8288,
      "step": 2318
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.320754314661158e-06,
      "loss": 0.8001,
      "step": 2319
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.3129460133747e-06,
      "loss": 0.2558,
      "step": 2320
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.305138770667266e-06,
      "loss": 0.6847,
      "step": 2321
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.297332591437674e-06,
      "loss": 0.7737,
      "step": 2322
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.289527480584095e-06,
      "loss": 0.803,
      "step": 2323
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.281723443004007e-06,
      "loss": 0.7607,
      "step": 2324
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.273920483594225e-06,
      "loss": 0.7886,
      "step": 2325
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.266118607250894e-06,
      "loss": 0.7998,
      "step": 2326
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.258317818869468e-06,
      "loss": 0.7907,
      "step": 2327
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.250518123344724e-06,
      "loss": 0.7502,
      "step": 2328
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.242719525570747e-06,
      "loss": 0.8025,
      "step": 2329
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.234922030440946e-06,
      "loss": 0.7266,
      "step": 2330
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.227125642848023e-06,
      "loss": 0.2906,
      "step": 2331
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.219330367683999e-06,
      "loss": 0.8342,
      "step": 2332
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.211536209840188e-06,
      "loss": 0.8013,
      "step": 2333
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.203743174207201e-06,
      "loss": 0.8019,
      "step": 2334
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.195951265674955e-06,
      "loss": 0.7967,
      "step": 2335
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.188160489132651e-06,
      "loss": 0.8392,
      "step": 2336
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.180370849468784e-06,
      "loss": 0.8098,
      "step": 2337
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.172582351571133e-06,
      "loss": 0.8256,
      "step": 2338
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.16479500032676e-06,
      "loss": 0.8117,
      "step": 2339
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.157008800622015e-06,
      "loss": 0.8292,
      "step": 2340
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.149223757342513e-06,
      "loss": 0.7891,
      "step": 2341
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.141439875373157e-06,
      "loss": 0.7897,
      "step": 2342
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.133657159598108e-06,
      "loss": 0.8431,
      "step": 2343
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.125875614900806e-06,
      "loss": 0.7881,
      "step": 2344
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.11809524616395e-06,
      "loss": 0.7696,
      "step": 2345
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.110316058269504e-06,
      "loss": 0.8818,
      "step": 2346
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.102538056098689e-06,
      "loss": 0.7956,
      "step": 2347
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.094761244531982e-06,
      "loss": 0.8006,
      "step": 2348
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.086985628449119e-06,
      "loss": 0.7374,
      "step": 2349
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.079211212729074e-06,
      "loss": 0.8347,
      "step": 2350
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.07143800225008e-06,
      "loss": 0.7908,
      "step": 2351
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.063666001889606e-06,
      "loss": 0.8313,
      "step": 2352
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.055895216524362e-06,
      "loss": 0.7926,
      "step": 2353
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.048125651030308e-06,
      "loss": 0.7619,
      "step": 2354
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.040357310282615e-06,
      "loss": 0.8597,
      "step": 2355
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.032590199155704e-06,
      "loss": 0.8041,
      "step": 2356
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.02482432252322e-06,
      "loss": 0.7794,
      "step": 2357
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.01705968525803e-06,
      "loss": 0.8413,
      "step": 2358
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.00929629223223e-06,
      "loss": 0.8158,
      "step": 2359
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.001534148317128e-06,
      "loss": 0.8646,
      "step": 2360
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.993773258383251e-06,
      "loss": 0.8488,
      "step": 2361
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.986013627300346e-06,
      "loss": 0.8376,
      "step": 2362
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.978255259937355e-06,
      "loss": 0.8022,
      "step": 2363
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.97049816116244e-06,
      "loss": 0.8082,
      "step": 2364
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.962742335842964e-06,
      "loss": 0.7937,
      "step": 2365
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.954987788845487e-06,
      "loss": 0.815,
      "step": 2366
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.947234525035772e-06,
      "loss": 0.787,
      "step": 2367
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.939482549278772e-06,
      "loss": 0.8004,
      "step": 2368
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.931731866438633e-06,
      "loss": 0.8047,
      "step": 2369
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.923982481378695e-06,
      "loss": 0.7786,
      "step": 2370
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.916234398961478e-06,
      "loss": 0.7664,
      "step": 2371
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.908487624048686e-06,
      "loss": 0.7866,
      "step": 2372
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.900742161501204e-06,
      "loss": 0.7569,
      "step": 2373
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.892998016179088e-06,
      "loss": 0.793,
      "step": 2374
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.885255192941579e-06,
      "loss": 0.7887,
      "step": 2375
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.877513696647073e-06,
      "loss": 0.7732,
      "step": 2376
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.869773532153148e-06,
      "loss": 0.8057,
      "step": 2377
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.862034704316539e-06,
      "loss": 0.7636,
      "step": 2378
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.854297217993138e-06,
      "loss": 0.8677,
      "step": 2379
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.84656107803801e-06,
      "loss": 0.8276,
      "step": 2380
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.838826289305356e-06,
      "loss": 0.833,
      "step": 2381
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.831092856648545e-06,
      "loss": 0.8607,
      "step": 2382
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.823360784920087e-06,
      "loss": 0.8321,
      "step": 2383
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.815630078971639e-06,
      "loss": 0.7795,
      "step": 2384
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.807900743654003e-06,
      "loss": 0.8244,
      "step": 2385
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.800172783817119e-06,
      "loss": 0.7929,
      "step": 2386
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.792446204310069e-06,
      "loss": 0.807,
      "step": 2387
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.784721009981058e-06,
      "loss": 0.2694,
      "step": 2388
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.776997205677435e-06,
      "loss": 0.77,
      "step": 2389
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.769274796245664e-06,
      "loss": 0.2764,
      "step": 2390
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.761553786531346e-06,
      "loss": 0.2509,
      "step": 2391
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.753834181379195e-06,
      "loss": 0.8305,
      "step": 2392
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.746115985633046e-06,
      "loss": 0.7433,
      "step": 2393
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.738399204135854e-06,
      "loss": 0.7998,
      "step": 2394
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.730683841729678e-06,
      "loss": 0.855,
      "step": 2395
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.722969903255695e-06,
      "loss": 0.8121,
      "step": 2396
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.71525739355418e-06,
      "loss": 0.8316,
      "step": 2397
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.707546317464523e-06,
      "loss": 0.8052,
      "step": 2398
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.699836679825205e-06,
      "loss": 0.7911,
      "step": 2399
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.692128485473807e-06,
      "loss": 0.7212,
      "step": 2400
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.684421739247001e-06,
      "loss": 0.7943,
      "step": 2401
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.67671644598056e-06,
      "loss": 0.7668,
      "step": 2402
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.669012610509334e-06,
      "loss": 0.7388,
      "step": 2403
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.661310237667268e-06,
      "loss": 0.8458,
      "step": 2404
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.65360933228738e-06,
      "loss": 0.7933,
      "step": 2405
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.645909899201775e-06,
      "loss": 0.8107,
      "step": 2406
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.638211943241632e-06,
      "loss": 0.7974,
      "step": 2407
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.630515469237202e-06,
      "loss": 0.7897,
      "step": 2408
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.622820482017804e-06,
      "loss": 0.8152,
      "step": 2409
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.615126986411827e-06,
      "loss": 0.7901,
      "step": 2410
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.607434987246728e-06,
      "loss": 0.887,
      "step": 2411
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.599744489349013e-06,
      "loss": 0.8618,
      "step": 2412
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.592055497544261e-06,
      "loss": 0.2533,
      "step": 2413
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.584368016657093e-06,
      "loss": 0.7661,
      "step": 2414
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.57668205151119e-06,
      "loss": 0.7624,
      "step": 2415
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.5689976069292806e-06,
      "loss": 0.7304,
      "step": 2416
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.5613146877331345e-06,
      "loss": 0.7729,
      "step": 2417
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.553633298743572e-06,
      "loss": 0.787,
      "step": 2418
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.545953444780445e-06,
      "loss": 0.8515,
      "step": 2419
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.53827513066265e-06,
      "loss": 0.828,
      "step": 2420
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.53059836120811e-06,
      "loss": 0.2859,
      "step": 2421
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.522923141233784e-06,
      "loss": 0.7675,
      "step": 2422
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.51524947555566e-06,
      "loss": 0.8016,
      "step": 2423
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.507577368988743e-06,
      "loss": 0.2622,
      "step": 2424
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.4999068263470675e-06,
      "loss": 0.7307,
      "step": 2425
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.492237852443682e-06,
      "loss": 0.8342,
      "step": 2426
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.484570452090655e-06,
      "loss": 0.7163,
      "step": 2427
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.47690463009906e-06,
      "loss": 0.8321,
      "step": 2428
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.46924039127899e-06,
      "loss": 0.8322,
      "step": 2429
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.46157774043954e-06,
      "loss": 0.2276,
      "step": 2430
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.4539166823888034e-06,
      "loss": 0.7683,
      "step": 2431
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.446257221933883e-06,
      "loss": 0.8085,
      "step": 2432
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.438599363880874e-06,
      "loss": 0.8177,
      "step": 2433
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.43094311303487e-06,
      "loss": 0.2645,
      "step": 2434
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.423288474199948e-06,
      "loss": 0.813,
      "step": 2435
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.415635452179183e-06,
      "loss": 0.797,
      "step": 2436
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.407984051774633e-06,
      "loss": 0.7749,
      "step": 2437
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.400334277787331e-06,
      "loss": 0.8113,
      "step": 2438
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.392686135017302e-06,
      "loss": 0.2697,
      "step": 2439
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.385039628263537e-06,
      "loss": 0.7672,
      "step": 2440
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.377394762324006e-06,
      "loss": 0.8048,
      "step": 2441
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.369751541995647e-06,
      "loss": 0.8435,
      "step": 2442
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.362109972074364e-06,
      "loss": 0.8012,
      "step": 2443
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.354470057355033e-06,
      "loss": 0.7774,
      "step": 2444
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.346831802631486e-06,
      "loss": 0.8507,
      "step": 2445
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.339195212696506e-06,
      "loss": 0.7711,
      "step": 2446
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.331560292341845e-06,
      "loss": 0.7884,
      "step": 2447
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.323927046358197e-06,
      "loss": 0.7753,
      "step": 2448
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.3162954795352155e-06,
      "loss": 0.8088,
      "step": 2449
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.308665596661488e-06,
      "loss": 0.8727,
      "step": 2450
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.301037402524556e-06,
      "loss": 0.8083,
      "step": 2451
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.2934109019108975e-06,
      "loss": 0.7866,
      "step": 2452
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.285786099605924e-06,
      "loss": 0.8047,
      "step": 2453
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.278163000393989e-06,
      "loss": 0.7811,
      "step": 2454
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.2705416090583705e-06,
      "loss": 0.7496,
      "step": 2455
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.262921930381279e-06,
      "loss": 0.769,
      "step": 2456
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.2553039691438475e-06,
      "loss": 0.7593,
      "step": 2457
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.247687730126135e-06,
      "loss": 0.7723,
      "step": 2458
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.240073218107118e-06,
      "loss": 0.788,
      "step": 2459
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.2324604378646844e-06,
      "loss": 0.8563,
      "step": 2460
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.224849394175646e-06,
      "loss": 0.8205,
      "step": 2461
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.2172400918157156e-06,
      "loss": 0.2436,
      "step": 2462
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.209632535559519e-06,
      "loss": 0.8572,
      "step": 2463
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.20202673018058e-06,
      "loss": 0.7651,
      "step": 2464
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.194422680451334e-06,
      "loss": 0.8362,
      "step": 2465
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.186820391143103e-06,
      "loss": 0.7982,
      "step": 2466
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.179219867026112e-06,
      "loss": 0.8249,
      "step": 2467
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.171621112869482e-06,
      "loss": 0.8436,
      "step": 2468
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.16402413344121e-06,
      "loss": 0.8424,
      "step": 2469
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.156428933508192e-06,
      "loss": 0.7955,
      "step": 2470
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.1488355178362e-06,
      "loss": 0.8209,
      "step": 2471
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.1412438911898925e-06,
      "loss": 0.8231,
      "step": 2472
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.133654058332798e-06,
      "loss": 0.7783,
      "step": 2473
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.126066024027327e-06,
      "loss": 0.7742,
      "step": 2474
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.118479793034758e-06,
      "loss": 0.7892,
      "step": 2475
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.110895370115234e-06,
      "loss": 0.8034,
      "step": 2476
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.103312760027774e-06,
      "loss": 0.7709,
      "step": 2477
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.095731967530246e-06,
      "loss": 0.8025,
      "step": 2478
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.088152997379392e-06,
      "loss": 0.8329,
      "step": 2479
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.080575854330795e-06,
      "loss": 0.8019,
      "step": 2480
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.073000543138904e-06,
      "loss": 0.8267,
      "step": 2481
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.0654270685570145e-06,
      "loss": 0.2777,
      "step": 2482
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.057855435337267e-06,
      "loss": 0.789,
      "step": 2483
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.050285648230653e-06,
      "loss": 0.8134,
      "step": 2484
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.042717711986996e-06,
      "loss": 0.8165,
      "step": 2485
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.035151631354968e-06,
      "loss": 0.7772,
      "step": 2486
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.027587411082069e-06,
      "loss": 0.7961,
      "step": 2487
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.020025055914635e-06,
      "loss": 0.8269,
      "step": 2488
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.012464570597837e-06,
      "loss": 0.7911,
      "step": 2489
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.004905959875663e-06,
      "loss": 0.7668,
      "step": 2490
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.997349228490926e-06,
      "loss": 0.8251,
      "step": 2491
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.989794381185268e-06,
      "loss": 0.7792,
      "step": 2492
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9822414226991405e-06,
      "loss": 0.793,
      "step": 2493
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.974690357771816e-06,
      "loss": 0.7939,
      "step": 2494
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.967141191141371e-06,
      "loss": 0.7818,
      "step": 2495
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.959593927544699e-06,
      "loss": 0.2689,
      "step": 2496
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.952048571717498e-06,
      "loss": 0.7787,
      "step": 2497
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9445051283942635e-06,
      "loss": 0.8335,
      "step": 2498
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.936963602308297e-06,
      "loss": 0.7823,
      "step": 2499
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.92942399819169e-06,
      "loss": 0.8382,
      "step": 2500
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.921886320775338e-06,
      "loss": 0.7641,
      "step": 2501
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.914350574788917e-06,
      "loss": 0.8416,
      "step": 2502
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9068167649608995e-06,
      "loss": 0.7596,
      "step": 2503
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.899284896018541e-06,
      "loss": 0.834,
      "step": 2504
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.891754972687873e-06,
      "loss": 0.7542,
      "step": 2505
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.884226999693715e-06,
      "loss": 0.8183,
      "step": 2506
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.876700981759656e-06,
      "loss": 0.8302,
      "step": 2507
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.869176923608065e-06,
      "loss": 0.7945,
      "step": 2508
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.861654829960073e-06,
      "loss": 0.8019,
      "step": 2509
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.854134705535584e-06,
      "loss": 0.7915,
      "step": 2510
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.846616555053266e-06,
      "loss": 0.778,
      "step": 2511
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.839100383230546e-06,
      "loss": 0.8095,
      "step": 2512
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.831586194783613e-06,
      "loss": 0.8135,
      "step": 2513
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.824073994427406e-06,
      "loss": 0.8023,
      "step": 2514
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.816563786875624e-06,
      "loss": 0.7463,
      "step": 2515
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.809055576840706e-06,
      "loss": 0.8396,
      "step": 2516
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.801549369033851e-06,
      "loss": 0.805,
      "step": 2517
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.794045168164984e-06,
      "loss": 0.7842,
      "step": 2518
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.7865429789427875e-06,
      "loss": 0.7395,
      "step": 2519
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.779042806074674e-06,
      "loss": 0.7795,
      "step": 2520
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.7715446542667864e-06,
      "loss": 0.7431,
      "step": 2521
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.764048528224012e-06,
      "loss": 0.8068,
      "step": 2522
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.7565544326499535e-06,
      "loss": 0.7742,
      "step": 2523
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.749062372246948e-06,
      "loss": 0.8184,
      "step": 2524
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.741572351716051e-06,
      "loss": 0.7613,
      "step": 2525
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.734084375757043e-06,
      "loss": 0.8149,
      "step": 2526
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.726598449068419e-06,
      "loss": 0.2579,
      "step": 2527
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.719114576347384e-06,
      "loss": 0.8134,
      "step": 2528
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.7116327622898634e-06,
      "loss": 0.8401,
      "step": 2529
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.704153011590479e-06,
      "loss": 0.8446,
      "step": 2530
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.6966753289425715e-06,
      "loss": 0.8007,
      "step": 2531
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.6891997190381705e-06,
      "loss": 0.8002,
      "step": 2532
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.681726186568014e-06,
      "loss": 0.8228,
      "step": 2533
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.674254736221537e-06,
      "loss": 0.7897,
      "step": 2534
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.666785372686863e-06,
      "loss": 0.819,
      "step": 2535
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.659318100650804e-06,
      "loss": 0.7806,
      "step": 2536
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.6518529247988695e-06,
      "loss": 0.7419,
      "step": 2537
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.644389849815241e-06,
      "loss": 0.8479,
      "step": 2538
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.636928880382795e-06,
      "loss": 0.7764,
      "step": 2539
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.629470021183075e-06,
