{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 20.0,
  "eval_steps": 500,
  "global_step": 3280,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 2e-05,
      "loss": 2.6798,
      "step": 1
    },
    {
      "epoch": 0.01,
      "learning_rate": 4e-05,
      "loss": 1.789,
      "step": 2
    },
    {
      "epoch": 0.02,
      "learning_rate": 6e-05,
      "loss": 2.2331,
      "step": 3
    },
    {
      "epoch": 0.02,
      "learning_rate": 8e-05,
      "loss": 1.6942,
      "step": 4
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001,
      "loss": 1.9346,
      "step": 5
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.996946564885497e-05,
      "loss": 1.3488,
      "step": 6
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.993893129770992e-05,
      "loss": 1.6845,
      "step": 7
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.990839694656488e-05,
      "loss": 1.3028,
      "step": 8
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.987786259541985e-05,
      "loss": 1.2087,
      "step": 9
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.984732824427481e-05,
      "loss": 1.2245,
      "step": 10
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.981679389312978e-05,
      "loss": 1.0479,
      "step": 11
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.978625954198474e-05,
      "loss": 1.245,
      "step": 12
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97557251908397e-05,
      "loss": 0.8715,
      "step": 13
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.972519083969466e-05,
      "loss": 1.1261,
      "step": 14
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.969465648854962e-05,
      "loss": 0.9244,
      "step": 15
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.966412213740459e-05,
      "loss": 0.8943,
      "step": 16
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.963358778625954e-05,
      "loss": 1.0608,
      "step": 17
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.960305343511452e-05,
      "loss": 0.8824,
      "step": 18
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.957251908396947e-05,
      "loss": 0.9156,
      "step": 19
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.954198473282443e-05,
      "loss": 0.78,
      "step": 20
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.95114503816794e-05,
      "loss": 0.8091,
      "step": 21
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.948091603053436e-05,
      "loss": 0.6894,
      "step": 22
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.945038167938931e-05,
      "loss": 0.8108,
      "step": 23
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.941984732824428e-05,
      "loss": 0.6459,
      "step": 24
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.938931297709924e-05,
      "loss": 0.7263,
      "step": 25
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.93587786259542e-05,
      "loss": 0.6389,
      "step": 26
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.932824427480917e-05,
      "loss": 0.7217,
      "step": 27
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.929770992366413e-05,
      "loss": 0.5818,
      "step": 28
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.926717557251909e-05,
      "loss": 0.5705,
      "step": 29
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.923664122137405e-05,
      "loss": 0.6901,
      "step": 30
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.920610687022901e-05,
      "loss": 0.7526,
      "step": 31
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.917557251908397e-05,
      "loss": 0.4792,
      "step": 32
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.914503816793893e-05,
      "loss": 0.8253,
      "step": 33
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.911450381679391e-05,
      "loss": 0.54,
      "step": 34
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.908396946564886e-05,
      "loss": 0.8134,
      "step": 35
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.905343511450382e-05,
      "loss": 0.6169,
      "step": 36
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.902290076335879e-05,
      "loss": 0.7937,
      "step": 37
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.899236641221374e-05,
      "loss": 0.793,
      "step": 38
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.89618320610687e-05,
      "loss": 0.6315,
      "step": 39
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.893129770992367e-05,
      "loss": 0.6736,
      "step": 40
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.890076335877863e-05,
      "loss": 0.4964,
      "step": 41
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.88702290076336e-05,
      "loss": 0.7035,
      "step": 42
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.883969465648856e-05,
      "loss": 0.6082,
      "step": 43
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.880916030534351e-05,
      "loss": 0.8418,
      "step": 44
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.877862595419848e-05,
      "loss": 0.7151,
      "step": 45
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.874809160305344e-05,
      "loss": 0.6075,
      "step": 46
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.87175572519084e-05,
      "loss": 0.7699,
      "step": 47
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.868702290076336e-05,
      "loss": 0.5344,
      "step": 48
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.865648854961832e-05,