      "loss": 0.2486,
      "step": 2540
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.62201327689631e-06,
      "loss": 0.803,
      "step": 2541
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.614558652201397e-06,
      "loss": 0.783,
      "step": 2542
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.6071061517759014e-06,
      "loss": 0.7962,
      "step": 2543
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.599655780296062e-06,
      "loss": 0.7614,
      "step": 2544
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.592207542436775e-06,
      "loss": 0.8229,
      "step": 2545
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.584761442871601e-06,
      "loss": 0.8437,
      "step": 2546
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.577317486272756e-06,
      "loss": 0.6993,
      "step": 2547
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.569875677311118e-06,
      "loss": 0.7956,
      "step": 2548
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.562436020656213e-06,
      "loss": 0.7832,
      "step": 2549
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.554998520976214e-06,
      "loss": 0.8224,
      "step": 2550
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.547563182937946e-06,
      "loss": 0.8341,
      "step": 2551
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.540130011206871e-06,
      "loss": 0.7173,
      "step": 2552
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.532699010447101e-06,
      "loss": 0.7413,
      "step": 2553
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.525270185321375e-06,
      "loss": 0.8057,
      "step": 2554
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.517843540491078e-06,
      "loss": 0.779,
      "step": 2555
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.510419080616215e-06,
      "loss": 0.7734,
      "step": 2556
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.502996810355429e-06,
      "loss": 0.8648,
      "step": 2557
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.495576734365989e-06,
      "loss": 0.8246,
      "step": 2558
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.488158857303779e-06,
      "loss": 0.8154,
      "step": 2559
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.480743183823314e-06,
      "loss": 0.808,
      "step": 2560
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.473329718577715e-06,
      "loss": 0.7582,
      "step": 2561
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.46591846621873e-06,
      "loss": 0.8279,
      "step": 2562
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.458509431396707e-06,
      "loss": 0.7781,
      "step": 2563
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.451102618760608e-06,
      "loss": 0.2771,
      "step": 2564
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.443698032958003e-06,
      "loss": 0.7959,
      "step": 2565
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.436295678635059e-06,
      "loss": 0.7975,
      "step": 2566
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.428895560436548e-06,
      "loss": 0.8107,
      "step": 2567
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.421497683005831e-06,
      "loss": 0.8021,
      "step": 2568
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.414102050984876e-06,
      "loss": 0.7993,
      "step": 2569
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.406708669014228e-06,
      "loss": 0.8921,
      "step": 2570
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.39931754173303e-06,
      "loss": 0.8018,
      "step": 2571
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.391928673779007e-06,
      "loss": 0.7722,
      "step": 2572
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.3845420697884645e-06,
      "loss": 0.7705,
      "step": 2573
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.377157734396291e-06,
      "loss": 0.8219,
      "step": 2574
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.369775672235948e-06,
      "loss": 0.2534,
      "step": 2575
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.362395887939475e-06,
      "loss": 0.7736,
      "step": 2576
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.355018386137475e-06,
      "loss": 0.2536,
      "step": 2577
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.347643171459127e-06,
      "loss": 0.8548,
      "step": 2578
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.340270248532172e-06,
      "loss": 0.8019,
      "step": 2579
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.332899621982911e-06,
      "loss": 0.7351,
      "step": 2580
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.325531296436202e-06,
      "loss": 0.8333,
      "step": 2581
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.318165276515467e-06,
      "loss": 0.7617,
      "step": 2582
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.310801566842671e-06,
      "loss": 0.8848,
      "step": 2583
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.303440172038342e-06,
      "loss": 0.8139,
      "step": 2584
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.296081096721545e-06,
      "loss": 0.7854,
      "step": 2585
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.288724345509891e-06,
      "loss": 0.7271,
      "step": 2586
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.281369923019543e-06,
      "loss": 0.7867,
      "step": 2587
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.274017833865185e-06,
      "loss": 0.78,
      "step": 2588
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.266668082660053e-06,
      "loss": 0.76,
      "step": 2589
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.259320674015905e-06,
      "loss": 0.8218,
      "step": 2590
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.25197561254304e-06,
      "loss": 0.848,
      "step": 2591
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.244632902850269e-06,
      "loss": 0.8197,
      "step": 2592
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.237292549544944e-06,
      "loss": 0.7901,
      "step": 2593
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.229954557232929e-06,
      "loss": 0.8374,
      "step": 2594
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.222618930518605e-06,
      "loss": 0.8185,
      "step": 2595
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.215285674004876e-06,
      "loss": 0.8713,
      "step": 2596
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.207954792293151e-06,
      "loss": 0.8183,
      "step": 2597
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.200626289983356e-06,
      "loss": 0.7516,
      "step": 2598
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1933001716739176e-06,
      "loss": 0.7339,
      "step": 2599
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.18597644196177e-06,
      "loss": 0.7842,
      "step": 2600
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1786551054423485e-06,
      "loss": 0.7299,
      "step": 2601
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.171336166709585e-06,
      "loss": 0.8378,
      "step": 2602
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.164019630355913e-06,
      "loss": 0.7511,
      "step": 2603
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.156705500972246e-06,
      "loss": 0.7565,
      "step": 2604
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.149393783148002e-06,
      "loss": 0.7608,
      "step": 2605
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.142084481471073e-06,
      "loss": 0.7522,
      "step": 2606
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.134777600527846e-06,
      "loss": 0.7997,
      "step": 2607
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.127473144903179e-06,
      "loss": 0.2444,
      "step": 2608
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.120171119180414e-06,
      "loss": 0.7799,
      "step": 2609
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1128715279413705e-06,
      "loss": 0.8636,
      "step": 2610
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.105574375766332e-06,
      "loss": 0.7845,
      "step": 2611
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.098279667234061e-06,
      "loss": 0.266,
      "step": 2612
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.09098740692178e-06,
      "loss": 0.8436,
      "step": 2613
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.083697599405181e-06,
      "loss": 0.7908,
      "step": 2614
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.076410249258406e-06,
      "loss": 0.7958,
      "step": 2615
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.069125361054069e-06,
      "loss": 0.74,
      "step": 2616
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.061842939363234e-06,
      "loss": 0.7445,
      "step": 2617
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.054562988755411e-06,
      "loss": 0.7691,
      "step": 2618
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.047285513798569e-06,
      "loss": 0.8607,
      "step": 2619
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.040010519059115e-06,
      "loss": 0.7766,
      "step": 2620
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.032738009101909e-06,
      "loss": 0.798,
      "step": 2621
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.025467988490244e-06,
      "loss": 0.7804,
      "step": 2622
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.018200461785851e-06,
      "loss": 0.8165,
      "step": 2623
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.010935433548907e-06,
      "loss": 0.7563,
      "step": 2624
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0036729083380095e-06,
      "loss": 0.8044,
      "step": 2625
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.996412890710184e-06,
      "loss": 0.8137,
      "step": 2626
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.989155385220895e-06,
      "loss": 0.7955,
      "step": 2627
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9819003964240175e-06,
      "loss": 0.7962,
      "step": 2628
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.974647928871859e-06,
      "loss": 0.7601,
      "step": 2629
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.967397987115134e-06,
      "loss": 0.2736,
      "step": 2630
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.960150575702981e-06,
      "loss": 0.779,
      "step": 2631
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.952905699182948e-06,
      "loss": 0.226,
      "step": 2632
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.945663362100989e-06,
      "loss": 0.835,
      "step": 2633
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.938423569001468e-06,
      "loss": 0.7883,
      "step": 2634
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.931186324427152e-06,
      "loss": 0.8269,
      "step": 2635
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9239516329192106e-06,
      "loss": 0.809,
      "step": 2636
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.916719499017206e-06,
      "loss": 0.8069,
      "step": 2637
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.909489927259101e-06,
      "loss": 0.8071,
      "step": 2638
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.9022629221812525e-06,
      "loss": 0.8583,
      "step": 2639
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8950384883183965e-06,
      "loss": 0.7873,
      "step": 2640
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.887816630203668e-06,
      "loss": 0.8267,
      "step": 2641
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.880597352368578e-06,
      "loss": 0.7936,
      "step": 2642
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.873380659343022e-06,
      "loss": 0.7998,
      "step": 2643
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8661665556552695e-06,
      "loss": 0.739,
      "step": 2644
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.858955045831969e-06,
      "loss": 0.8332,
      "step": 2645
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.851746134398144e-06,
      "loss": 0.2529,
      "step": 2646
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.84453982587718e-06,
      "loss": 0.7979,
      "step": 2647
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.837336124790834e-06,
      "loss": 0.7881,
      "step": 2648
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.83013503565923e-06,
      "loss": 0.7724,
      "step": 2649
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.822936563000845e-06,
      "loss": 0.8518,
      "step": 2650
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.815740711332521e-06,
      "loss": 0.848,
      "step": 2651
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.808547485169447e-06,
      "loss": 0.7673,
      "step": 2652
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8013568890251804e-06,
      "loss": 0.2766,
      "step": 2653
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.794168927411612e-06,
      "loss": 0.2891,
      "step": 2654
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.786983604838984e-06,
      "loss": 0.7716,
      "step": 2655
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.779800925815892e-06,
      "loss": 0.7342,
      "step": 2656
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.772620894849261e-06,
      "loss": 0.7981,
      "step": 2657
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.76544351644436e-06,
      "loss": 0.255,
      "step": 2658
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7582687951047864e-06,
      "loss": 0.7475,
      "step": 2659
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.751096735332486e-06,
      "loss": 0.8062,
      "step": 2660
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.743927341627722e-06,
      "loss": 0.8033,
      "step": 2661
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.736760618489083e-06,
      "loss": 0.7092,
      "step": 2662
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.729596570413493e-06,
      "loss": 0.8509,
      "step": 2663
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7224352018961885e-06,
      "loss": 0.781,
      "step": 2664
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.715276517430726e-06,
      "loss": 0.7872,
      "step": 2665
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.708120521508978e-06,
      "loss": 0.7871,
      "step": 2666
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.700967218621134e-06,
      "loss": 0.8161,
      "step": 2667
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.693816613255693e-06,
      "loss": 0.7578,
      "step": 2668
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.68666870989945e-06,
      "loss": 0.8119,
      "step": 2669
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.679523513037528e-06,
      "loss": 0.8076,
      "step": 2670
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6723810271533265e-06,
      "loss": 0.7549,
      "step": 2671
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.665241256728553e-06,
      "loss": 0.8145,
      "step": 2672
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.658104206243222e-06,
      "loss": 0.8071,
      "step": 2673
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.650969880175628e-06,
      "loss": 0.7823,
      "step": 2674
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.643838283002358e-06,
      "loss": 0.8175,
      "step": 2675
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.63670941919829e-06,
      "loss": 0.7698,
      "step": 2676
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.62958329323659e-06,
      "loss": 0.7348,
      "step": 2677
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6224599095887e-06,
      "loss": 0.8289,
      "step": 2678
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.615339272724338e-06,
      "loss": 0.7631,
      "step": 2679
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.608221387111514e-06,
      "loss": 0.7396,
      "step": 2680
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.601106257216495e-06,
      "loss": 0.7798,
      "step": 2681
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.593993887503826e-06,
      "loss": 0.8317,
      "step": 2682
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5868842824363145e-06,
      "loss": 0.8522,
      "step": 2683
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5797774464750475e-06,
      "loss": 0.7839,
      "step": 2684
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.572673384079361e-06,
      "loss": 0.7691,
      "step": 2685
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.565572099706851e-06,
      "loss": 0.7618,
      "step": 2686
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.55847359781338e-06,
      "loss": 0.8155,
      "step": 2687
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5513778828530595e-06,
      "loss": 0.8356,
      "step": 2688
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.544284959278246e-06,
      "loss": 0.7731,
      "step": 2689
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.537194831539548e-06,
      "loss": 0.8098,
      "step": 2690
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.53010750408583e-06,
      "loss": 0.7736,
      "step": 2691
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.523022981364187e-06,
      "loss": 0.2582,
      "step": 2692
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.515941267819953e-06,
      "loss": 0.2657,
      "step": 2693
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.508862367896714e-06,
      "loss": 0.7893,
      "step": 2694
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.501786286036277e-06,
      "loss": 0.7952,
      "step": 2695
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.494713026678684e-06,
      "loss": 0.819,
      "step": 2696
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.487642594262203e-06,
      "loss": 0.8231,
      "step": 2697
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.48057499322334e-06,
      "loss": 0.766,
      "step": 2698
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.473510227996812e-06,
      "loss": 0.2634,
      "step": 2699
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.466448303015557e-06,
      "loss": 0.7312,
      "step": 2700
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.459389222710742e-06,
      "loss": 0.7464,
      "step": 2701
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.452332991511738e-06,
      "loss": 0.7484,
      "step": 2702
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.445279613846132e-06,
      "loss": 0.7847,
      "step": 2703
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4382290941397195e-06,
      "loss": 0.7422,
      "step": 2704
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.431181436816507e-06,
      "loss": 0.2509,
      "step": 2705
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.424136646298701e-06,
      "loss": 0.8471,
      "step": 2706
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4170947270067045e-06,
      "loss": 0.7786,
      "step": 2707
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.410055683359131e-06,
      "loss": 0.7951,
      "step": 2708
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.403019519772781e-06,
      "loss": 0.7831,
      "step": 2709
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.395986240662648e-06,
      "loss": 0.7846,
      "step": 2710
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.388955850441915e-06,
      "loss": 0.8529,
      "step": 2711
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.381928353521961e-06,
      "loss": 0.8366,
      "step": 2712
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.37490375431234e-06,
      "loss": 0.7925,
      "step": 2713
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.367882057220785e-06,
      "loss": 0.7908,
      "step": 2714
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.360863266653228e-06,
      "loss": 0.7766,
      "step": 2715
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.353847387013751e-06,
      "loss": 0.7677,
      "step": 2716
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3468344227046195e-06,
      "loss": 0.8215,
      "step": 2717
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.339824378126282e-06,
      "loss": 0.8025,
      "step": 2718
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3328172576773395e-06,
      "loss": 0.7636,
      "step": 2719
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.325813065754563e-06,
      "loss": 0.7962,
      "step": 2720
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.318811806752884e-06,
      "loss": 0.8444,
      "step": 2721
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.311813485065402e-06,
      "loss": 0.8208,
      "step": 2722
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.304818105083367e-06,
      "loss": 0.7312,
      "step": 2723
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.297825671196175e-06,
      "loss": 0.7675,
      "step": 2724
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.290836187791392e-06,
      "loss": 0.7252,
      "step": 2725
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.283849659254718e-06,
      "loss": 0.7928,
      "step": 2726
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2768660899700056e-06,
      "loss": 0.2503,
      "step": 2727
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.269885484319241e-06,
      "loss": 0.8112,
      "step": 2728
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.262907846682571e-06,
      "loss": 0.2409,
      "step": 2729
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.25593318143826e-06,
      "loss": 0.7313,
      "step": 2730
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.248961492962711e-06,
      "loss": 0.8347,
      "step": 2731
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.241992785630472e-06,
      "loss": 0.7841,
      "step": 2732
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.235027063814206e-06,
      "loss": 0.8527,
      "step": 2733
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2280643318847076e-06,
      "loss": 0.2532,
      "step": 2734
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.221104594210893e-06,
      "loss": 0.7737,
      "step": 2735
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.21414785515981e-06,
      "loss": 0.8133,
      "step": 2736
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.207194119096613e-06,
      "loss": 0.8497,
      "step": 2737
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2002433903845714e-06,
      "loss": 0.8106,
      "step": 2738
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.193295673385081e-06,
      "loss": 0.7442,
      "step": 2739
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.186350972457636e-06,
      "loss": 0.819,
      "step": 2740
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.179409291959841e-06,
      "loss": 0.8172,
      "step": 2741
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1724706362473995e-06,
      "loss": 0.826,
      "step": 2742
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.165535009674133e-06,
      "loss": 0.7507,
      "step": 2743
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.158602416591949e-06,
      "loss": 0.7666,
      "step": 2744
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.151672861350849e-06,
      "loss": 0.7968,
      "step": 2745
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.144746348298946e-06,
      "loss": 0.7885,
      "step": 2746
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1378228817824235e-06,
      "loss": 0.8086,
      "step": 2747
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.130902466145567e-06,
      "loss": 0.8187,
      "step": 2748
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.123985105730738e-06,
      "loss": 0.747,
      "step": 2749
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.117070804878393e-06,
      "loss": 0.7943,
      "step": 2750
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1101595679270566e-06,
      "loss": 0.8146,
      "step": 2751
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.103251399213334e-06,
      "loss": 0.7421,
      "step": 2752
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.096346303071915e-06,
      "loss": 0.7649,
      "step": 2753
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.089444283835549e-06,
      "loss": 0.787,
      "step": 2754
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.082545345835059e-06,
      "loss": 0.8269,
      "step": 2755
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.075649493399332e-06,
      "loss": 0.7842,
      "step": 2756
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.068756730855329e-06,
      "loss": 0.8217,
      "step": 2757
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.061867062528061e-06,
      "loss": 0.7879,
      "step": 2758
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.054980492740598e-06,
      "loss": 0.7814,
      "step": 2759
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.0480970258140805e-06,
      "loss": 0.762,
      "step": 2760
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.04121666606768e-06,
      "loss": 0.6977,
      "step": 2761
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.034339417818628e-06,