      "loss": 0.5892,
      "step": 49
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.862595419847329e-05,
      "loss": 0.7846,
      "step": 50
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.859541984732825e-05,
      "loss": 0.717,
      "step": 51
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.856488549618321e-05,
      "loss": 0.6906,
      "step": 52
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.853435114503818e-05,
      "loss": 0.9293,
      "step": 53
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.850381679389313e-05,
      "loss": 0.7572,
      "step": 54
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.84732824427481e-05,
      "loss": 0.7748,
      "step": 55
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.844274809160306e-05,
      "loss": 0.8314,
      "step": 56
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.841221374045801e-05,
      "loss": 1.0392,
      "step": 57
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.838167938931299e-05,
      "loss": 0.7564,
      "step": 58
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.835114503816795e-05,
      "loss": 0.8133,
      "step": 59
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.83206106870229e-05,
      "loss": 0.6421,
      "step": 60
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.829007633587787e-05,
      "loss": 0.8475,
      "step": 61
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.825954198473283e-05,
      "loss": 0.4835,
      "step": 62
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.822900763358778e-05,
      "loss": 0.6597,
      "step": 63
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.819847328244275e-05,
      "loss": 0.5103,
      "step": 64
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.816793893129771e-05,
      "loss": 0.5987,
      "step": 65
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.813740458015268e-05,
      "loss": 0.8821,
      "step": 66
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.810687022900764e-05,
      "loss": 0.6065,
      "step": 67
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.80763358778626e-05,
      "loss": 0.972,
      "step": 68
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.804580152671756e-05,
      "loss": 0.5441,
      "step": 69
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.801526717557252e-05,
      "loss": 0.6741,
      "step": 70
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.798473282442749e-05,
      "loss": 0.6272,
      "step": 71
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.795419847328244e-05,
      "loss": 0.479,
      "step": 72
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.79236641221374e-05,
      "loss": 0.7349,
      "step": 73
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.789312977099238e-05,
      "loss": 0.5815,
      "step": 74
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.786259541984733e-05,
      "loss": 0.6201,
      "step": 75
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.78320610687023e-05,
      "loss": 0.6399,
      "step": 76
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.780152671755726e-05,
      "loss": 0.6588,
      "step": 77
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.777099236641222e-05,
      "loss": 0.5014,
      "step": 78
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.774045801526717e-05,
      "loss": 0.5912,
      "step": 79
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.770992366412214e-05,
      "loss": 0.5236,
      "step": 80
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.76793893129771e-05,
      "loss": 0.6688,
      "step": 81
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.764885496183207e-05,
      "loss": 0.4234,
      "step": 82
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.761832061068703e-05,
      "loss": 0.8562,
      "step": 83
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.7587786259542e-05,
      "loss": 0.6259,
      "step": 84
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.755725190839695e-05,
      "loss": 0.5996,
      "step": 85
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.752671755725191e-05,
      "loss": 0.7328,
      "step": 86
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.749618320610688e-05,
      "loss": 0.6371,
      "step": 87
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.746564885496183e-05,
      "loss": 0.768,
      "step": 88
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.743511450381679e-05,
      "loss": 0.5099,
      "step": 89
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.740458015267177e-05,
      "loss": 0.6689,
      "step": 90
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.737404580152672e-05,
      "loss": 0.4946,
      "step": 91
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.734351145038169e-05,
      "loss": 0.5682,
      "step": 92
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.731297709923665e-05,
      "loss": 0.6664,
      "step": 93
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.72824427480916e-05,
      "loss": 0.542,
      "step": 94
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.725190839694657e-05,
      "loss": 0.723,
      "step": 95
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.722137404580153e-05,
      "loss": 0.6148,
      "step": 96
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.719083969465648e-05,
      "loss": 0.6123,
      "step": 97
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.716030534351146e-05,
      "loss": 0.5557,
      "step": 98
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.712977099236642e-05,
      "loss": 0.5546,
      "step": 99
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.709923664122138e-05,
      "loss": 0.9345,
      "step": 100
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.706870229007634e-05,
      "loss": 0.63,
      "step": 101
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.70381679389313e-05,
      "loss": 0.8046,
      "step": 102
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.700763358778626e-05,
      "loss": 0.7358,
      "step": 103
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.697709923664122e-05,
      "loss": 0.5665,
      "step": 104
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.694656488549618e-05,
      "loss": 0.9,
      "step": 105
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.691603053435115e-05,
      "loss": 0.6869,
      "step": 106
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.688549618320611e-05,
      "loss": 0.6947,
      "step": 107
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.685496183206108e-05,
      "loss": 0.6668,
      "step": 108
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.682442748091604e-05,
      "loss": 0.6632,
      "step": 109
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.679389312977099e-05,
      "loss": 0.5953,
      "step": 110
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.676335877862596e-05,
      "loss": 0.6907,
      "step": 111
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.673282442748092e-05,
      "loss": 0.995,
      "step": 112
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.670229007633587e-05,
      "loss": 0.4654,
      "step": 113
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.667175572519085e-05,
      "loss": 0.5927,
      "step": 114
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.664122137404582e-05,
      "loss": 0.618,
      "step": 115
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.661068702290077e-05,
      "loss": 0.5931,
      "step": 116
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.658015267175573e-05,
      "loss": 0.8848,
      "step": 117
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.65496183206107e-05,
      "loss": 0.7746,
      "step": 118
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.651908396946565e-05,
      "loss": 0.5337,
      "step": 119
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.648854961832061e-05,
      "loss": 0.8552,
      "step": 120
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.645801526717558e-05,
      "loss": 0.585,
      "step": 121
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.642748091603054e-05,
      "loss": 0.7983,
      "step": 122
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.63969465648855e-05,
      "loss": 0.7295,
      "step": 123
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.636641221374047e-05,
      "loss": 0.6791,
      "step": 124
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.633587786259542e-05,
      "loss": 0.7956,
      "step": 125
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.630534351145038e-05,
      "loss": 0.4942,
      "step": 126
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.627480916030535e-05,
      "loss": 0.8309,
      "step": 127
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.62442748091603e-05,
      "loss": 0.6278,
      "step": 128
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.621374045801526e-05,
      "loss": 0.5997,
      "step": 129
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.618320610687024e-05,
      "loss": 0.6082,
      "step": 130
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.61526717557252e-05,
      "loss": 0.5213,
      "step": 131
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.612213740458016e-05,
      "loss": 0.5972,
      "step": 132
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.609160305343512e-05,
      "loss": 0.6867,
      "step": 133
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.606106870229007e-05,
      "loss": 0.6782,
      "step": 134
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.603053435114504e-05,
      "loss": 0.5501,
      "step": 135
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.6e-05,
      "loss": 0.706,
      "step": 136
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.596946564885497e-05,
      "loss": 0.7133,
      "step": 137
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.593893129770993e-05,
      "loss": 0.7482,
      "step": 138
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.59083969465649e-05,
      "loss": 0.6487,
      "step": 139
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.587786259541986e-05,
      "loss": 0.5677,
      "step": 140
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.584732824427481e-05,
      "loss": 0.6623,
      "step": 141
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.581679389312978e-05,
      "loss": 0.7414,
      "step": 142
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.578625954198474e-05,
      "loss": 0.566,
      "step": 143
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.575572519083969e-05,
      "loss": 0.746,
      "step": 144
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.572519083969466e-05,
      "loss": 0.6697,
      "step": 145
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.569465648854963e-05,
      "loss": 0.47,
      "step": 146