      "loss": 0.7633,
      "step": 2762
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.027465285382213e-06,
      "loss": 0.8245,
      "step": 2763
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.020594273071757e-06,
      "loss": 0.7942,
      "step": 2764
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.013726385198623e-06,
      "loss": 0.8153,
      "step": 2765
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.006861626072224e-06,
      "loss": 0.776,
      "step": 2766
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.000000000000003e-06,
      "loss": 0.7814,
      "step": 2767
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.993141511287435e-06,
      "loss": 0.8378,
      "step": 2768
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.986286164238026e-06,
      "loss": 0.2508,
      "step": 2769
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.979433963153322e-06,
      "loss": 0.7945,
      "step": 2770
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.972584912332883e-06,
      "loss": 0.8366,
      "step": 2771
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.965739016074295e-06,
      "loss": 0.7673,
      "step": 2772
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.958896278673163e-06,
      "loss": 0.8342,
      "step": 2773
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.952056704423122e-06,
      "loss": 0.732,
      "step": 2774
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.945220297615806e-06,
      "loss": 0.2529,
      "step": 2775
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.938387062540865e-06,
      "loss": 0.7988,
      "step": 2776
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.931557003485972e-06,
      "loss": 0.7671,
      "step": 2777
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.924730124736791e-06,
      "loss": 0.828,
      "step": 2778
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.917906430576999e-06,
      "loss": 0.7908,
      "step": 2779
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.911085925288266e-06,
      "loss": 0.819,
      "step": 2780
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.904268613150278e-06,
      "loss": 0.8208,
      "step": 2781
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.897454498440703e-06,
      "loss": 0.7514,
      "step": 2782
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8906435854352e-06,
      "loss": 0.6982,
      "step": 2783
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.883835878407436e-06,
      "loss": 0.7697,
      "step": 2784
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.877031381629053e-06,
      "loss": 0.8244,
      "step": 2785
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.870230099369679e-06,
      "loss": 0.7343,
      "step": 2786
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.863432035896924e-06,
      "loss": 0.7801,
      "step": 2787
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.856637195476394e-06,
      "loss": 0.7936,
      "step": 2788
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.849845582371652e-06,
      "loss": 0.763,
      "step": 2789
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.843057200844242e-06,
      "loss": 0.8009,
      "step": 2790
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.836272055153691e-06,
      "loss": 0.2714,
      "step": 2791
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8294901495574856e-06,
      "loss": 0.8182,
      "step": 2792
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.822711488311077e-06,
      "loss": 0.8035,
      "step": 2793
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.815936075667884e-06,
      "loss": 0.788,
      "step": 2794
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.809163915879295e-06,
      "loss": 0.7936,
      "step": 2795
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.802395013194646e-06,
      "loss": 0.8387,
      "step": 2796
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.795629371861229e-06,
      "loss": 0.7589,
      "step": 2797
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.788866996124304e-06,
      "loss": 0.7783,
      "step": 2798
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.782107890227066e-06,
      "loss": 0.8526,
      "step": 2799
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.775352058410664e-06,
      "loss": 0.7747,
      "step": 2800
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.768599504914189e-06,
      "loss": 0.8279,
      "step": 2801
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.76185023397469e-06,
      "loss": 0.7782,
      "step": 2802
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.755104249827136e-06,
      "loss": 0.8333,
      "step": 2803
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.748361556704443e-06,
      "loss": 0.7632,
      "step": 2804
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.74162215883747e-06,
      "loss": 0.8138,
      "step": 2805
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7348860604549905e-06,
      "loss": 0.76,
      "step": 2806
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.728153265783717e-06,
      "loss": 0.7964,
      "step": 2807
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.721423779048296e-06,
      "loss": 0.7902,
      "step": 2808
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.714697604471289e-06,
      "loss": 0.7661,
      "step": 2809
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.707974746273176e-06,
      "loss": 0.2577,
      "step": 2810
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.701255208672372e-06,
      "loss": 0.819,
      "step": 2811
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.694538995885192e-06,
      "loss": 0.8309,
      "step": 2812
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.687826112125872e-06,
      "loss": 0.8459,
      "step": 2813
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.681116561606554e-06,
      "loss": 0.8001,
      "step": 2814
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6744103485373005e-06,
      "loss": 0.8038,
      "step": 2815
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.667707477126069e-06,
      "loss": 0.7869,
      "step": 2816
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.661007951578722e-06,
      "loss": 0.7977,
      "step": 2817
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.65431177609902e-06,
      "loss": 0.7837,
      "step": 2818
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.647618954888633e-06,
      "loss": 0.7558,
      "step": 2819
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.640929492147116e-06,
      "loss": 0.7722,
      "step": 2820
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.634243392071912e-06,
      "loss": 0.7897,
      "step": 2821
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.627560658858372e-06,
      "loss": 0.7609,
      "step": 2822
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.620881296699719e-06,
      "loss": 0.7269,
      "step": 2823
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.614205309787063e-06,
      "loss": 0.8069,
      "step": 2824
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.607532702309397e-06,
      "loss": 0.7614,
      "step": 2825
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.600863478453602e-06,
      "loss": 0.7793,
      "step": 2826
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.594197642404426e-06,
      "loss": 0.2665,
      "step": 2827
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.587535198344487e-06,
      "loss": 0.7895,
      "step": 2828
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.580876150454292e-06,
      "loss": 0.8477,
      "step": 2829
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.574220502912202e-06,
      "loss": 0.7877,
      "step": 2830
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.567568259894449e-06,
      "loss": 0.7423,
      "step": 2831
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5609194255751255e-06,
      "loss": 0.7547,
      "step": 2832
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.554274004126196e-06,
      "loss": 0.7672,
      "step": 2833
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.54763199971747e-06,
      "loss": 0.7755,
      "step": 2834
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.540993416516618e-06,
      "loss": 0.2413,
      "step": 2835
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.53435825868917e-06,
      "loss": 0.7591,
      "step": 2836
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5277265303984995e-06,
      "loss": 0.776,
      "step": 2837
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5210982358058285e-06,
      "loss": 0.6938,
      "step": 2838
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.514473379070222e-06,
      "loss": 0.7903,
      "step": 2839
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.507851964348599e-06,
      "loss": 0.7586,
      "step": 2840
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.501233995795709e-06,
      "loss": 0.7632,
      "step": 2841
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.494619477564137e-06,
      "loss": 0.7474,
      "step": 2842
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4880084138043145e-06,
      "loss": 0.7814,
      "step": 2843
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.481400808664494e-06,
      "loss": 0.7816,
      "step": 2844
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.474796666290765e-06,
      "loss": 0.7356,
      "step": 2845
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.468195990827035e-06,
      "loss": 0.7852,
      "step": 2846
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.461598786415052e-06,
      "loss": 0.258,
      "step": 2847
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.455005057194372e-06,
      "loss": 0.7669,
      "step": 2848
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.44841480730237e-06,
      "loss": 0.8303,
      "step": 2849
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.441828040874256e-06,
      "loss": 0.7767,
      "step": 2850
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.435244762043029e-06,
      "loss": 0.7385,
      "step": 2851
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.428664974939513e-06,
      "loss": 0.7517,
      "step": 2852
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4220886836923445e-06,
      "loss": 0.7977,
      "step": 2853
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.415515892427961e-06,
      "loss": 0.7398,
      "step": 2854
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.408946605270597e-06,
      "loss": 0.7905,
      "step": 2855
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.402380826342307e-06,
      "loss": 0.803,
      "step": 2856
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.395818559762928e-06,
      "loss": 0.7388,
      "step": 2857
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.389259809650096e-06,
      "loss": 0.7163,
      "step": 2858
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.382704580119242e-06,
      "loss": 0.8362,
      "step": 2859
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.376152875283595e-06,
      "loss": 0.7457,
      "step": 2860
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3696046992541606e-06,
      "loss": 0.8313,
      "step": 2861
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.363060056139737e-06,
      "loss": 0.7983,
      "step": 2862
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.356518950046901e-06,
      "loss": 0.7578,
      "step": 2863
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.349981385080019e-06,
      "loss": 0.7711,
      "step": 2864
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.343447365341226e-06,
      "loss": 0.8207,
      "step": 2865
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.336916894930433e-06,
      "loss": 0.7974,
      "step": 2866
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.330389977945333e-06,
      "loss": 0.7449,
      "step": 2867
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3238666184813816e-06,
      "loss": 0.7204,
      "step": 2868
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.317346820631801e-06,
      "loss": 0.7587,
      "step": 2869
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.31083058848758e-06,
      "loss": 0.8143,
      "step": 2870
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.304317926137478e-06,
      "loss": 0.778,
      "step": 2871
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2978088376680036e-06,
      "loss": 0.7753,
      "step": 2872
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2913033271634255e-06,
      "loss": 0.7648,
      "step": 2873
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2848013987057735e-06,
      "loss": 0.746,
      "step": 2874
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.278303056374822e-06,
      "loss": 0.7669,
      "step": 2875
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.271808304248099e-06,
      "loss": 0.2885,
      "step": 2876
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.265317146400877e-06,
      "loss": 0.8023,
      "step": 2877
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2588295869061814e-06,
      "loss": 0.7601,
      "step": 2878
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.252345629834769e-06,
      "loss": 0.693,
      "step": 2879
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.245865279255138e-06,
      "loss": 0.7331,
      "step": 2880
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.239388539233534e-06,
      "loss": 0.747,
      "step": 2881
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.232915413833926e-06,
      "loss": 0.8363,
      "step": 2882
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.2264459071180184e-06,
      "loss": 0.7119,
      "step": 2883
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.219980023145241e-06,
      "loss": 0.7479,
      "step": 2884
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.213517765972762e-06,
      "loss": 0.7438,
      "step": 2885
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.207059139655464e-06,
      "loss": 0.8049,
      "step": 2886
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.200604148245948e-06,
      "loss": 0.7753,
      "step": 2887
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.194152795794549e-06,
      "loss": 0.7818,
      "step": 2888
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.187705086349304e-06,
      "loss": 0.7945,
      "step": 2889
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1812610239559705e-06,
      "loss": 0.795,
      "step": 2890
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.174820612658015e-06,
      "loss": 0.7873,
      "step": 2891
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.168383856496619e-06,
      "loss": 0.7292,
      "step": 2892
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.161950759510664e-06,
      "loss": 0.8633,
      "step": 2893
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.155521325736737e-06,
      "loss": 0.7485,
      "step": 2894
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.149095559209133e-06,
      "loss": 0.7886,
      "step": 2895
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1426734639598325e-06,
      "loss": 0.7594,
      "step": 2896
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.136255044018519e-06,
      "loss": 0.6805,
      "step": 2897
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.129840303412578e-06,
      "loss": 0.8057,
      "step": 2898
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.123429246167077e-06,
      "loss": 0.2554,
      "step": 2899
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.117021876304767e-06,
      "loss": 0.7792,
      "step": 2900
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.110618197846105e-06,
      "loss": 0.7035,
      "step": 2901
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.104218214809214e-06,
      "loss": 0.7535,
      "step": 2902
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.097821931209902e-06,
      "loss": 0.7733,
      "step": 2903
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.091429351061655e-06,
      "loss": 0.8321,
      "step": 2904
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.085040478375646e-06,
      "loss": 0.7453,
      "step": 2905
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.078655317160708e-06,
      "loss": 0.7977,
      "step": 2906
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.072273871423348e-06,
      "loss": 0.7654,
      "step": 2907
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.065896145167749e-06,
      "loss": 0.2584,
      "step": 2908
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.059522142395756e-06,
      "loss": 0.7855,
      "step": 2909
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.053151867106871e-06,
      "loss": 0.8045,
      "step": 2910
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.046785323298264e-06,
      "loss": 0.8068,
      "step": 2911
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.040422514964769e-06,
      "loss": 0.7964,
      "step": 2912
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.034063446098864e-06,
      "loss": 0.2973,
      "step": 2913
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.027708120690688e-06,
      "loss": 0.8374,
      "step": 2914
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.021356542728025e-06,
      "loss": 0.7884,
      "step": 2915
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.015008716196319e-06,
      "loss": 0.8153,
      "step": 2916
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.00866464507865e-06,
      "loss": 0.7372,
      "step": 2917
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.00232433335574e-06,
      "loss": 0.7552,
      "step": 2918
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.995987785005966e-06,
      "loss": 0.8068,
      "step": 2919
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.989655004005327e-06,
      "loss": 0.8345,
      "step": 2920
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.9833259943274686e-06,
      "loss": 0.7705,
      "step": 2921
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.977000759943661e-06,
      "loss": 0.713,
      "step": 2922
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.97067930482282e-06,
      "loss": 0.8561,
      "step": 2923
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.964361632931477e-06,
      "loss": 0.7154,
      "step": 2924
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.95804774823379e-06,
      "loss": 0.7375,
      "step": 2925
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.951737654691551e-06,
      "loss": 0.7504,
      "step": 2926
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.945431356264164e-06,
      "loss": 0.7821,
      "step": 2927
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.939128856908655e-06,
      "loss": 0.7426,
      "step": 2928
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.93283016057966e-06,
      "loss": 0.7092,
      "step": 2929
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.926535271229443e-06,
      "loss": 0.7365,
      "step": 2930
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.920244192807865e-06,
      "loss": 0.25,
      "step": 2931
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.913956929262398e-06,
      "loss": 0.7773,
      "step": 2932
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9076734845381326e-06,
      "loss": 0.716,
      "step": 2933
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.901393862577748e-06,
      "loss": 0.781,
      "step": 2934
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.895118067321532e-06,
      "loss": 0.8033,
      "step": 2935
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.888846102707365e-06,
      "loss": 0.7805,
      "step": 2936
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.882577972670737e-06,
      "loss": 0.7542,
      "step": 2937
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.876313681144721e-06,
      "loss": 0.8213,
      "step": 2938
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.870053232059979e-06,
      "loss": 0.7406,
      "step": 2939
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8637966293447785e-06,
      "loss": 0.7573,
      "step": 2940
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.857543876924949e-06,
      "loss": 0.7888,
      "step": 2941
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.851294978723922e-06,
      "loss": 0.7622,
      "step": 2942
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.845049938662709e-06,
      "loss": 0.7485,
      "step": 2943
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.838808760659895e-06,
      "loss": 0.7998,
      "step": 2944
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.832571448631641e-06,
      "loss": 0.8868,
      "step": 2945
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.826338006491691e-06,
      "loss": 0.2157,
      "step": 2946
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.820108438151352e-06,
      "loss": 0.2624,
      "step": 2947
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8138827475195037e-06,
      "loss": 0.2557,
      "step": 2948
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8076609385025886e-06,
      "loss": 0.7987,
      "step": 2949
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.801443015004623e-06,
      "loss": 0.7961,
      "step": 2950
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.795228980927178e-06,
      "loss": 0.8185,
      "step": 2951
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7890188401693805e-06,
      "loss": 0.7987,
      "step": 2952
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.782812596627927e-06,
      "loss": 0.7578,
      "step": 2953
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7766102541970572e-06,
      "loss": 0.7946,
      "step": 2954
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7704118167685676e-06,
      "loss": 0.7892,
      "step": 2955
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7642172882318007e-06,
      "loss": 0.6947,
      "step": 2956
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7580266724736545e-06,
      "loss": 0.7757,
      "step": 2957
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.751839973378564e-06,
      "loss": 0.7149,
      "step": 2958
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.745657194828509e-06,
      "loss": 0.7508,
      "step": 2959
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7394783407030066e-06,
      "loss": 0.7874,
      "step": 2960
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7333034148791213e-06,
      "loss": 0.7885,
      "step": 2961
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.727132421231442e-06,
      "loss": 0.7954,
      "step": 2962
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.720965363632091e-06,
      "loss": 0.744,
      "step": 2963
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7148022459507292e-06,
      "loss": 0.7768,
      "step": 2964
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.7086430720545374e-06,
      "loss": 0.7879,
      "step": 2965
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.702487845808225e-06,
      "loss": 0.808,
      "step": 2966
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6963365710740195e-06,
      "loss": 0.7312,
      "step": 2967
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6901892517116812e-06,
      "loss": 0.8292,
      "step": 2968
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6840458915784747e-06,
      "loss": 0.8028,
      "step": 2969
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6779064945291843e-06,
      "loss": 0.8423,
      "step": 2970
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6717710644161166e-06,
      "loss": 0.7889,
      "step": 2971
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.665639605089076e-06,
      "loss": 0.7472,
      "step": 2972
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.659512120395384e-06,
      "loss": 0.2644,
      "step": 2973
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6533886141798603e-06,
      "loss": 0.2949,
      "step": 2974
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6472690902848408e-06,
      "loss": 0.779,
      "step": 2975
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6411535525501505e-06,
      "loss": 0.8047,
      "step": 2976
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.635042004813114e-06,
      "loss": 0.8228,
      "step": 2977
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6289344509085643e-06,
      "loss": 0.8076,
      "step": 2978
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6228308946688163e-06,
      "loss": 0.8298,
      "step": 2979
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.616731339923679e-06,
      "loss": 0.7864,
      "step": 2980
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6106357905004497e-06,
      "loss": 0.7626,
      "step": 2981
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.60454425022392e-06,
      "loss": 0.7546,
      "step": 2982
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.598456722916359e-06,
      "loss": 0.8121,
      "step": 2983
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.592373212397514e-06,