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.566412213740458e-05,
      "loss": 0.5837,
      "step": 147
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.563358778625955e-05,
      "loss": 0.655,
      "step": 148
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.560305343511451e-05,
      "loss": 0.6164,
      "step": 149
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.557251908396946e-05,
      "loss": 0.5205,
      "step": 150
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.554198473282443e-05,
      "loss": 0.5372,
      "step": 151
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.55114503816794e-05,
      "loss": 0.7136,
      "step": 152
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.548091603053434e-05,
      "loss": 0.72,
      "step": 153
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.545038167938932e-05,
      "loss": 0.7468,
      "step": 154
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.541984732824429e-05,
      "loss": 0.5839,
      "step": 155
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.538931297709924e-05,
      "loss": 0.6264,
      "step": 156
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.53587786259542e-05,
      "loss": 0.8154,
      "step": 157
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.532824427480917e-05,
      "loss": 0.5565,
      "step": 158
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.529770992366412e-05,
      "loss": 0.8639,
      "step": 159
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.526717557251908e-05,
      "loss": 0.6717,
      "step": 160
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.523664122137405e-05,
      "loss": 0.5948,
      "step": 161
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.520610687022901e-05,
      "loss": 0.518,
      "step": 162
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.517557251908398e-05,
      "loss": 0.5955,
      "step": 163
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.514503816793894e-05,
      "loss": 0.8849,
      "step": 164
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.511450381679389e-05,
      "loss": 0.4757,
      "step": 165
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.508396946564886e-05,
      "loss": 0.4877,
      "step": 166
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.505343511450382e-05,
      "loss": 0.4565,
      "step": 167
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.502290076335879e-05,
      "loss": 0.4855,
      "step": 168
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.499236641221374e-05,
      "loss": 0.4974,
      "step": 169
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.496183206106871e-05,
      "loss": 0.5089,
      "step": 170
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.493129770992368e-05,
      "loss": 0.4643,
      "step": 171
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.490076335877863e-05,
      "loss": 0.4323,
      "step": 172
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.48702290076336e-05,
      "loss": 0.5418,
      "step": 173
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.483969465648856e-05,
      "loss": 0.4877,
      "step": 174
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.480916030534351e-05,
      "loss": 0.4546,
      "step": 175
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.477862595419847e-05,
      "loss": 0.6339,
      "step": 176
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.474809160305344e-05,
      "loss": 0.5701,
      "step": 177
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.47175572519084e-05,
      "loss": 0.6023,
      "step": 178
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.468702290076337e-05,
      "loss": 0.5727,
      "step": 179
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.465648854961833e-05,
      "loss": 0.6136,
      "step": 180
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.462595419847328e-05,
      "loss": 0.3309,
      "step": 181
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.459541984732825e-05,
      "loss": 0.3104,
      "step": 182
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.456488549618321e-05,
      "loss": 0.3441,
      "step": 183
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.453435114503816e-05,
      "loss": 0.423,
      "step": 184
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.450381679389313e-05,
      "loss": 0.5573,
      "step": 185
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.44732824427481e-05,
      "loss": 0.4989,
      "step": 186
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.444274809160306e-05,
      "loss": 0.5366,
      "step": 187
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.441221374045802e-05,
      "loss": 0.5411,
      "step": 188
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.438167938931299e-05,
      "loss": 0.449,
      "step": 189
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.435114503816794e-05,
      "loss": 0.4212,
      "step": 190
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.43206106870229e-05,
      "loss": 0.4869,
      "step": 191
    },
    {
      "epoch": 1.17,
      "learning_rate": 9.429007633587787e-05,
      "loss": 0.4349,
      "step": 192
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.425954198473283e-05,
      "loss": 0.4398,
      "step": 193
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.42290076335878e-05,
      "loss": 0.4411,
      "step": 194
    },
    {
      "epoch": 1.19,
      "learning_rate": 9.419847328244276e-05,
      "loss": 0.3411,
      "step": 195
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.416793893129771e-05,
      "loss": 0.4828,
      "step": 196
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.413740458015267e-05,
      "loss": 0.4931,
      "step": 197
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.410687022900764e-05,
      "loss": 0.4969,
      "step": 198
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.40763358778626e-05,
      "loss": 0.4848,
      "step": 199
    },
    {
      "epoch": 1.22,
      "learning_rate": 9.404580152671755e-05,
      "loss": 0.6431,
      "step": 200
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.401526717557252e-05,
      "loss": 0.2808,
      "step": 201
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.39847328244275e-05,
      "loss": 0.5772,
      "step": 202
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.395419847328245e-05,
      "loss": 0.4815,
      "step": 203
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.392366412213741e-05,
      "loss": 0.426,
      "step": 204
    },
    {
      "epoch": 1.25,
      "learning_rate": 9.389312977099238e-05,
      "loss": 0.5129,
      "step": 205
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.386259541984733e-05,
      "loss": 0.4378,
      "step": 206
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.383206106870229e-05,
      "loss": 0.4231,
      "step": 207
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.380152671755726e-05,
      "loss": 0.4086,
      "step": 208
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.377099236641221e-05,
      "loss": 0.5333,
      "step": 209
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.374045801526719e-05,
      "loss": 0.4343,
      "step": 210
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.370992366412215e-05,
      "loss": 0.4015,
      "step": 211
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.36793893129771e-05,
      "loss": 0.3846,
      "step": 212
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.364885496183207e-05,
      "loss": 0.4687,
      "step": 213
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.361832061068703e-05,
      "loss": 0.5908,
      "step": 214
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.358778625954198e-05,
      "loss": 0.4907,
      "step": 215
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.355725190839695e-05,
      "loss": 0.4705,
      "step": 216
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.352671755725191e-05,
      "loss": 0.503,
      "step": 217
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.349618320610687e-05,
      "loss": 0.4133,
      "step": 218
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.346564885496184e-05,
      "loss": 0.4069,
      "step": 219
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.34351145038168e-05,
      "loss": 0.4424,
      "step": 220
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.340458015267175e-05,
      "loss": 0.375,
      "step": 221
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.337404580152672e-05,
      "loss": 0.5487,
      "step": 222
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.334351145038168e-05,
      "loss": 0.434,
      "step": 223
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.331297709923665e-05,
      "loss": 0.4031,
      "step": 224
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.32824427480916e-05,
      "loss": 0.4452,
      "step": 225
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.325190839694658e-05,
      "loss": 0.4341,
      "step": 226
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.322137404580153e-05,
      "loss": 0.5052,
      "step": 227
    },
    {
      "epoch": 1.39,
      "learning_rate": 9.319083969465649e-05,
      "loss": 0.3961,
      "step": 228
    },
    {
      "epoch": 1.4,
      "learning_rate": 9.316030534351146e-05,
      "loss": 0.3346,
      "step": 229
    },
    {
      "epoch": 1.4,
      "learning_rate": 9.312977099236642e-05,
      "loss": 0.4737,
      "step": 230
    },
    {
      "epoch": 1.41,
      "learning_rate": 9.309923664122137e-05,
      "loss": 0.408,
      "step": 231
    },
    {
      "epoch": 1.41,
      "learning_rate": 9.306870229007634e-05,
      "loss": 0.545,
      "step": 232
    },
    {
      "epoch": 1.42,
      "learning_rate": 9.30381679389313e-05,
      "loss": 0.5353,
      "step": 233
    },
    {
      "epoch": 1.43,
      "learning_rate": 9.300763358778627e-05,
      "loss": 0.4431,
      "step": 234
    },
    {
      "epoch": 1.43,
      "learning_rate": 9.297709923664123e-05,
      "loss": 0.5402,
      "step": 235
    },
    {
      "epoch": 1.44,
      "learning_rate": 9.29465648854962e-05,
      "loss": 0.3853,
      "step": 236
    },
    {
      "epoch": 1.45,
      "learning_rate": 9.291603053435115e-05,
      "loss": 0.3737,
      "step": 237
    },
    {
      "epoch": 1.45,
      "learning_rate": 9.288549618320611e-05,
      "loss": 0.4921,
      "step": 238
    },
    {
      "epoch": 1.46,
      "learning_rate": 9.285496183206108e-05,
      "loss": 0.3359,
      "step": 239
    },
    {
      "epoch": 1.46,
      "learning_rate": 9.282442748091603e-05,
      "loss": 0.3831,
      "step": 240
    },
    {
      "epoch": 1.47,
      "learning_rate": 9.279389312977099e-05,
      "loss": 0.5496,
      "step": 241
    },
    {
      "epoch": 1.48,
      "learning_rate": 9.276335877862597e-05,
      "loss": 0.4593,
      "step": 242