      "loss": 0.8354,
      "step": 2984
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5862937224846284e-06,
      "loss": 0.8338,
      "step": 2985
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5802182569924014e-06,
      "loss": 0.8007,
      "step": 2986
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5741468197330154e-06,
      "loss": 0.7581,
      "step": 2987
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5680794145161356e-06,
      "loss": 0.7597,
      "step": 2988
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.562016045148884e-06,
      "loss": 0.735,
      "step": 2989
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.555956715435851e-06,
      "loss": 0.7792,
      "step": 2990
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5499014291791035e-06,
      "loss": 0.7868,
      "step": 2991
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5438501901781607e-06,
      "loss": 0.781,
      "step": 2992
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5378030022300037e-06,
      "loss": 0.7397,
      "step": 2993
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.531759869129071e-06,
      "loss": 0.7624,
      "step": 2994
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5257207946672657e-06,
      "loss": 0.7709,
      "step": 2995
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5196857826339337e-06,
      "loss": 0.7957,
      "step": 2996
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.513654836815872e-06,
      "loss": 0.7929,
      "step": 2997
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5076279609973375e-06,
      "loss": 0.8078,
      "step": 2998
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.501605158960022e-06,
      "loss": 0.761,
      "step": 2999
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4955864344830635e-06,
      "loss": 0.8183,
      "step": 3000
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4895717913430406e-06,
      "loss": 0.8187,
      "step": 3001
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.483561233313979e-06,
      "loss": 0.8033,
      "step": 3002
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4775547641673337e-06,
      "loss": 0.7782,
      "step": 3003
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4715523876719924e-06,
      "loss": 0.8172,
      "step": 3004
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4655541075942766e-06,
      "loss": 0.7287,
      "step": 3005
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4595599276979453e-06,
      "loss": 0.7408,
      "step": 3006
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.453569851744174e-06,
      "loss": 0.7651,
      "step": 3007
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.447583883491564e-06,
      "loss": 0.7493,
      "step": 3008
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4416020266961503e-06,
      "loss": 0.2459,
      "step": 3009
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.435624285111375e-06,
      "loss": 0.7975,
      "step": 3010
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4296506624881044e-06,
      "loss": 0.7811,
      "step": 3011
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.423681162574615e-06,
      "loss": 0.7974,
      "step": 3012
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4177157891166067e-06,
      "loss": 0.7975,
      "step": 3013
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.411754545857182e-06,
      "loss": 0.763,
      "step": 3014
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4057974365368496e-06,
      "loss": 0.7573,
      "step": 3015
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3998444648935346e-06,
      "loss": 0.7602,
      "step": 3016
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3938956346625575e-06,
      "loss": 0.7959,
      "step": 3017
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3879509495766404e-06,
      "loss": 0.8294,
      "step": 3018
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.382010413365905e-06,
      "loss": 0.8119,
      "step": 3019
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3760740297578763e-06,
      "loss": 0.7455,
      "step": 3020
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3701418024774656e-06,
      "loss": 0.7981,
      "step": 3021
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.364213735246976e-06,
      "loss": 0.7605,
      "step": 3022
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.358289831786109e-06,
      "loss": 0.7144,
      "step": 3023
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3523700958119454e-06,
      "loss": 0.7679,
      "step": 3024
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.346454531038953e-06,
      "loss": 0.2264,
      "step": 3025
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.340543141178979e-06,
      "loss": 0.7953,
      "step": 3026
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3346359299412624e-06,
      "loss": 0.7986,
      "step": 3027
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3287329010324077e-06,
      "loss": 0.7764,
      "step": 3028
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.322834058156397e-06,
      "loss": 0.7953,
      "step": 3029
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3169394050145987e-06,
      "loss": 0.7702,
      "step": 3030
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3110489453057316e-06,
      "loss": 0.7749,
      "step": 3031
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3051626827258955e-06,
      "loss": 0.6661,
      "step": 3032
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.29928062096856e-06,
      "loss": 0.8181,
      "step": 3033
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.293402763724551e-06,
      "loss": 0.7254,
      "step": 3034
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2875291146820553e-06,
      "loss": 0.8014,
      "step": 3035
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2816596775266296e-06,
      "loss": 0.7269,
      "step": 3036
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.275794455941178e-06,
      "loss": 0.7535,
      "step": 3037
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2699334536059622e-06,
      "loss": 0.7347,
      "step": 3038
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.264076674198594e-06,
      "loss": 0.796,
      "step": 3039
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2582241213940437e-06,
      "loss": 0.7301,
      "step": 3040
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.25237579886462e-06,
      "loss": 0.7261,
      "step": 3041
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.246531710279979e-06,
      "loss": 0.83,
      "step": 3042
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.240691859307129e-06,
      "loss": 0.8026,
      "step": 3043
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2348562496104085e-06,
      "loss": 0.7455,
      "step": 3044
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2290248848514993e-06,
      "loss": 0.8284,
      "step": 3045
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2231977686894156e-06,
      "loss": 0.255,
      "step": 3046
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.2173749047805158e-06,
      "loss": 0.8413,
      "step": 3047
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.2115562967784806e-06,
      "loss": 0.8443,
      "step": 3048
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.20574194833432e-06,
      "loss": 0.7585,
      "step": 3049
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1999318630963793e-06,
      "loss": 0.7497,
      "step": 3050
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1941260447103228e-06,
      "loss": 0.74,
      "step": 3051
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1883244968191384e-06,
      "loss": 0.8105,
      "step": 3052
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.18252722306313e-06,
      "loss": 0.2558,
      "step": 3053
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1767342270799304e-06,
      "loss": 0.7837,
      "step": 3054
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.170945512504481e-06,
      "loss": 0.7587,
      "step": 3055
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.165161082969036e-06,
      "loss": 0.8278,
      "step": 3056
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1593809421031583e-06,
      "loss": 0.7252,
      "step": 3057
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1536050935337336e-06,
      "loss": 0.8205,
      "step": 3058
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1478335408849392e-06,
      "loss": 0.7413,
      "step": 3059
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1420662877782604e-06,
      "loss": 0.7186,
      "step": 3060
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1363033378324916e-06,
      "loss": 0.8123,
      "step": 3061
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.13054469466372e-06,
      "loss": 0.7566,
      "step": 3062
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1247903618853326e-06,
      "loss": 0.8196,
      "step": 3063
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1190403431080073e-06,
      "loss": 0.803,
      "step": 3064
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1132946419397283e-06,
      "loss": 0.7515,
      "step": 3065
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.107553261985756e-06,
      "loss": 0.8258,
      "step": 3066
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1018162068486434e-06,
      "loss": 0.8379,
      "step": 3067
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.096083480128237e-06,
      "loss": 0.7978,
      "step": 3068
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0903550854216603e-06,
      "loss": 0.8371,
      "step": 3069
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0846310263233194e-06,
      "loss": 0.785,
      "step": 3070
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0789113064248967e-06,
      "loss": 0.8142,
      "step": 3071
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0731959293153622e-06,
      "loss": 0.7159,
      "step": 3072
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0674848985809536e-06,
      "loss": 0.7985,
      "step": 3073
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.061778217805177e-06,
      "loss": 0.7597,
      "step": 3074
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.056075890568825e-06,
      "loss": 0.7665,
      "step": 3075
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.050377920449937e-06,
      "loss": 0.8196,
      "step": 3076
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.044684311023831e-06,
      "loss": 0.7241,
      "step": 3077
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.038995065863093e-06,
      "loss": 0.7891,
      "step": 3078
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0333101885375613e-06,
      "loss": 0.7919,
      "step": 3079
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.027629682614334e-06,
      "loss": 0.7249,
      "step": 3080
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.021953551657774e-06,
      "loss": 0.2298,
      "step": 3081
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0162817992294925e-06,
      "loss": 0.7936,
      "step": 3082
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0106144288883553e-06,
      "loss": 0.8003,
      "step": 3083
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0049514441904736e-06,
      "loss": 0.7993,
      "step": 3084
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.999292848689218e-06,
      "loss": 0.7787,
      "step": 3085
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.9936386459351963e-06,
      "loss": 0.7655,
      "step": 3086
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.987988839476258e-06,
      "loss": 0.8176,
      "step": 3087
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.982343432857504e-06,
      "loss": 0.7867,
      "step": 3088
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9767024296212653e-06,
      "loss": 0.7609,
      "step": 3089
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9710658333071154e-06,
      "loss": 0.802,
      "step": 3090
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.965433647451854e-06,
      "loss": 0.7856,
      "step": 3091
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9598058755895276e-06,
      "loss": 0.8322,
      "step": 3092
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9541825212514007e-06,
      "loss": 0.7033,
      "step": 3093
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9485635879659693e-06,
      "loss": 0.7454,
      "step": 3094
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.942949079258961e-06,
      "loss": 0.8332,
      "step": 3095
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9373389986533195e-06,
      "loss": 0.7359,
      "step": 3096
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9317333496692123e-06,
      "loss": 0.8211,
      "step": 3097
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.926132135824026e-06,
      "loss": 0.7905,
      "step": 3098
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9205353606323683e-06,
      "loss": 0.763,
      "step": 3099
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9149430276060573e-06,
      "loss": 0.2551,
      "step": 3100
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9093551402541242e-06,
      "loss": 0.7794,
      "step": 3101
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9037717020828094e-06,
      "loss": 0.8167,
      "step": 3102
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8981927165955682e-06,
      "loss": 0.7809,
      "step": 3103
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8926181872930557e-06,
      "loss": 0.8173,
      "step": 3104
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.887048117673129e-06,
      "loss": 0.7343,
      "step": 3105
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8814825112308576e-06,
      "loss": 0.7712,
      "step": 3106
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8759213714584976e-06,
      "loss": 0.7852,
      "step": 3107
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8703647018455107e-06,
      "loss": 0.7575,
      "step": 3108
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.864812505878547e-06,
      "loss": 0.7846,
      "step": 3109
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.859264787041459e-06,
      "loss": 0.7765,
      "step": 3110
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8537215488152837e-06,
      "loss": 0.7898,
      "step": 3111
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8481827946782427e-06,
      "loss": 0.7669,
      "step": 3112
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8426485281057546e-06,
      "loss": 0.7299,
      "step": 3113
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8371187525704138e-06,
      "loss": 0.7439,
      "step": 3114
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.831593471542e-06,
      "loss": 0.7359,
      "step": 3115
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.826072688487468e-06,
      "loss": 0.828,
      "step": 3116
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.82055640687096e-06,
      "loss": 0.7826,
      "step": 3117
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.815044630153787e-06,
      "loss": 0.8488,
      "step": 3118
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.809537361794428e-06,
      "loss": 0.7876,
      "step": 3119
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.804034605248548e-06,
      "loss": 0.7497,
      "step": 3120
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.79853636396897e-06,
      "loss": 0.7579,
      "step": 3121
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.793042641405679e-06,
      "loss": 0.7778,
      "step": 3122
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7875534410058392e-06,
      "loss": 0.8313,
      "step": 3123
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.782068766213768e-06,
      "loss": 0.7789,
      "step": 3124
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7765886204709415e-06,
      "loss": 0.7504,
      "step": 3125
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.771113007216003e-06,
      "loss": 0.7907,
      "step": 3126
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7656419298847426e-06,
      "loss": 0.7836,
      "step": 3127
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7601753919101094e-06,
      "loss": 0.8028,
      "step": 3128
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7547133967221985e-06,
      "loss": 0.818,
      "step": 3129
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.749255947748265e-06,
      "loss": 0.7402,
      "step": 3130
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.743803048412701e-06,
      "loss": 0.8013,
      "step": 3131
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.738354702137046e-06,
      "loss": 0.7925,
      "step": 3132
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7329109123399888e-06,
      "loss": 0.789,
      "step": 3133
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7274716824373537e-06,
      "loss": 0.7623,
      "step": 3134
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.722037015842103e-06,
      "loss": 0.7852,
      "step": 3135
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.716606915964336e-06,
      "loss": 0.777,
      "step": 3136
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7111813862112924e-06,
      "loss": 0.7228,
      "step": 3137
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7057604299873375e-06,
      "loss": 0.739,
      "step": 3138
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.700344050693967e-06,
      "loss": 0.7975,
      "step": 3139
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6949322517298117e-06,
      "loss": 0.8083,
      "step": 3140
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.68952503649062e-06,
      "loss": 0.7664,
      "step": 3141
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6841224083692698e-06,
      "loss": 0.8392,
      "step": 3142
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.678724370755754e-06,
      "loss": 0.7905,
      "step": 3143
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6733309270371965e-06,
      "loss": 0.8064,
      "step": 3144
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6679420805978283e-06,
      "loss": 0.802,
      "step": 3145
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6625578348189994e-06,
      "loss": 0.8045,
      "step": 3146
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6571781930791706e-06,
      "loss": 0.8251,
      "step": 3147
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6518031587539207e-06,
      "loss": 0.7686,
      "step": 3148
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6464327352159303e-06,
      "loss": 0.7643,
      "step": 3149
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6410669258349875e-06,
      "loss": 0.8456,
      "step": 3150
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6357057339779914e-06,
      "loss": 0.8099,
      "step": 3151
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6303491630089385e-06,
      "loss": 0.7834,
      "step": 3152
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.624997216288925e-06,
      "loss": 0.2526,
      "step": 3153
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6196498971761442e-06,
      "loss": 0.7937,
      "step": 3154
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.614307209025896e-06,
      "loss": 0.7486,
      "step": 3155
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.608969155190563e-06,
      "loss": 0.7706,
      "step": 3156
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.603635739019622e-06,
      "loss": 0.8182,
      "step": 3157
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.598306963859648e-06,
      "loss": 0.8016,
      "step": 3158
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.592982833054294e-06,
      "loss": 0.7997,
      "step": 3159
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.587663349944304e-06,
      "loss": 0.7616,
      "step": 3160
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.582348517867499e-06,
      "loss": 0.798,
      "step": 3161
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.577038340158796e-06,
      "loss": 0.7193,
      "step": 3162
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5717328201501767e-06,
      "loss": 0.7699,
      "step": 3163
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5664319611707046e-06,
      "loss": 0.7862,
      "step": 3164
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5611357665465254e-06,
      "loss": 0.7829,
      "step": 3165
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5558442396008532e-06,
      "loss": 0.7846,
      "step": 3166
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5505573836539622e-06,
      "loss": 0.8051,
      "step": 3167
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5452752020232164e-06,
      "loss": 0.7522,
      "step": 3168
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5399976980230335e-06,
      "loss": 0.7614,
      "step": 3169
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5347248749648958e-06,
      "loss": 0.7586,
      "step": 3170
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.529456736157356e-06,
      "loss": 0.8099,
      "step": 3171
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5241932849060234e-06,
      "loss": 0.782,
      "step": 3172
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.518934524513564e-06,
      "loss": 0.8045,
      "step": 3173
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5136804582797002e-06,
      "loss": 0.8248,
      "step": 3174
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5084310895012178e-06,
      "loss": 0.8015,
      "step": 3175
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5031864214719424e-06,
      "loss": 0.2592,
      "step": 3176
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.497946457482756e-06,
      "loss": 0.8288,
      "step": 3177
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.492711200821594e-06,
      "loss": 0.804,
      "step": 3178
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4874806547734286e-06,
      "loss": 0.818,
      "step": 3179
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4822548226202837e-06,
      "loss": 0.7514,
      "step": 3180
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4770337076412164e-06,
      "loss": 0.7591,
      "step": 3181
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4718173131123367e-06,
      "loss": 0.8058,
      "step": 3182
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4666056423067832e-06,
      "loss": 0.7567,
      "step": 3183
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.461398698494729e-06,
      "loss": 0.7875,
      "step": 3184
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.456196484943393e-06,
      "loss": 0.8176,
      "step": 3185
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.450999004917013e-06,
      "loss": 0.7641,
      "step": 3186
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.445806261676863e-06,
      "loss": 0.7684,
      "step": 3187
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.440618258481241e-06,
      "loss": 0.7631,
      "step": 3188
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.43543499858548e-06,
      "loss": 0.7435,
      "step": 3189
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4302564852419243e-06,
      "loss": 0.7726,
      "step": 3190
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.425082721699945e-06,
      "loss": 0.6848,
      "step": 3191
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.419913711205939e-06,
      "loss": 0.7995,
      "step": 3192
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4147494570033116e-06,
      "loss": 0.8141,
      "step": 3193
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4095899623324892e-06,
      "loss": 0.8193,
      "step": 3194
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.404435230430905e-06,
      "loss": 0.7248,
      "step": 3195
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3992852645330145e-06,
      "loss": 0.7632,
      "step": 3196
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3941400678702754e-06,
      "loss": 0.7549,
      "step": 3197
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3889996436711525e-06,
      "loss": 0.2548,
      "step": 3198
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3838639951611155e-06,
      "loss": 0.8079,
      "step": 3199
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3787331255626454e-06,
      "loss": 0.7306,
      "step": 3200
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.373607038095217e-06,
      "loss": 0.7312,
      "step": 3201
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.368485735975302e-06,
      "loss": 0.7652,
      "step": 3202
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3633692224163807e-06,
      "loss": 0.7871,
      "step": 3203
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3582575006289164e-06,
      "loss": 0.7828,
      "step": 3204
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3531505738203742e-06,
      "loss": 0.797,
      "step": 3205
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3480484451952022e-06,