    },
    {
      "epoch": 1.48,
      "learning_rate": 9.273282442748092e-05,
      "loss": 0.3947,
      "step": 243
    },
    {
      "epoch": 1.49,
      "learning_rate": 9.270229007633588e-05,
      "loss": 0.357,
      "step": 244
    },
    {
      "epoch": 1.49,
      "learning_rate": 9.267175572519085e-05,
      "loss": 0.5329,
      "step": 245
    },
    {
      "epoch": 1.5,
      "learning_rate": 9.26412213740458e-05,
      "loss": 0.4296,
      "step": 246
    },
    {
      "epoch": 1.51,
      "learning_rate": 9.261068702290076e-05,
      "loss": 0.4837,
      "step": 247
    },
    {
      "epoch": 1.51,
      "learning_rate": 9.258015267175573e-05,
      "loss": 0.4717,
      "step": 248
    },
    {
      "epoch": 1.52,
      "learning_rate": 9.254961832061069e-05,
      "loss": 0.3721,
      "step": 249
    },
    {
      "epoch": 1.52,
      "learning_rate": 9.251908396946566e-05,
      "loss": 0.4464,
      "step": 250
    },
    {
      "epoch": 1.53,
      "learning_rate": 9.248854961832062e-05,
      "loss": 0.398,
      "step": 251
    },
    {
      "epoch": 1.54,
      "learning_rate": 9.245801526717557e-05,
      "loss": 0.6588,
      "step": 252
    },
    {
      "epoch": 1.54,
      "learning_rate": 9.242748091603054e-05,
      "loss": 0.3834,
      "step": 253
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.23969465648855e-05,
      "loss": 0.4862,
      "step": 254
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.236641221374047e-05,
      "loss": 0.483,
      "step": 255
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.233587786259542e-05,
      "loss": 0.4341,
      "step": 256
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.230534351145038e-05,
      "loss": 0.4634,
      "step": 257
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.227480916030535e-05,
      "loss": 0.6184,
      "step": 258
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.224427480916031e-05,
      "loss": 0.4843,
      "step": 259
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.221374045801528e-05,
      "loss": 0.5829,
      "step": 260
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.218320610687024e-05,
      "loss": 0.4005,
      "step": 261
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.215267175572519e-05,
      "loss": 0.5087,
      "step": 262
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.212213740458016e-05,
      "loss": 0.4909,
      "step": 263
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.209160305343512e-05,
      "loss": 0.4743,
      "step": 264
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.206106870229007e-05,
      "loss": 0.4658,
      "step": 265
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.203053435114505e-05,
      "loss": 0.2896,
      "step": 266
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.200000000000001e-05,
      "loss": 0.4785,
      "step": 267
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.196946564885496e-05,
      "loss": 0.5576,
      "step": 268
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.193893129770993e-05,
      "loss": 0.5445,
      "step": 269
    },
    {
      "epoch": 1.65,
      "learning_rate": 9.19083969465649e-05,
      "loss": 0.3801,
      "step": 270
    },
    {
      "epoch": 1.65,
      "learning_rate": 9.187786259541984e-05,
      "loss": 0.4339,
      "step": 271
    },
    {
      "epoch": 1.66,
      "learning_rate": 9.184732824427481e-05,
      "loss": 0.375,
      "step": 272
    },
    {
      "epoch": 1.66,
      "learning_rate": 9.181679389312977e-05,
      "loss": 0.4907,
      "step": 273
    },
    {
      "epoch": 1.67,
      "learning_rate": 9.178625954198474e-05,
      "loss": 0.5036,
      "step": 274
    },
    {
      "epoch": 1.68,
      "learning_rate": 9.17557251908397e-05,
      "loss": 0.4389,
      "step": 275
    },
    {
      "epoch": 1.68,
      "learning_rate": 9.172519083969467e-05,
      "loss": 0.5531,
      "step": 276
    },
    {
      "epoch": 1.69,
      "learning_rate": 9.169465648854962e-05,
      "loss": 0.6006,
      "step": 277
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.166412213740458e-05,
      "loss": 0.5804,
      "step": 278
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.163358778625955e-05,
      "loss": 0.5696,
      "step": 279
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.160305343511451e-05,
      "loss": 0.4212,
      "step": 280
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.157251908396946e-05,
      "loss": 0.518,
      "step": 281
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.154198473282444e-05,
      "loss": 0.4382,
      "step": 282
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.151145038167939e-05,
      "loss": 0.4195,
      "step": 283
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.148091603053436e-05,
      "loss": 0.4913,
      "step": 284
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.145038167938932e-05,
      "loss": 0.5461,
      "step": 285
    },
    {
      "epoch": 1.74,
      "learning_rate": 9.141984732824428e-05,
      "loss": 0.3565,
      "step": 286
    },
    {
      "epoch": 1.75,
      "learning_rate": 9.138931297709924e-05,
      "loss": 0.3885,
      "step": 287
    },
    {
      "epoch": 1.76,
      "learning_rate": 9.13587786259542e-05,
      "loss": 0.4252,
      "step": 288
    },
    {
      "epoch": 1.76,
      "learning_rate": 9.132824427480916e-05,
      "loss": 0.4245,
      "step": 289
    },
    {
      "epoch": 1.77,