      "loss": 0.8153,
      "step": 3206
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3429511179548504e-06,
      "loss": 0.7475,
      "step": 3207
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.337858595297744e-06,
      "loss": 0.258,
      "step": 3208
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3327708804192963e-06,
      "loss": 0.7723,
      "step": 3209
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.327687976511911e-06,
      "loss": 0.7531,
      "step": 3210
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.322609886764967e-06,
      "loss": 0.8136,
      "step": 3211
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.317536614364816e-06,
      "loss": 0.6963,
      "step": 3212
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.312468162494801e-06,
      "loss": 0.7965,
      "step": 3213
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3074045343352314e-06,
      "loss": 0.8219,
      "step": 3214
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.302345733063388e-06,
      "loss": 0.8,
      "step": 3215
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2972917618535327e-06,
      "loss": 0.7882,
      "step": 3216
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.292242623876889e-06,
      "loss": 0.2835,
      "step": 3217
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.287198322301648e-06,
      "loss": 0.7854,
      "step": 3218
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2821588602929635e-06,
      "loss": 0.7726,
      "step": 3219
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2771242410129646e-06,
      "loss": 0.2274,
      "step": 3220
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2720944676207292e-06,
      "loss": 0.7577,
      "step": 3221
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2670695432722957e-06,
      "loss": 0.2632,
      "step": 3222
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2620494711206697e-06,
      "loss": 0.7882,
      "step": 3223
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.257034254315802e-06,
      "loss": 0.7653,
      "step": 3224
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2520238960046016e-06,
      "loss": 0.7951,
      "step": 3225
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2470183993309237e-06,
      "loss": 0.7442,
      "step": 3226
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2420177674355825e-06,
      "loss": 0.7977,
      "step": 3227
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.237022003456333e-06,
      "loss": 0.7944,
      "step": 3228
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2320311105278713e-06,
      "loss": 0.8157,
      "step": 3229
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.227045091781851e-06,
      "loss": 0.7641,
      "step": 3230
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2220639503468542e-06,
      "loss": 0.7934,
      "step": 3231
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2170876893484084e-06,
      "loss": 0.7998,
      "step": 3232
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.212116311908975e-06,
      "loss": 0.815,
      "step": 3233
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2071498211479582e-06,
      "loss": 0.7953,
      "step": 3234
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.20218822018169e-06,
      "loss": 0.7348,
      "step": 3235
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.197231512123432e-06,
      "loss": 0.7842,
      "step": 3236
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.192279700083385e-06,
      "loss": 0.7267,
      "step": 3237
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.18733278716867e-06,
      "loss": 0.7662,
      "step": 3238
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.182390776483336e-06,
      "loss": 0.7847,
      "step": 3239
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1774536711283535e-06,
      "loss": 0.7529,
      "step": 3240
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.172521474201621e-06,
      "loss": 0.7756,
      "step": 3241
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1675941887979536e-06,
      "loss": 0.7777,
      "step": 3242
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1626718180090824e-06,
      "loss": 0.7637,
      "step": 3243
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1577543649236542e-06,
      "loss": 0.7142,
      "step": 3244
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1528418326272403e-06,
      "loss": 0.8379,
      "step": 3245
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1479342242023137e-06,
      "loss": 0.7839,
      "step": 3246
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1430315427282556e-06,
      "loss": 0.7654,
      "step": 3247
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1381337912813693e-06,
      "loss": 0.7889,
      "step": 3248
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1332409729348526e-06,
      "loss": 0.8263,
      "step": 3249
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.128353090758811e-06,
      "loss": 0.8117,
      "step": 3250
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.123470147820252e-06,
      "loss": 0.8426,
      "step": 3251
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1185921471830895e-06,
      "loss": 0.7652,
      "step": 3252
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.113719091908131e-06,
      "loss": 0.7616,
      "step": 3253
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.108850985053077e-06,
      "loss": 0.8175,
      "step": 3254
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1039878296725335e-06,
      "loss": 0.7811,
      "step": 3255
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0991296288179965e-06,
      "loss": 0.7453,
      "step": 3256
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.094276385537839e-06,
      "loss": 0.7895,
      "step": 3257
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0894281028773456e-06,
      "loss": 0.8117,
      "step": 3258
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.084584783878675e-06,
      "loss": 0.6938,
      "step": 3259
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.079746431580869e-06,
      "loss": 0.7619,
      "step": 3260
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0749130490198655e-06,
      "loss": 0.8054,
      "step": 3261
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0700846392284725e-06,
      "loss": 0.747,
      "step": 3262
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.065261205236382e-06,
      "loss": 0.2415,
      "step": 3263
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0604427500701594e-06,
      "loss": 0.7979,
      "step": 3264
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.055629276753258e-06,
      "loss": 0.7368,
      "step": 3265
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.050820788305994e-06,
      "loss": 0.8144,
      "step": 3266
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0460172877455543e-06,
      "loss": 0.7247,
      "step": 3267
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.041218778086007e-06,
      "loss": 0.787,
      "step": 3268
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.036425262338281e-06,
      "loss": 0.7979,
      "step": 3269
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.03163674351017e-06,
      "loss": 0.7767,
      "step": 3270
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0268532246063356e-06,
      "loss": 0.275,
      "step": 3271
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0220747086283056e-06,
      "loss": 0.7704,
      "step": 3272
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0173011985744608e-06,
      "loss": 0.7649,
      "step": 3273
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.012532697440044e-06,
      "loss": 0.7468,
      "step": 3274
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0077692082171595e-06,
      "loss": 0.7624,
      "step": 3275
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.003010733894761e-06,
      "loss": 0.7982,
      "step": 3276
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9982572774586585e-06,
      "loss": 0.778,
      "step": 3277
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.993508841891509e-06,
      "loss": 0.8068,
      "step": 3278
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9887654301728287e-06,
      "loss": 0.7703,
      "step": 3279
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9840270452789713e-06,
      "loss": 0.227,
      "step": 3280
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9792936901831383e-06,
      "loss": 0.7604,
      "step": 3281
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.974565367855382e-06,
      "loss": 0.7411,
      "step": 3282
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9698420812625906e-06,
      "loss": 0.7163,
      "step": 3283
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.965123833368493e-06,
      "loss": 0.7639,
      "step": 3284
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.960410627133654e-06,
      "loss": 0.8084,
      "step": 3285
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9557024655154845e-06,
      "loss": 0.7249,
      "step": 3286
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.950999351468221e-06,
      "loss": 0.7513,
      "step": 3287
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9463012879429345e-06,
      "loss": 0.7561,
      "step": 3288
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.941608277887528e-06,
      "loss": 0.7713,
      "step": 3289
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.936920324246735e-06,
      "loss": 0.739,
      "step": 3290
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.932237429962116e-06,
      "loss": 0.7617,
      "step": 3291
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9275595979720506e-06,
      "loss": 0.7539,
      "step": 3292
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9228868312117523e-06,
      "loss": 0.7831,
      "step": 3293
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9182191326132506e-06,
      "loss": 0.8107,
      "step": 3294
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9135565051053928e-06,
      "loss": 0.7276,
      "step": 3295
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9088989516138466e-06,
      "loss": 0.7688,
      "step": 3296
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9042464750610989e-06,
      "loss": 0.7557,
      "step": 3297
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8995990783664475e-06,
      "loss": 0.7864,
      "step": 3298
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8949567644459998e-06,
      "loss": 0.7343,
      "step": 3299
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8903195362126824e-06,
      "loss": 0.7532,
      "step": 3300
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8856873965762268e-06,
      "loss": 0.2473,
      "step": 3301
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8810603484431622e-06,
      "loss": 0.7583,
      "step": 3302
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8764383947168386e-06,
      "loss": 0.2342,
      "step": 3303
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8718215382974015e-06,
      "loss": 0.8055,
      "step": 3304
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8672097820817935e-06,
      "loss": 0.8506,
      "step": 3305
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8626031289637703e-06,
      "loss": 0.73,
      "step": 3306
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8580015818338725e-06,
      "loss": 0.8041,
      "step": 3307
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8534051435794432e-06,
      "loss": 0.7266,
      "step": 3308
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8488138170846148e-06,
      "loss": 0.7239,
      "step": 3309
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8442276052303222e-06,
      "loss": 0.221,
      "step": 3310
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8396465108942807e-06,
      "loss": 0.8299,
      "step": 3311
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8350705369509957e-06,
      "loss": 0.7685,
      "step": 3312
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8304996862717695e-06,
      "loss": 0.7992,
      "step": 3313
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8259339617246774e-06,
      "loss": 0.7967,
      "step": 3314
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8213733661745858e-06,
      "loss": 0.2461,
      "step": 3315
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8168179024831366e-06,
      "loss": 0.7669,
      "step": 3316
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8122675735087602e-06,
      "loss": 0.8325,
      "step": 3317
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8077223821066592e-06,
      "loss": 0.8274,
      "step": 3318
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8031823311288088e-06,
      "loss": 0.8227,
      "step": 3319
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7986474234239703e-06,
      "loss": 0.75,
      "step": 3320
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7941176618376687e-06,
      "loss": 0.8315,
      "step": 3321
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7895930492122015e-06,
      "loss": 0.7441,
      "step": 3322
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7850735883866332e-06,
      "loss": 0.7406,
      "step": 3323
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7805592821968054e-06,
      "loss": 0.7805,
      "step": 3324
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7760501334753133e-06,
      "loss": 0.2661,
      "step": 3325
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7715461450515215e-06,
      "loss": 0.8018,
      "step": 3326
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7670473197515592e-06,
      "loss": 0.8382,
      "step": 3327
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7625536603983118e-06,
      "loss": 0.2543,
      "step": 3328
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7580651698114238e-06,
      "loss": 0.7771,
      "step": 3329
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7535818508072944e-06,
      "loss": 0.7577,
      "step": 3330
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7491037061990856e-06,
      "loss": 0.7604,
      "step": 3331
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7446307387967043e-06,
      "loss": 0.8029,
      "step": 3332
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.740162951406812e-06,
      "loss": 0.814,
      "step": 3333
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7357003468328216e-06,
      "loss": 0.7557,
      "step": 3334
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.731242927874891e-06,
      "loss": 0.7566,
      "step": 3335
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7267906973299265e-06,
      "loss": 0.7517,
      "step": 3336
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7223436579915754e-06,
      "loss": 0.7715,
      "step": 3337
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7179018126502322e-06,
      "loss": 0.7341,
      "step": 3338
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7134651640930311e-06,
      "loss": 0.7404,
      "step": 3339
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7090337151038417e-06,
      "loss": 0.8345,
      "step": 3340
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.704607468463274e-06,
      "loss": 0.7527,
      "step": 3341
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7001864269486768e-06,
      "loss": 0.8128,
      "step": 3342
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6957705933341273e-06,
      "loss": 0.783,
      "step": 3343
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6913599703904348e-06,
      "loss": 0.2744,
      "step": 3344
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6869545608851467e-06,
      "loss": 0.7266,
      "step": 3345
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6825543675825329e-06,
      "loss": 0.767,
      "step": 3346
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6781593932435847e-06,
      "loss": 0.7211,
      "step": 3347
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6737696406260318e-06,
      "loss": 0.7512,
      "step": 3348
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6693851124843197e-06,
      "loss": 0.2689,
      "step": 3349
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6650058115696122e-06,
      "loss": 0.7909,
      "step": 3350
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6606317406298045e-06,
      "loss": 0.7363,
      "step": 3351
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6562629024095e-06,
      "loss": 0.7597,
      "step": 3352
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6518992996500217e-06,
      "loss": 0.7982,
      "step": 3353
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6475409350894056e-06,
      "loss": 0.7488,
      "step": 3354
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.643187811462409e-06,
      "loss": 0.793,
      "step": 3355
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.638839931500491e-06,
      "loss": 0.8168,
      "step": 3356
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.634497297931823e-06,
      "loss": 0.7345,
      "step": 3357
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6301599134812895e-06,
      "loss": 0.7615,
      "step": 3358
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6258277808704758e-06,
      "loss": 0.754,
      "step": 3359
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.621500902817673e-06,
      "loss": 0.7593,
      "step": 3360
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.617179282037873e-06,
      "loss": 0.7772,
      "step": 3361
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6128629212427771e-06,
      "loss": 0.8301,
      "step": 3362
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6085518231407781e-06,
      "loss": 0.7752,
      "step": 3363
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6042459904369667e-06,
      "loss": 0.7656,
      "step": 3364
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.599945425833136e-06,
      "loss": 0.796,
      "step": 3365
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5956501320277674e-06,
      "loss": 0.7605,
      "step": 3366
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5913601117160371e-06,
      "loss": 0.2528,
      "step": 3367
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.587075367589812e-06,
      "loss": 0.8028,
      "step": 3368
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5827959023376505e-06,
      "loss": 0.7472,
      "step": 3369
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5785217186447964e-06,
      "loss": 0.7559,
      "step": 3370
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5742528191931773e-06,
      "loss": 0.7527,
      "step": 3371
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5699892066614131e-06,
      "loss": 0.7433,
      "step": 3372
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5657308837247976e-06,
      "loss": 0.8013,
      "step": 3373
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5614778530553087e-06,
      "loss": 0.7883,
      "step": 3374
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.557230117321602e-06,
      "loss": 0.7925,
      "step": 3375
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5529876791890174e-06,
      "loss": 0.7989,
      "step": 3376
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5487505413195637e-06,
      "loss": 0.7575,
      "step": 3377
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5445187063719225e-06,
      "loss": 0.7564,
      "step": 3378
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5402921770014567e-06,
      "loss": 0.7698,
      "step": 3379
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.536070955860194e-06,
      "loss": 0.7983,
      "step": 3380
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5318550455968285e-06,
      "loss": 0.8025,
      "step": 3381
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5276444488567267e-06,
      "loss": 0.7519,
      "step": 3382
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5234391682819227e-06,
      "loss": 0.2311,
      "step": 3383
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5192392065111106e-06,
      "loss": 0.7718,
      "step": 3384
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5150445661796476e-06,
      "loss": 0.771,
      "step": 3385
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5108552499195517e-06,
      "loss": 0.833,
      "step": 3386
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5066712603595045e-06,
      "loss": 0.8132,
      "step": 3387
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5024926001248409e-06,
      "loss": 0.7933,
      "step": 3388
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4983192718375494e-06,
      "loss": 0.8222,
      "step": 3389
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4941512781162825e-06,
      "loss": 0.7003,
      "step": 3390
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4899886215763382e-06,
      "loss": 0.7266,
      "step": 3391
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4858313048296602e-06,
      "loss": 0.7882,
      "step": 3392
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4816793304848543e-06,
      "loss": 0.6801,
      "step": 3393
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4775327011471662e-06,
      "loss": 0.7636,
      "step": 3394
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4733914194184873e-06,
      "loss": 0.7818,
      "step": 3395
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4692554878973586e-06,
      "loss": 0.7369,
      "step": 3396
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.465124909178961e-06,
      "loss": 0.7981,
      "step": 3397
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4609996858551156e-06,
      "loss": 0.7822,
      "step": 3398
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.456879820514282e-06,
      "loss": 0.7283,
      "step": 3399
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4527653157415644e-06,
      "loss": 0.7285,
      "step": 3400
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4486561741186967e-06,
      "loss": 0.2685,
      "step": 3401
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4445523982240483e-06,
      "loss": 0.8006,
      "step": 3402
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4404539906326288e-06,
      "loss": 0.8075,
      "step": 3403
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4363609539160694e-06,
      "loss": 0.7813,
      "step": 3404
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4322732906426362e-06,
      "loss": 0.7625,
      "step": 3405
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4281910033772228e-06,
      "loss": 0.7795,
      "step": 3406
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4241140946813526e-06,
      "loss": 0.7918,
      "step": 3407
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.42004256711317e-06,
      "loss": 0.8316,
      "step": 3408
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4159764232274408e-06,
      "loss": 0.7631,
      "step": 3409
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4119156655755618e-06,
      "loss": 0.7329,
      "step": 3410
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.407860296705542e-06,
      "loss": 0.7946,
      "step": 3411
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4038103191620123e-06,
      "loss": 0.7329,
      "step": 3412
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.399765735486216e-06,
      "loss": 0.8181,
      "step": 3413
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.395726548216021e-06,
      "loss": 0.7447,
      "step": 3414
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.391692759885902e-06,
      "loss": 0.7269,
      "step": 3415
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3876643730269457e-06,
      "loss": 0.8039,
      "step": 3416
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3836413901668566e-06,
      "loss": 0.696,
      "step": 3417
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.37962381382994e-06,
      "loss": 0.8294,
      "step": 3418
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3756116465371128e-06,
      "loss": 0.7474,
      "step": 3419
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3716048908058965e-06,
      "loss": 0.7193,
      "step": 3420
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3676035491504202e-06,
      "loss": 0.7419,
      "step": 3421
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3636076240814122e-06,
      "loss": 0.725,
      "step": 3422
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3596171181062024e-06,
      "loss": 0.7671,
      "step": 3423
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3556320337287232e-06,
      "loss": 0.7409,
      "step": 3424
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3516523734495035e-06,
      "loss": 0.8065,
      "step": 3425
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3476781397656657e-06,
      "loss": 0.286,
      "step": 3426
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3437093351709296e-06,
      "loss": 0.7546,
      "step": 3427
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.339745962155613e-06,