      "learning_rate": 9.129770992366413e-05,
      "loss": 0.4556,
      "step": 290
    },
    {
      "epoch": 1.77,
      "learning_rate": 9.12671755725191e-05,
      "loss": 0.4611,
      "step": 291
    },
    {
      "epoch": 1.78,
      "learning_rate": 9.123664122137406e-05,
      "loss": 0.3452,
      "step": 292
    },
    {
      "epoch": 1.79,
      "learning_rate": 9.120610687022901e-05,
      "loss": 0.3204,
      "step": 293
    },
    {
      "epoch": 1.79,
      "learning_rate": 9.117557251908397e-05,
      "loss": 0.4647,
      "step": 294
    },
    {
      "epoch": 1.8,
      "learning_rate": 9.114503816793894e-05,
      "loss": 0.4932,
      "step": 295
    },
    {
      "epoch": 1.8,
      "learning_rate": 9.111450381679389e-05,
      "loss": 0.4036,
      "step": 296
    },
    {
      "epoch": 1.81,
      "learning_rate": 9.108396946564885e-05,
      "loss": 0.5425,
      "step": 297
    },
    {
      "epoch": 1.82,
      "learning_rate": 9.105343511450383e-05,
      "loss": 0.4853,
      "step": 298
    },
    {
      "epoch": 1.82,
      "learning_rate": 9.102290076335878e-05,
      "loss": 0.4889,
      "step": 299
    },
    {
      "epoch": 1.83,
      "learning_rate": 9.099236641221375e-05,
      "loss": 0.3967,
      "step": 300
    },
    {
      "epoch": 1.84,
      "learning_rate": 9.096183206106871e-05,
      "loss": 0.4943,
      "step": 301
    },
    {
      "epoch": 1.84,
      "learning_rate": 9.093129770992366e-05,
      "loss": 0.4495,
      "step": 302
    },
    {
      "epoch": 1.85,
      "learning_rate": 9.090076335877863e-05,
      "loss": 0.3957,
      "step": 303
    },
    {
      "epoch": 1.85,
      "learning_rate": 9.087022900763359e-05,
      "loss": 0.5299,
      "step": 304
    },
    {
      "epoch": 1.86,
      "learning_rate": 9.083969465648856e-05,
      "loss": 0.5227,
      "step": 305
    },
    {
      "epoch": 1.87,
      "learning_rate": 9.080916030534351e-05,
      "loss": 0.4052,
      "step": 306
    },
    {
      "epoch": 1.87,
      "learning_rate": 9.077862595419849e-05,
      "loss": 0.3982,
      "step": 307
    },
    {
      "epoch": 1.88,
      "learning_rate": 9.074809160305344e-05,
      "loss": 0.3826,
      "step": 308
    },
    {
      "epoch": 1.88,
      "learning_rate": 9.07175572519084e-05,
      "loss": 0.5636,
      "step": 309
    },
    {
      "epoch": 1.89,
      "learning_rate": 9.068702290076337e-05,
      "loss": 0.3812,
      "step": 310
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.065648854961833e-05,
      "loss": 0.4326,
      "step": 311
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.062595419847328e-05,
      "loss": 0.3939,
      "step": 312
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.059541984732825e-05,
      "loss": 0.4156,
      "step": 313
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.056488549618321e-05,
      "loss": 0.5742,
      "step": 314
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.053435114503817e-05,
      "loss": 0.4274,
      "step": 315
    },
    {
      "epoch": 1.93,
      "learning_rate": 9.050381679389314e-05,
      "loss": 0.5505,
      "step": 316
    },
    {
      "epoch": 1.93,
      "learning_rate": 9.04732824427481e-05,
      "loss": 0.531,
      "step": 317
    },
    {
      "epoch": 1.94,
      "learning_rate": 9.044274809160305e-05,
      "loss": 0.3835,
      "step": 318
    },
    {
      "epoch": 1.95,
      "learning_rate": 9.041221374045802e-05,
      "loss": 0.5963,
      "step": 319
    },
    {
      "epoch": 1.95,
      "learning_rate": 9.038167938931298e-05,
      "loss": 0.479,
      "step": 320
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.035114503816793e-05,
      "loss": 0.3185,
      "step": 321
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.03206106870229e-05,
      "loss": 0.3605,
      "step": 322
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.029007633587788e-05,
      "loss": 0.4668,
      "step": 323
    },
    {
      "epoch": 1.98,
      "learning_rate": 9.025954198473283e-05,
      "loss": 0.4028,
      "step": 324
    },
    {
      "epoch": 1.98,
      "learning_rate": 9.022900763358779e-05,
      "loss": 0.3362,
      "step": 325
    },
    {
      "epoch": 1.99,
      "learning_rate": 9.019847328244276e-05,
      "loss": 0.4877,
      "step": 326
    },
    {
      "epoch": 1.99,
      "learning_rate": 9.016793893129771e-05,
      "loss": 0.3382,
      "step": 327
    },
    {
      "epoch": 2.0,
      "learning_rate": 9.013740458015267e-05,
      "loss": 0.4349,
      "step": 328
    },
    {
      "epoch": 2.01,
      "learning_rate": 9.010687022900764e-05,
      "loss": 0.4012,
      "step": 329
    },
    {
      "epoch": 2.01,
      "learning_rate": 9.007633587786259e-05,
      "loss": 0.1814,
      "step": 330
    },
    {
      "epoch": 2.02,
      "learning_rate": 9.004580152671757e-05,
      "loss": 0.2782,
      "step": 331
    },
    {
      "epoch": 2.02,
      "learning_rate": 9.001526717557253e-05,
      "loss": 0.1797,
      "step": 332
    },
    {
      "epoch": 2.03,
      "learning_rate": 8.998473282442748e-05,
      "loss": 0.3017,
      "step": 333
    },
    {
      "epoch": 2.04,
      "learning_rate": 8.995419847328245e-05,
      "loss": 0.1886,
      "step": 334
    },
    {
      "epoch": 2.04,
      "learning_rate": 8.992366412213741e-05,
      "loss": 0.2897,
      "step": 335
    },
    {
      "epoch": 2.05,
      "learning_rate": 8.989312977099237e-05,
      "loss": 0.211,
      "step": 336