      "loss": 0.7709,
      "step": 3428
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.335788023206619e-06,
      "loss": 0.7381,
      "step": 3429
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3318355208074428e-06,
      "loss": 0.778,
      "step": 3430
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3278884574381657e-06,
      "loss": 0.7204,
      "step": 3431
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3239468355754648e-06,
      "loss": 0.7542,
      "step": 3432
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3200106576925952e-06,
      "loss": 0.75,
      "step": 3433
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3160799262593947e-06,
      "loss": 0.7823,
      "step": 3434
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3121546437422916e-06,
      "loss": 0.8194,
      "step": 3435
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.308234812604291e-06,
      "loss": 0.7471,
      "step": 3436
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3043204353049709e-06,
      "loss": 0.7905,
      "step": 3437
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.300411514300498e-06,
      "loss": 0.7595,
      "step": 3438
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2965080520436101e-06,
      "loss": 0.7066,
      "step": 3439
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.292610050983617e-06,
      "loss": 0.765,
      "step": 3440
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2887175135664088e-06,
      "loss": 0.2447,
      "step": 3441
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2848304422344427e-06,
      "loss": 0.7624,
      "step": 3442
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2809488394267455e-06,
      "loss": 0.7665,
      "step": 3443
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2770727075789136e-06,
      "loss": 0.7444,
      "step": 3444
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2732020491231133e-06,
      "loss": 0.769,
      "step": 3445
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2693368664880735e-06,
      "loss": 0.762,
      "step": 3446
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2654771620990846e-06,
      "loss": 0.7326,
      "step": 3447
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2616229383780076e-06,
      "loss": 0.2609,
      "step": 3448
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2577741977432567e-06,
      "loss": 0.7312,
      "step": 3449
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2539309426098101e-06,
      "loss": 0.7139,
      "step": 3450
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2500931753891998e-06,
      "loss": 0.8623,
      "step": 3451
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2462608984895207e-06,
      "loss": 0.7413,
      "step": 3452
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2424341143154173e-06,
      "loss": 0.2414,
      "step": 3453
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2386128252680873e-06,
      "loss": 0.7847,
      "step": 3454
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2347970337452865e-06,
      "loss": 0.7564,
      "step": 3455
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2309867421413147e-06,
      "loss": 0.7956,
      "step": 3456
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.227181952847022e-06,
      "loss": 0.7541,
      "step": 3457
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2233826682498061e-06,
      "loss": 0.803,
      "step": 3458
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2195888907336162e-06,
      "loss": 0.7807,
      "step": 3459
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2158006226789376e-06,
      "loss": 0.7745,
      "step": 3460
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2120178664628002e-06,
      "loss": 0.7959,
      "step": 3461
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.208240624458782e-06,
      "loss": 0.8313,
      "step": 3462
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2044688990369946e-06,
      "loss": 0.7683,
      "step": 3463
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2007026925640897e-06,
      "loss": 0.8093,
      "step": 3464
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1969420074032535e-06,
      "loss": 0.7802,
      "step": 3465
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1931868459142148e-06,
      "loss": 0.7768,
      "step": 3466
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1894372104532292e-06,
      "loss": 0.7585,
      "step": 3467
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.185693103373088e-06,
      "loss": 0.7896,
      "step": 3468
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1819545270231147e-06,
      "loss": 0.7149,
      "step": 3469
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.17822148374916e-06,
      "loss": 0.7672,
      "step": 3470
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1744939758936047e-06,
      "loss": 0.7856,
      "step": 3471
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.170772005795352e-06,
      "loss": 0.7336,
      "step": 3472
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1670555757898382e-06,
      "loss": 0.7181,
      "step": 3473
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.163344688209017e-06,
      "loss": 0.7977,
      "step": 3474
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1596393453813647e-06,
      "loss": 0.7965,
      "step": 3475
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1559395496318826e-06,
      "loss": 0.2334,
      "step": 3476
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1522453032820868e-06,
      "loss": 0.716,
      "step": 3477
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1485566086500144e-06,
      "loss": 0.2686,
      "step": 3478
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.144873468050215e-06,
      "loss": 0.8162,
      "step": 3479
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1411958837937598e-06,
      "loss": 0.8054,
      "step": 3480
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1375238581882298e-06,
      "loss": 0.7445,
      "step": 3481
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1338573935377128e-06,
      "loss": 0.7153,
      "step": 3482
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1301964921428165e-06,
      "loss": 0.2492,
      "step": 3483
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1265411563006545e-06,
      "loss": 0.7898,
      "step": 3484
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.122891388304843e-06,
      "loss": 0.7696,
      "step": 3485
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1192471904455138e-06,
      "loss": 0.7831,
      "step": 3486
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1156085650092975e-06,
      "loss": 0.7306,
      "step": 3487
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1119755142793276e-06,
      "loss": 0.79,
      "step": 3488
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.108348040535242e-06,
      "loss": 0.7576,
      "step": 3489
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1047261460531799e-06,
      "loss": 0.7435,
      "step": 3490
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.101109833105779e-06,
      "loss": 0.7864,
      "step": 3491
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0974991039621707e-06,
      "loss": 0.2597,
      "step": 3492
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0938939608879905e-06,
      "loss": 0.8341,
      "step": 3493
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0902944061453625e-06,
      "loss": 0.7132,
      "step": 3494
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0867004419929061e-06,
      "loss": 0.7864,
      "step": 3495
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0831120706857312e-06,
      "loss": 0.7169,
      "step": 3496
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0795292944754442e-06,
      "loss": 0.792,
      "step": 3497
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0759521156101337e-06,
      "loss": 0.7526,
      "step": 3498
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0723805363343788e-06,
      "loss": 0.2319,
      "step": 3499
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0688145588892473e-06,
      "loss": 0.7672,
      "step": 3500
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0652541855122888e-06,
      "loss": 0.7911,
      "step": 3501
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0616994184375373e-06,
      "loss": 0.7433,
      "step": 3502
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0581502598955084e-06,
      "loss": 0.7832,
      "step": 3503
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.054606712113203e-06,
      "loss": 0.7149,
      "step": 3504
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0510687773140948e-06,
      "loss": 0.7875,
      "step": 3505
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.047536457718138e-06,
      "loss": 0.7569,
      "step": 3506
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.044009755541766e-06,
      "loss": 0.2551,
      "step": 3507
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.040488672997887e-06,
      "loss": 0.2359,
      "step": 3508
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0369732122958775e-06,
      "loss": 0.7637,
      "step": 3509
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0334633756415912e-06,
      "loss": 0.7682,
      "step": 3510
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0299591652373542e-06,
      "loss": 0.7247,
      "step": 3511
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0264605832819596e-06,
      "loss": 0.6913,
      "step": 3512
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0229676319706671e-06,
      "loss": 0.263,
      "step": 3513
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0194803134952091e-06,
      "loss": 0.8053,
      "step": 3514
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0159986300437785e-06,
      "loss": 0.7776,
      "step": 3515
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0125225838010333e-06,
      "loss": 0.7555,
      "step": 3516
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0090521769480943e-06,
      "loss": 0.7819,
      "step": 3517
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0055874116625463e-06,
      "loss": 0.7693,
      "step": 3518
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0021282901184315e-06,
      "loss": 0.7742,
      "step": 3519
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.986748144862502e-07,
      "loss": 0.7924,
      "step": 3520
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.952269869329645e-07,
      "loss": 0.7864,
      "step": 3521
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.917848096219873e-07,
      "loss": 0.7332,
      "step": 3522
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.883482847131886e-07,
      "loss": 0.7537,
      "step": 3523
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.849174143628902e-07,
      "loss": 0.7821,
      "step": 3524
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.814922007238692e-07,
      "loss": 0.7385,
      "step": 3525
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.780726459453516e-07,
      "loss": 0.7909,
      "step": 3526
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.746587521730067e-07,
      "loss": 0.7572,
      "step": 3527
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.712505215489632e-07,
      "loss": 0.7001,
      "step": 3528
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.67847956211786e-07,
      "loss": 0.7482,
      "step": 3529
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.644510582964905e-07,
      "loss": 0.7876,
      "step": 3530
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.610598299345364e-07,
      "loss": 0.844,
      "step": 3531
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.576742732538235e-07,
      "loss": 0.229,
      "step": 3532
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.542943903786929e-07,
      "loss": 0.7664,
      "step": 3533
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.509201834299242e-07,
      "loss": 0.7144,
      "step": 3534
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.475516545247421e-07,
      "loss": 0.7088,
      "step": 3535
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.441888057768023e-07,
      "loss": 0.7855,
      "step": 3536
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.408316392961947e-07,
      "loss": 0.7925,
      "step": 3537
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.374801571894521e-07,
      "loss": 0.7646,
      "step": 3538
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.341343615595333e-07,
      "loss": 0.7719,
      "step": 3539
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.307942545058301e-07,
      "loss": 0.7505,
      "step": 3540
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.274598381241651e-07,
      "loss": 0.7303,
      "step": 3541
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.241311145067955e-07,
      "loss": 0.7233,
      "step": 3542
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.208080857423985e-07,
      "loss": 0.797,
      "step": 3543
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.174907539160793e-07,
      "loss": 0.7438,
      "step": 3544
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.141791211093764e-07,
      "loss": 0.7473,
      "step": 3545
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.108731894002421e-07,
      "loss": 0.7289,
      "step": 3546
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.075729608630568e-07,
      "loss": 0.8015,
      "step": 3547
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.042784375686198e-07,
      "loss": 0.6975,
      "step": 3548
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.009896215841562e-07,
      "loss": 0.7442,
      "step": 3549
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.977065149733022e-07,
      "loss": 0.7836,
      "step": 3550
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.944291197961141e-07,
      "loss": 0.2842,
      "step": 3551
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.911574381090682e-07,
      "loss": 0.7303,
      "step": 3552
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.878914719650533e-07,
      "loss": 0.721,
      "step": 3553
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.846312234133692e-07,
      "loss": 0.7494,
      "step": 3554
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.813766944997293e-07,
      "loss": 0.7543,
      "step": 3555
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.781278872662635e-07,
      "loss": 0.7647,
      "step": 3556
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.748848037515034e-07,
      "loss": 0.7199,
      "step": 3557
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.716474459903934e-07,
      "loss": 0.7998,
      "step": 3558
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.684158160142874e-07,
      "loss": 0.7652,
      "step": 3559
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.651899158509402e-07,
      "loss": 0.7436,
      "step": 3560
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.619697475245136e-07,
      "loss": 0.8102,
      "step": 3561
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.587553130555714e-07,
      "loss": 0.7831,
      "step": 3562
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.555466144610858e-07,
      "loss": 0.7366,
      "step": 3563
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.523436537544205e-07,
      "loss": 0.7525,
      "step": 3564
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.491464329453436e-07,
      "loss": 0.7976,
      "step": 3565
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.459549540400247e-07,
      "loss": 0.8282,
      "step": 3566
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.427692190410252e-07,
      "loss": 0.7539,
      "step": 3567
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.395892299473052e-07,
      "loss": 0.8213,
      "step": 3568
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.36414988754215e-07,
      "loss": 0.785,
      "step": 3569
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.332464974535071e-07,
      "loss": 0.8033,
      "step": 3570
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.300837580333199e-07,
      "loss": 0.7601,
      "step": 3571
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.269267724781793e-07,
      "loss": 0.7585,
      "step": 3572
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.237755427690097e-07,
      "loss": 0.7934,
      "step": 3573
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.206300708831172e-07,
      "loss": 0.7851,
      "step": 3574
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.174903587941963e-07,
      "loss": 0.7446,
      "step": 3575
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.14356408472331e-07,
      "loss": 0.7969,
      "step": 3576
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.112282218839851e-07,
      "loss": 0.7831,
      "step": 3577
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.081058009920095e-07,
      "loss": 0.7231,
      "step": 3578
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.049891477556326e-07,
      "loss": 0.2469,
      "step": 3579
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.018782641304701e-07,
      "loss": 0.7353,
      "step": 3580
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.987731520685127e-07,
      "loss": 0.764,
      "step": 3581
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.956738135181297e-07,
      "loss": 0.7804,
      "step": 3582
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.925802504240721e-07,
      "loss": 0.7708,
      "step": 3583
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.894924647274627e-07,
      "loss": 0.7814,
      "step": 3584
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.864104583657995e-07,
      "loss": 0.2369,
      "step": 3585
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.833342332729543e-07,
      "loss": 0.7497,
      "step": 3586
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.802637913791755e-07,
      "loss": 0.8215,
      "step": 3587
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.771991346110752e-07,
      "loss": 0.2413,
      "step": 3588
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.741402648916407e-07,
      "loss": 0.7638,
      "step": 3589
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.71087184140229e-07,
      "loss": 0.7502,
      "step": 3590
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.680398942725609e-07,
      "loss": 0.7984,
      "step": 3591
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.649983972007247e-07,
      "loss": 0.7847,
      "step": 3592
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.619626948331737e-07,
      "loss": 0.7809,
      "step": 3593
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.5893278907473e-07,
      "loss": 0.7928,
      "step": 3594
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.559086818265704e-07,
      "loss": 0.741,
      "step": 3595
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.52890374986237e-07,
      "loss": 0.6956,
      "step": 3596
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.498778704476373e-07,
      "loss": 0.7594,
      "step": 3597
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.468711701010289e-07,
      "loss": 0.7246,
      "step": 3598
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.438702758330329e-07,
      "loss": 0.7642,
      "step": 3599
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.408751895266264e-07,
      "loss": 0.7227,
      "step": 3600
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.378859130611426e-07,
      "loss": 0.7464,
      "step": 3601
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.349024483122691e-07,
      "loss": 0.7852,
      "step": 3602
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.319247971520427e-07,
      "loss": 0.8107,
      "step": 3603
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.2895296144886e-07,
      "loss": 0.7936,
      "step": 3604
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.259869430674637e-07,
      "loss": 0.7794,
      "step": 3605
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.230267438689465e-07,
      "loss": 0.8103,
      "step": 3606
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.200723657107489e-07,
      "loss": 0.7445,
      "step": 3607
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.171238104466627e-07,
      "loss": 0.7716,
      "step": 3608
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.141810799268222e-07,
      "loss": 0.7526,
      "step": 3609
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.112441759977073e-07,
      "loss": 0.7665,
      "step": 3610
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.083131005021449e-07,
      "loss": 0.2689,
      "step": 3611
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.053878552793014e-07,
      "loss": 0.7261,
      "step": 3612
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.024684421646877e-07,
      "loss": 0.825,
      "step": 3613
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.995548629901494e-07,
      "loss": 0.8156,
      "step": 3614
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.966471195838808e-07,
      "loss": 0.7744,
      "step": 3615
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.937452137704082e-07,
      "loss": 0.7738,
      "step": 3616
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.908491473705914e-07,
      "loss": 0.7791,
      "step": 3617
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.879589222016359e-07,
      "loss": 0.7151,
      "step": 3618
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.850745400770742e-07,
      "loss": 0.799,
      "step": 3619
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.821960028067754e-07,
      "loss": 0.7946,
      "step": 3620
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.793233121969422e-07,
      "loss": 0.7401,
      "step": 3621
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.764564700501075e-07,
      "loss": 0.7775,
      "step": 3622
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.735954781651333e-07,
      "loss": 0.786,
      "step": 3623
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.707403383372091e-07,
      "loss": 0.788,
      "step": 3624
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.678910523578608e-07,
      "loss": 0.801,
      "step": 3625
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.650476220149305e-07,
      "loss": 0.704,
      "step": 3626
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.62210049092592e-07,
      "loss": 0.8324,
      "step": 3627
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.593783353713445e-07,
      "loss": 0.2312,
      "step": 3628
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.56552482628009e-07,
      "loss": 0.7848,
      "step": 3629
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.537324926357258e-07,
      "loss": 0.7262,
      "step": 3630
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.509183671639607e-07,
      "loss": 0.8302,
      "step": 3631
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.481101079785002e-07,
      "loss": 0.2223,
      "step": 3632
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.453077168414456e-07,
      "loss": 0.7785,
      "step": 3633
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.425111955112196e-07,
      "loss": 0.8249,
      "step": 3634
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.397205457425626e-07,
      "loss": 0.7154,
      "step": 3635
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.369357692865263e-07,
      "loss": 0.7849,
      "step": 3636
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.34156867890482e-07,
      "loss": 0.7854,
      "step": 3637
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.313838432981101e-07,
      "loss": 0.7928,
      "step": 3638
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.286166972494079e-07,
      "loss": 0.7716,
      "step": 3639
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.258554314806808e-07,
      "loss": 0.7464,
      "step": 3640
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.231000477245441e-07,
      "loss": 0.8286,
      "step": 3641
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.20350547709927e-07,
      "loss": 0.7286,
      "step": 3642
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.176069331620615e-07,
      "loss": 0.8093,
      "step": 3643
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.148692058024886e-07,
      "loss": 0.2494,
      "step": 3644
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.121373673490549e-07,
      "loss": 0.775,
      "step": 3645
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.094114195159151e-07,
      "loss": 0.773,
      "step": 3646
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.06691364013523e-07,
      "loss": 0.7377,
      "step": 3647
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.039772025486357e-07,
      "loss": 0.7053,
      "step": 3648
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.012689368243163e-07,
      "loss": 0.8606,
      "step": 3649
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.985665685399245e-07,
      "loss": 0.776,
      "step": 3650
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.958700993911193e-07,
      "loss": 0.7674,
      "step": 3651
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.931795310698596e-07,
      "loss": 0.7579,
      "step": 3652
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.904948652644027e-07,
      "loss": 0.7679,
      "step": 3653
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.878161036593011e-07,
      "loss": 0.7682,
      "step": 3654
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.851432479354003e-07,
      "loss": 0.7542,
      "step": 3655
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.824762997698452e-07,
      "loss": 0.7863,
      "step": 3656
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.798152608360696e-07,
      "loss": 0.7515,
      "step": 3657
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.771601328038012e-07,
      "loss": 0.757,
      "step": 3658
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.745109173390551e-07,
      "loss": 0.7325,
      "step": 3659
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.718676161041447e-07,
      "loss": 0.7623,
      "step": 3660
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.692302307576669e-07,
      "loss": 0.798,
      "step": 3661
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.665987629545012e-07,
      "loss": 0.7697,
      "step": 3662
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.639732143458254e-07,
      "loss": 0.7946,
      "step": 3663
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.613535865790954e-07,
      "loss": 0.7906,
      "step": 3664
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.587398812980538e-07,
      "loss": 0.7687,
      "step": 3665
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.561321001427289e-07,
      "loss": 0.7889,
      "step": 3666
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.535302447494295e-07,
      "loss": 0.8153,
      "step": 3667
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.509343167507463e-07,
      "loss": 0.7971,
      "step": 3668
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.483443177755499e-07,
      "loss": 0.77,
      "step": 3669
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.457602494489955e-07,
      "loss": 0.7849,
      "step": 3670
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.431821133925108e-07,
      "loss": 0.8095,
      "step": 3671
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.406099112238039e-07,
      "loss": 0.7724,
      "step": 3672
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.380436445568615e-07,
      "loss": 0.8114,
      "step": 3673
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.354833150019423e-07,
      "loss": 0.7933,
      "step": 3674
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.329289241655811e-07,
      "loss": 0.7562,
      "step": 3675
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.303804736505857e-07,
      "loss": 0.8151,
      "step": 3676
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.278379650560395e-07,
      "loss": 0.7839,
      "step": 3677
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.253013999772949e-07,
      "loss": 0.737,
      "step": 3678
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.227707800059722e-07,
      "loss": 0.8015,
      "step": 3679
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.202461067299692e-07,
      "loss": 0.8092,
      "step": 3680
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.177273817334439e-07,
      "loss": 0.6891,
      "step": 3681
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.152146065968266e-07,
      "loss": 0.7846,
      "step": 3682
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.12707782896813e-07,
      "loss": 0.8373,
      "step": 3683
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.102069122063646e-07,
      "loss": 0.7802,
      "step": 3684
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.077119960947086e-07,
      "loss": 0.7512,
      "step": 3685
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.052230361273336e-07,
      "loss": 0.2381,
      "step": 3686
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.027400338659926e-07,
      "loss": 0.7706,
      "step": 3687
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.002629908687007e-07,
      "loss": 0.7454,
      "step": 3688
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.977919086897309e-07,
      "loss": 0.7955,
      "step": 3689
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.953267888796176e-07,
      "loss": 0.7404,
      "step": 3690
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.928676329851578e-07,
      "loss": 0.7699,
      "step": 3691
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.904144425494006e-07,
      "loss": 0.7449,
      "step": 3692
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.879672191116524e-07,
      "loss": 0.7496,
      "step": 3693
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.855259642074805e-07,
      "loss": 0.7357,
      "step": 3694
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.830906793687018e-07,
      "loss": 0.2523,
      "step": 3695
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.806613661233905e-07,
      "loss": 0.7363,
      "step": 3696
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.782380259958697e-07,
      "loss": 0.7339,
      "step": 3697
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.758206605067206e-07,
      "loss": 0.7691,
      "step": 3698
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.734092711727711e-07,
      "loss": 0.7154,
      "step": 3699
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.710038595070976e-07,
      "loss": 0.7911,
      "step": 3700
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.6860442701903375e-07,
      "loss": 0.7594,
      "step": 3701
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.6621097521415304e-07,
      "loss": 0.7718,
      "step": 3702
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.6382350559427857e-07,
      "loss": 0.8247,
      "step": 3703
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.614420196574809e-07,
      "loss": 0.7674,
      "step": 3704
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5906651889807697e-07,
      "loss": 0.7227,
      "step": 3705
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.566970048066266e-07,
      "loss": 0.7517,
      "step": 3706
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.543334788699316e-07,
      "loss": 0.7735,
      "step": 3707
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5197594257103907e-07,
      "loss": 0.7523,
      "step": 3708
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.496243973892389e-07,
      "loss": 0.7345,
      "step": 3709
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4727884480005534e-07,
      "loss": 0.8114,
      "step": 3710
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4493928627526105e-07,
      "loss": 0.2431,
      "step": 3711
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.426057232828607e-07,
      "loss": 0.7437,
      "step": 3712
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4027815728710177e-07,
      "loss": 0.7885,
      "step": 3713
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3795658974846277e-07,
      "loss": 0.7454,
      "step": 3714
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3564102212366734e-07,
      "loss": 0.7564,
      "step": 3715
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.333314558656654e-07,
      "loss": 0.7635,
      "step": 3716
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3102789242364553e-07,
      "loss": 0.8302,
      "step": 3717
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.2873033324303037e-07,
      "loss": 0.7436,
      "step": 3718
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.2643877976547343e-07,
      "loss": 0.7083,
      "step": 3719
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.2415323342886006e-07,
      "loss": 0.2678,
      "step": 3720
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.2187369566730416e-07,
      "loss": 0.7574,
      "step": 3721
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.196001679111561e-07,
      "loss": 0.8441,
      "step": 3722
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.17332651586988e-07,
      "loss": 0.7095,
      "step": 3723
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1507114811760174e-07,
      "loss": 0.8058,
      "step": 3724
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1281565892203e-07,
      "loss": 0.7557,
      "step": 3725
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.105661854155274e-07,
      "loss": 0.7967,
      "step": 3726
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.083227290095759e-07,
      "loss": 0.8077,
      "step": 3727
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.060852911118796e-07,
      "loss": 0.2384,
      "step": 3728
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.0385387312637194e-07,
      "loss": 0.8284,
      "step": 3729
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.016284764532008e-07,
      "loss": 0.8052,
      "step": 3730
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9940910248874254e-07,
      "loss": 0.7201,
      "step": 3731
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9719575262559093e-07,
      "loss": 0.7376,
      "step": 3732
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.949884282525629e-07,
      "loss": 0.7497,
      "step": 3733
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9278713075469045e-07,
      "loss": 0.8261,
      "step": 3734
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9059186151322534e-07,
      "loss": 0.7381,
      "step": 3735
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8840262190563915e-07,
      "loss": 0.8259,
      "step": 3736
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.862194133056185e-07,
      "loss": 0.8057,
      "step": 3737
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8404223708306434e-07,
      "loss": 0.7386,
      "step": 3738
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8187109460409375e-07,
      "loss": 0.7617,
      "step": 3739
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7970598723103936e-07,
      "loss": 0.7667,
      "step": 3740
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7754691632244323e-07,
      "loss": 0.7863,
      "step": 3741
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7539388323306056e-07,
      "loss": 0.8607,
      "step": 3742
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.732468893138641e-07,
      "loss": 0.2124,
      "step": 3743
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.711059359120273e-07,
      "loss": 0.8375,
      "step": 3744
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.689710243709399e-07,
      "loss": 0.7763,
      "step": 3745
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6684215603020044e-07,
      "loss": 0.7801,
      "step": 3746
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6471933222561375e-07,
      "loss": 0.7516,
      "step": 3747
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.626025542891909e-07,
      "loss": 0.7897,
      "step": 3748
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.604918235491495e-07,
      "loss": 0.2607,
      "step": 3749
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.583871413299167e-07,
      "loss": 0.7538,
      "step": 3750
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.5628850895212174e-07,
      "loss": 0.2323,
      "step": 3751
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.541959277325935e-07,
      "loss": 0.7924,
      "step": 3752
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.521093989843716e-07,
      "loss": 0.7574,
      "step": 3753
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.500289240166932e-07,
      "loss": 0.6998,
      "step": 3754
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.479545041349963e-07,
      "loss": 0.8158,
      "step": 3755
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.458861406409253e-07,
      "loss": 0.7024,
      "step": 3756
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.438238348323164e-07,
      "loss": 0.7674,
      "step": 3757
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.417675880032101e-07,
      "loss": 0.7942,
      "step": 3758
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.397174014438431e-07,
      "loss": 0.7592,
      "step": 3759
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3767327644064984e-07,
      "loss": 0.8051,
      "step": 3760
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.356352142762631e-07,
      "loss": 0.7438,
      "step": 3761
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3360321622950664e-07,
      "loss": 0.7653,
      "step": 3762
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.31577283575405e-07,
      "loss": 0.6962,
      "step": 3763
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2955741758517345e-07,
      "loss": 0.7507,
      "step": 3764
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2754361952621936e-07,
      "loss": 0.8382,
      "step": 3765
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.255358906621464e-07,
      "loss": 0.8096,
      "step": 3766
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2353423225274683e-07,
      "loss": 0.7345,
      "step": 3767
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2153864555400705e-07,
      "loss": 0.7365,
      "step": 3768
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.195491318180988e-07,
      "loss": 0.7985,
      "step": 3769
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1756569229338897e-07,
      "loss": 0.7582,
      "step": 3770
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1558832822442874e-07,
      "loss": 0.286,
      "step": 3771
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.13617040851959e-07,
      "loss": 0.7645,
      "step": 3772
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.116518314129058e-07,
      "loss": 0.2555,
      "step": 3773
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.09692701140385e-07,
      "loss": 0.8281,
      "step": 3774
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.077396512636954e-07,
      "loss": 0.8008,
      "step": 3775
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.057926830083191e-07,
      "loss": 0.7544,
      "step": 3776
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0385179759592767e-07,
      "loss": 0.7743,
      "step": 3777
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.019169962443691e-07,
      "loss": 0.7434,
      "step": 3778
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.999882801676779e-07,
      "loss": 0.7711,
      "step": 3779
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.9806565057606707e-07,
      "loss": 0.7795,
      "step": 3780
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.9614910867593717e-07,
      "loss": 0.7049,
      "step": 3781
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.942386556698618e-07,
      "loss": 0.7883,
      "step": 3782
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.9233429275659643e-07,
      "loss": 0.2391,
      "step": 3783
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.904360211310764e-07,
      "loss": 0.7909,
      "step": 3784
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.885438419844133e-07,
      "loss": 0.7301,
      "step": 3785
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8665775650389637e-07,
      "loss": 0.707,
      "step": 3786
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.847777658729911e-07,
      "loss": 0.741,
      "step": 3787
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8290387127133947e-07,
      "loss": 0.7303,
      "step": 3788
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8103607387475753e-07,
      "loss": 0.7606,
      "step": 3789
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7917437485523447e-07,
      "loss": 0.7082,
      "step": 3790
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.77318775380937e-07,
      "loss": 0.7563,
      "step": 3791
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7546927661619925e-07,
      "loss": 0.7652,
      "step": 3792
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7362587972153075e-07,
      "loss": 0.8015,
      "step": 3793
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7178858585360955e-07,
      "loss": 0.7935,
      "step": 3794
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6995739616528905e-07,
      "loss": 0.2318,
      "step": 3795
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6813231180558786e-07,
      "loss": 0.761,
      "step": 3796
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6631333391969437e-07,
      "loss": 0.8106,
      "step": 3797
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6450046364896785e-07,
      "loss": 0.7994,
      "step": 3798
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6269370213093283e-07,
      "loss": 0.8084,
      "step": 3799
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6089305049928013e-07,
      "loss": 0.2425,
      "step": 3800
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5909850988386943e-07,
      "loss": 0.7687,
      "step": 3801
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.573100814107254e-07,
      "loss": 0.7997,
      "step": 3802
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.55527766202035e-07,
      "loss": 0.7293,
      "step": 3803
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.537515653761502e-07,
      "loss": 0.7947,
      "step": 3804
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.519814800475906e-07,
      "loss": 0.714,
      "step": 3805
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.502175113270322e-07,
      "loss": 0.7888,
      "step": 3806
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.484596603213163e-07,
      "loss": 0.2681,
      "step": 3807
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4670792813344725e-07,
      "loss": 0.7393,
      "step": 3808
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4496231586258466e-07,
      "loss": 0.7295,
      "step": 3809
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4322282460405356e-07,
      "loss": 0.7472,
      "step": 3810
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.414894554493341e-07,
      "loss": 0.7665,
      "step": 3811
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.397622094860674e-07,
      "loss": 0.7284,
      "step": 3812
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3804108779805325e-07,
      "loss": 0.8308,
      "step": 3813
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3632609146524431e-07,
      "loss": 0.7361,
      "step": 3814
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3461722156375654e-07,
      "loss": 0.7837,
      "step": 3815
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.329144791658544e-07,
      "loss": 0.7338,
      "step": 3816
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3121786533996328e-07,
      "loss": 0.7655,
      "step": 3817
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.295273811506582e-07,
      "loss": 0.2543,
      "step": 3818
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2784302765867294e-07,
      "loss": 0.7607,
      "step": 3819
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2616480592089096e-07,
      "loss": 0.8065,
      "step": 3820
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.244927169903488e-07,
      "loss": 0.8038,
      "step": 3821
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.228267619162361e-07,
      "loss": 0.7866,
      "step": 3822
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.211669417438933e-07,
      "loss": 0.8114,
      "step": 3823
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1951325751481068e-07,
      "loss": 0.7636,
      "step": 3824
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1786571026662705e-07,
      "loss": 0.8094,
      "step": 3825
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1622430103313442e-07,
      "loss": 0.7297,
      "step": 3826
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1458903084427108e-07,
      "loss": 0.7236,
      "step": 3827
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.129599007261196e-07,
      "loss": 0.785,
      "step": 3828
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1133691170091786e-07,
      "loss": 0.7474,
      "step": 3829
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0972006478704342e-07,
      "loss": 0.7712,
      "step": 3830
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0810936099902368e-07,
      "loss": 0.7657,
      "step": 3831
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0650480134752793e-07,
      "loss": 0.7697,
      "step": 3832
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0490638683937415e-07,
      "loss": 0.2719,
      "step": 3833
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0331411847752226e-07,
      "loss": 0.7634,
      "step": 3834
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0172799726107528e-07,
      "loss": 0.7621,
      "step": 3835
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0014802418528156e-07,
      "loss": 0.7248,
      "step": 3836
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9857420024152808e-07,
      "loss": 0.7991,
      "step": 3837
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9700652641734485e-07,
      "loss": 0.7797,
      "step": 3838
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9544500369640396e-07,
      "loss": 0.7566,
      "step": 3839
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9388963305851827e-07,
      "loss": 0.233,
      "step": 3840
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.923404154796371e-07,
      "loss": 0.7817,
      "step": 3841
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9079735193185288e-07,
      "loss": 0.826,
      "step": 3842
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8926044338339332e-07,
      "loss": 0.7795,
      "step": 3843
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.877296907986259e-07,
      "loss": 0.7232,
      "step": 3844
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.862050951380545e-07,
      "loss": 0.7385,
      "step": 3845
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8468665735832058e-07,
      "loss": 0.8147,
      "step": 3846
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8317437841220087e-07,
      "loss": 0.7574,
      "step": 3847
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8166825924860853e-07,
      "loss": 0.729,
      "step": 3848
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8016830081259095e-07,
      "loss": 0.7028,
      "step": 3849
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.786745040453286e-07,
      "loss": 0.7331,
      "step": 3850
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7718686988413946e-07,
      "loss": 0.7428,
      "step": 3851
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.757053992624691e-07,
      "loss": 0.2644,
      "step": 3852
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.742300931099017e-07,
      "loss": 0.7917,
      "step": 3853
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7276095235215006e-07,
      "loss": 0.7825,
      "step": 3854
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7129797791105686e-07,
      "loss": 0.7665,
      "step": 3855
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6984117070459882e-07,
      "loss": 0.7648,
      "step": 3856
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6839053164688145e-07,
      "loss": 0.2645,
      "step": 3857
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6694606164813998e-07,
      "loss": 0.7706,
      "step": 3858
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6550776161473825e-07,
      "loss": 0.7664,
      "step": 3859
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.640756324491699e-07,
      "loss": 0.8297,
      "step": 3860
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6264967505005613e-07,
      "loss": 0.8059,
      "step": 3861
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6122989031214453e-07,
      "loss": 0.7538,
      "step": 3862
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.5981627912630914e-07,
      "loss": 0.7979,
      "step": 3863
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.5840884237955378e-07,
      "loss": 0.7688,
      "step": 3864
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5700758095500314e-07,
      "loss": 0.7245,
      "step": 3865
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.556124957319105e-07,
      "loss": 0.7666,
      "step": 3866
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5422358758565348e-07,
      "loss": 0.7403,
      "step": 3867
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5284085738773158e-07,
      "loss": 0.7924,
      "step": 3868
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5146430600576965e-07,
      "loss": 0.7378,
      "step": 3869
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.500939343035146e-07,
      "loss": 0.7945,
      "step": 3870
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4872974314083744e-07,
      "loss": 0.7544,
      "step": 3871
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.47371733373729e-07,
      "loss": 0.2867,
      "step": 3872
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4601990585430214e-07,
      "loss": 0.7639,
      "step": 3873
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4467426143079168e-07,
      "loss": 0.815,
      "step": 3874
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4333480094754992e-07,
      "loss": 0.7682,
      "step": 3875
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.420015252450524e-07,
      "loss": 0.7809,
      "step": 3876
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.40674435159891e-07,
      "loss": 0.7563,
      "step": 3877
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3935353152477848e-07,
      "loss": 0.771,
      "step": 3878
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3803881516854412e-07,
      "loss": 0.7442,
      "step": 3879
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3673028691613576e-07,
      "loss": 0.7355,
      "step": 3880
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3542794758861887e-07,
      "loss": 0.7342,
      "step": 3881
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.341317980031731e-07,
      "loss": 0.2416,
      "step": 3882
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3284183897309788e-07,
      "loss": 0.7431,
      "step": 3883
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.315580713078046e-07,
      "loss": 0.2464,
      "step": 3884
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3028049581282231e-07,
      "loss": 0.7697,
      "step": 3885
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2900911328979416e-07,
      "loss": 0.7531,
      "step": 3886
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2774392453647644e-07,
      "loss": 0.77,
      "step": 3887
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.264849303467397e-07,
      "loss": 0.7408,
      "step": 3888
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2523213151056758e-07,
      "loss": 0.745,
      "step": 3889
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2398552881405456e-07,
      "loss": 0.758,
      "step": 3890
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2274512303941165e-07,
      "loss": 0.8167,
      "step": 3891
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2151091496495736e-07,
      "loss": 0.7589,
      "step": 3892
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2028290536512223e-07,
      "loss": 0.7183,
      "step": 3893
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.190610950104476e-07,
      "loss": 0.8091,
      "step": 3894
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1784548466758583e-07,
      "loss": 0.7486,
      "step": 3895
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1663607509929898e-07,
      "loss": 0.7587,
      "step": 3896
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1543286706445556e-07,
      "loss": 0.2453,
      "step": 3897
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.142358613180361e-07,
      "loss": 0.8064,
      "step": 3898
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.130450586111298e-07,
      "loss": 0.7351,
      "step": 3899
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1186045969092895e-07,
      "loss": 0.7657,
      "step": 3900
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1068206530073678e-07,
      "loss": 0.7897,
      "step": 3901
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.09509876179964e-07,
      "loss": 0.7807,
      "step": 3902
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0834389306412674e-07,
      "loss": 0.7684,
      "step": 3903
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0718411668484308e-07,
      "loss": 0.7191,
      "step": 3904
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0603054776984534e-07,
      "loss": 0.793,
      "step": 3905
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0488318704296118e-07,
      "loss": 0.8231,
      "step": 3906
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0374203522413029e-07,
      "loss": 0.8065,
      "step": 3907
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0260709302939098e-07,
      "loss": 0.7515,
      "step": 3908
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0147836117088916e-07,
      "loss": 0.7493,
      "step": 3909
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0035584035687273e-07,
      "loss": 0.8021,
      "step": 3910
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.92395312916905e-08,
      "loss": 0.7449,
      "step": 3911
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.81294346757955e-08,
      "loss": 0.7236,
      "step": 3912
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.702555120574386e-08,
      "loss": 0.7287,
      "step": 3913
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.59278815741893e-08,
      "loss": 0.7712,
      "step": 3914
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.483642646988978e-08,
      "loss": 0.2303,
      "step": 3915
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.375118657770188e-08,
      "loss": 0.7287,
      "step": 3916
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.267216257858535e-08,
      "loss": 0.7621,
      "step": 3917
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.159935514959417e-08,
      "loss": 0.816,
      "step": 3918
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.053276496388763e-08,
      "loss": 0.7269,
      "step": 3919
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.947239269072039e-08,
      "loss": 0.7711,
      "step": 3920
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.841823899544577e-08,
      "loss": 0.2623,
      "step": 3921
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.737030453951578e-08,
      "loss": 0.7262,
      "step": 3922
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.632858998047999e-08,
      "loss": 0.2301,
      "step": 3923
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.529309597198554e-08,
      "loss": 0.7746,
      "step": 3924
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.426382316377379e-08,
      "loss": 0.8177,
      "step": 3925
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.324077220168813e-08,
      "loss": 0.8322,
      "step": 3926
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.222394372766173e-08,
      "loss": 0.7554,
      "step": 3927
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.121333837972534e-08,
      "loss": 0.2828,
      "step": 3928
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.020895679200724e-08,
      "loss": 0.7929,
      "step": 3929
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.921079959472888e-08,
      "loss": 0.7173,
      "step": 3930
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.821886741420482e-08,
      "loss": 0.7295,
      "step": 3931
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.723316087284494e-08,
      "loss": 0.2386,
      "step": 3932
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.625368058915228e-08,
      "loss": 0.7269,
      "step": 3933
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.5280427177723e-08,
      "loss": 0.7906,
      "step": 3934
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.431340124924635e-08,
      "loss": 0.7703,
      "step": 3935
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.335260341050476e-08,
      "loss": 0.7644,
      "step": 3936
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.239803426437042e-08,
      "loss": 0.7695,
      "step": 3937
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.144969440981087e-08,
      "loss": 0.6996,
      "step": 3938
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.050758444188122e-08,
      "loss": 0.7723,
      "step": 3939
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.95717049517286e-08,
      "loss": 0.7737,
      "step": 3940
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.864205652659328e-08,
      "loss": 0.8176,
      "step": 3941
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.771863974980086e-08,
      "loss": 0.8277,
      "step": 3942
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.680145520077119e-08,
      "loss": 0.7796,
      "step": 3943
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.589050345501058e-08,
      "loss": 0.7444,
      "step": 3944
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.498578508411734e-08,
      "loss": 0.8117,
      "step": 3945
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.408730065577517e-08,
      "loss": 0.7377,
      "step": 3946
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.319505073375976e-08,
      "loss": 0.2758,
      "step": 3947
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.230903587793103e-08,
      "loss": 0.7235,
      "step": 3948
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.142925664423872e-08,
      "loss": 0.7625,
      "step": 3949
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.055571358472124e-08,
      "loss": 0.8351,
      "step": 3950
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.968840724750013e-08,
      "loss": 0.8205,
      "step": 3951
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.8827338176787876e-08,
      "loss": 0.8035,
      "step": 3952
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.797250691287892e-08,
      "loss": 0.7734,
      "step": 3953
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.7123913992157555e-08,
      "loss": 0.7744,
      "step": 3954
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.628155994709117e-08,
      "loss": 0.8097,
      "step": 3955
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.5445445306233636e-08,
      "loss": 0.776,
      "step": 3956
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.4615570594223066e-08,
      "loss": 0.7949,
      "step": 3957
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.379193633178403e-08,
      "loss": 0.2435,
      "step": 3958
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.2974543035722026e-08,
      "loss": 0.2472,
      "step": 3959
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.2163391218929e-08,
      "loss": 0.7953,
      "step": 3960
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.135848139038002e-08,
      "loss": 0.7491,
      "step": 3961
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.0559814055133326e-08,
      "loss": 0.7358,
      "step": 3962
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.976738971433026e-08,
      "loss": 0.7412,
      "step": 3963
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.898120886519642e-08,
      "loss": 0.8133,
      "step": 3964
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.820127200103497e-08,
      "loss": 0.7503,
      "step": 3965
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.742757961123778e-08,
      "loss": 0.8271,
      "step": 3966
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.6660132181272076e-08,
      "loss": 0.7787,
      "step": 3967
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5898930192691536e-08,
      "loss": 0.8409,
      "step": 3968
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5143974123129655e-08,
      "loss": 0.7486,
      "step": 3969
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.439526444629749e-08,
      "loss": 0.8124,
      "step": 3970
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.365280163199148e-08,
      "loss": 0.761,
      "step": 3971
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.291658614608563e-08,
      "loss": 0.7331,
      "step": 3972
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.218661845053484e-08,
      "loss": 0.8032,
      "step": 3973
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.146289900337275e-08,
      "loss": 0.7336,
      "step": 3974
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.074542825871275e-08,
      "loss": 0.8071,
      "step": 3975
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.003420666674918e-08,
      "loss": 0.791,
      "step": 3976
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.9329234673751714e-08,
      "loss": 0.7858,
      "step": 3977
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.8630512722072076e-08,
      "loss": 0.7698,
      "step": 3978
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.7938041250137335e-08,
      "loss": 0.7745,
      "step": 3979
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.725182069245437e-08,
      "loss": 0.7941,
      "step": 3980
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.6571851479607624e-08,
      "loss": 0.7736,
      "step": 3981
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.5898134038259146e-08,
      "loss": 0.7497,
      "step": 3982
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.523066879114634e-08,
      "loss": 0.7752,
      "step": 3983
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.45694561570864e-08,
      "loss": 0.8007,
      "step": 3984
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.3914496550969674e-08,
      "loss": 0.769,
      "step": 3985
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.326579038376743e-08,
      "loss": 0.8,
      "step": 3986
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.262333806252294e-08,
      "loss": 0.7696,
      "step": 3987
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.198713999035707e-08,
      "loss": 0.7815,
      "step": 3988
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.1357196566467184e-08,
      "loss": 0.7545,
      "step": 3989
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.073350818612486e-08,
      "loss": 0.7689,
      "step": 3990
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.0116075240675946e-08,
      "loss": 0.7722,
      "step": 3991
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.9504898117542758e-08,
      "loss": 0.7238,
      "step": 3992
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.889997720022297e-08,
      "loss": 0.7234,
      "step": 3993
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.8301312868285192e-08,
      "loss": 0.7835,
      "step": 3994
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.7708905497377812e-08,
      "loss": 0.7002,
      "step": 3995
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.7122755459216833e-08,
      "loss": 0.7291,
      "step": 3996
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.654286312159582e-08,
      "loss": 0.7811,
      "step": 3997
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.596922884838149e-08,
      "loss": 0.7625,
      "step": 3998
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.540185299951259e-08,
      "loss": 0.7569,
      "step": 3999
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.4840735931002115e-08,
      "loss": 0.7448,
      "step": 4000
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.428587799493509e-08,
      "loss": 0.7761,
      "step": 4001
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.3737279539469695e-08,
      "loss": 0.7931,
      "step": 4002
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.3194940908835007e-08,
      "loss": 0.8273,
      "step": 4003
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2658862443333263e-08,
      "loss": 0.7481,
      "step": 4004
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2129044479339833e-08,
      "loss": 0.7805,
      "step": 4005
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1605487349299904e-08,
      "loss": 0.7308,
      "step": 4006
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1088191381731794e-08,
      "loss": 0.7827,
      "step": 4007
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.0577156901222528e-08,
      "loss": 0.7829,
      "step": 4008
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.0072384228432275e-08,
      "loss": 0.7505,
      "step": 4009
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9573873680093225e-08,
      "loss": 0.8073,
      "step": 4010
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.908162556900628e-08,
      "loss": 0.7814,
      "step": 4011
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8595640204041033e-08,
      "loss": 0.7851,
      "step": 4012
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8115917890142443e-08,
      "loss": 0.7848,
      "step": 4013
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.764245892832195e-08,
      "loss": 0.7196,
      "step": 4014
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.717526361566191e-08,
      "loss": 0.7519,
      "step": 4015
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.671433224531338e-08,
      "loss": 0.8021,
      "step": 4016
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6259665106498346e-08,
      "loss": 0.7882,
      "step": 4017
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5811262484507484e-08,
      "loss": 0.8237,
      "step": 4018
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5369124660701283e-08,
      "loss": 0.2673,
      "step": 4019
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.493325191250894e-08,
      "loss": 0.721,
      "step": 4020
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4503644513428339e-08,
      "loss": 0.721,
      "step": 4021
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4080302733026074e-08,
      "loss": 0.7677,
      "step": 4022
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.3663226836936327e-08,
      "loss": 0.783,
      "step": 4023
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.325241708686309e-08,
      "loss": 0.755,
      "step": 4024
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2847873740577944e-08,
      "loss": 0.7383,
      "step": 4025
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2449597051920059e-08,
      "loss": 0.8101,
      "step": 4026
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.205758727079842e-08,
      "loss": 0.7046,
      "step": 4027
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.167184464318738e-08,
      "loss": 0.7757,
      "step": 4028
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1292369411127768e-08,
      "loss": 0.8205,
      "step": 4029
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0919161812732449e-08,
      "loss": 0.7653,
      "step": 4030
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0552222082177433e-08,
      "loss": 0.2472,
      "step": 4031
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0191550449706322e-08,
      "loss": 0.7327,
      "step": 4032
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.837147141632531e-09,
      "loss": 0.7529,
      "step": 4033
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.489012380332618e-09,
      "loss": 0.8292,
      "step": 4034
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.147146384250739e-09,
      "loss": 0.7944,
      "step": 4035
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.811549367899741e-09,
      "loss": 0.7304,
      "step": 4036
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.482221541856739e-09,
      "loss": 0.8201,
      "step": 4037
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.159163112765322e-09,
      "loss": 0.2422,
      "step": 4038
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.842374283336673e-09,
      "loss": 0.7528,
      "step": 4039
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.531855252346232e-09,
      "loss": 0.787,
      "step": 4040
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.227606214635918e-09,
      "loss": 0.255,
      "step": 4041
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.929627361113023e-09,
      "loss": 0.7628,
      "step": 4042
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.637918878751315e-09,
      "loss": 0.7733,
      "step": 4043
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.352480950589934e-09,
      "loss": 0.7684,
      "step": 4044
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.07331375573228e-09,
      "loss": 0.7797,
      "step": 4045
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.80041746934934e-09,
      "loss": 0.7755,
      "step": 4046
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.5337922626752526e-09,
      "loss": 0.7879,
      "step": 4047
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.273438303009526e-09,
      "loss": 0.7543,
      "step": 4048
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.019355753717036e-09,
      "loss": 0.7399,
      "step": 4049
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.771544774228032e-09,
      "loss": 0.8185,
      "step": 4050
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5300055200359076e-09,
      "loss": 0.759,
      "step": 4051
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.294738142701649e-09,
      "loss": 0.806,
      "step": 4052
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.065742789846061e-09,
      "loss": 0.7533,
      "step": 4053
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.843019605159759e-09,
      "loss": 0.2589,
      "step": 4054
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.6265687283942862e-09,
      "loss": 0.7167,
      "step": 4055
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.416390295366556e-09,
      "loss": 0.7646,
      "step": 4056
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.2124844379577415e-09,
      "loss": 0.7575,
      "step": 4057
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.0148512841132736e-09,
      "loss": 0.8074,
      "step": 4058
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.8234909578417347e-09,
      "loss": 0.7257,
      "step": 4059
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.638403579217075e-09,
      "loss": 0.7371,
      "step": 4060
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.459589264375284e-09,
      "loss": 0.7947,
      "step": 4061
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2870481255188314e-09,
      "loss": 0.7554,
      "step": 4062
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.120780270912226e-09,
      "loss": 0.7456,
      "step": 4063
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.960785804883125e-09,
      "loss": 0.7831,
      "step": 4064
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.807064827823446e-09,
      "loss": 0.7511,
      "step": 4065
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.6596174361893647e-09,
      "loss": 0.7891,
      "step": 4066
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.518443722500207e-09,
      "loss": 0.8167,
      "step": 4067
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.383543775337337e-09,
      "loss": 0.8062,
      "step": 4068
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.2549176793485995e-09,
      "loss": 0.7535,
      "step": 4069
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.132565515241657e-09,
      "loss": 0.7855,
      "step": 4070
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.016487359789542e-09,
      "loss": 0.7481,
      "step": 4071
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.066832858273255e-10,
      "loss": 0.78,
      "step": 4072
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.031533622554488e-10,
      "loss": 0.8015,
      "step": 4073
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.058976540352813e-10,
      "loss": 0.7859,
      "step": 4074
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.14916222192452e-10,
      "loss": 0.7647,
      "step": 4075
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.302091238135187e-10,
      "loss": 0.2676,
      "step": 4076
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5177641205262915e-10,
      "loss": 0.7592,
      "step": 4077
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.79618136121529e-10,
      "loss": 0.7511,
      "step": 4078
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.137343412984439e-10,
      "loss": 0.2615,
      "step": 4079
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5412506892474874e-10,
      "loss": 0.7497,
      "step": 4080
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.007903564027469e-10,
      "loss": 0.6978,
      "step": 4081
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.537302371967808e-10,
      "loss": 0.742,
      "step": 4082
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1294474083878293e-10,
      "loss": 0.7423,
      "step": 4083
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.843389291828374e-11,
      "loss": 0.7218,
      "step": 4084
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.0197715091293563e-11,
      "loss": 0.8287,
      "step": 4085
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.823622507364121e-11,
      "loss": 0.7523,
      "step": 4086
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.2549436647635305e-11,
      "loss": 0.7309,
      "step": 4087
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.1373596542927375e-12,
      "loss": 0.5558,
      "step": 4088
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0,
      "loss": 0.4724,
      "step": 4089
    },
    {
      "epoch": 1.0,
      "step": 4089,
      "total_flos": 1782928753164288.0,
      "train_loss": 0.7879507731519663,
      "train_runtime": 207131.3046,
      "train_samples_per_second": 2.527,
      "train_steps_per_second": 0.02
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 4089,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 50000,
  "total_flos": 1782928753164288.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}