    },
    {
      "epoch": 2.05,
      "learning_rate": 8.986259541984733e-05,
      "loss": 0.2853,
      "step": 337
    },
    {
      "epoch": 2.06,
      "learning_rate": 8.983206106870229e-05,
      "loss": 0.2478,
      "step": 338
    },
    {
      "epoch": 2.07,
      "learning_rate": 8.980152671755725e-05,
      "loss": 0.2558,
      "step": 339
    },
    {
      "epoch": 2.07,
      "learning_rate": 8.977099236641222e-05,
      "loss": 0.2562,
      "step": 340
    },
    {
      "epoch": 2.08,
      "learning_rate": 8.974045801526718e-05,
      "loss": 0.2225,
      "step": 341
    },
    {
      "epoch": 2.09,
      "learning_rate": 8.970992366412215e-05,
      "loss": 0.4078,
      "step": 342
    },
    {
      "epoch": 2.09,
      "learning_rate": 8.96793893129771e-05,
      "loss": 0.2403,
      "step": 343
    },
    {
      "epoch": 2.1,
      "learning_rate": 8.964885496183206e-05,
      "loss": 0.2007,
      "step": 344
    },
    {
      "epoch": 2.1,
      "learning_rate": 8.961832061068703e-05,
      "loss": 0.2818,
      "step": 345
    },
    {
      "epoch": 2.11,
      "learning_rate": 8.958778625954198e-05,
      "loss": 0.2593,
      "step": 346
    },
    {
      "epoch": 2.12,
      "learning_rate": 8.955725190839696e-05,
      "loss": 0.258,
      "step": 347
    },
    {
      "epoch": 2.12,
      "learning_rate": 8.952671755725192e-05,
      "loss": 0.265,
      "step": 348
    },
    {
      "epoch": 2.13,
      "learning_rate": 8.949618320610687e-05,
      "loss": 0.3107,
      "step": 349
    },
    {
      "epoch": 2.13,
      "learning_rate": 8.946564885496184e-05,
      "loss": 0.3126,
      "step": 350
    },
    {
      "epoch": 2.14,
      "learning_rate": 8.94351145038168e-05,
      "loss": 0.2321,
      "step": 351
    },
    {
      "epoch": 2.15,
      "learning_rate": 8.940458015267175e-05,
      "loss": 0.2611,
      "step": 352
    },
    {
      "epoch": 2.15,
      "learning_rate": 8.937404580152672e-05,
      "loss": 0.24,
      "step": 353
    },
    {
      "epoch": 2.16,
      "learning_rate": 8.934351145038168e-05,
      "loss": 0.1878,
      "step": 354
    },
    {
      "epoch": 2.16,
      "learning_rate": 8.931297709923665e-05,
      "loss": 0.2853,
      "step": 355
    },
    {
      "epoch": 2.17,
      "learning_rate": 8.928244274809161e-05,
      "loss": 0.2292,
      "step": 356
    },
    {
      "epoch": 2.18,
      "learning_rate": 8.925190839694658e-05,
      "loss": 0.1908,
      "step": 357
    },
    {
      "epoch": 2.18,
      "learning_rate": 8.922137404580153e-05,
      "loss": 0.1887,
      "step": 358
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.919083969465649e-05,
      "loss": 0.2512,
      "step": 359
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.916030534351145e-05,
      "loss": 0.2896,
      "step": 360
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.91297709923664e-05,
      "loss": 0.2679,
      "step": 361
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.909923664122137e-05,
      "loss": 0.1891,
      "step": 362
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.906870229007635e-05,
      "loss": 0.2209,
      "step": 363
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.90381679389313e-05,
      "loss": 0.262,
      "step": 364
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.900763358778626e-05,
      "loss": 0.1617,
      "step": 365
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.897709923664123e-05,
      "loss": 0.175,
      "step": 366
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.894656488549619e-05,
      "loss": 0.2925,
      "step": 367
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.891603053435114e-05,
      "loss": 0.2674,
      "step": 368
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.888549618320611e-05,
      "loss": 0.2234,
      "step": 369
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.885496183206107e-05,
      "loss": 0.2929,
      "step": 370
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.882442748091604e-05,
      "loss": 0.3237,
      "step": 371
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.8793893129771e-05,
      "loss": 0.2561,
      "step": 372
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.876335877862597e-05,
      "loss": 0.1837,
      "step": 373
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.873282442748092e-05,
      "loss": 0.1659,
      "step": 374
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.870229007633588e-05,
      "loss": 0.237,
      "step": 375
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.867175572519085e-05,
      "loss": 0.2729,
      "step": 376
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.86412213740458e-05,
      "loss": 0.2853,
      "step": 377
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.861068702290076e-05,
      "loss": 0.2203,
      "step": 378
    },
    {
      "epoch": 2.31,
      "learning_rate": 8.858015267175574e-05,
      "loss": 0.2985,
      "step": 379
    },
    {
      "epoch": 2.32,
      "learning_rate": 8.854961832061069e-05,
      "loss": 0.1779,
      "step": 380
    },
    {
      "epoch": 2.32,
      "learning_rate": 8.851908396946566e-05,
      "loss": 0.3121,
      "step": 381
    },
    {
      "epoch": 2.33,
      "learning_rate": 8.848854961832062e-05,
      "loss": 0.2587,
      "step": 382
    },
    {
      "epoch": 2.34,
      "learning_rate": 8.845801526717557e-05,
      "loss": 0.1986,
      "step": 383
    },
    {
      "epoch": 2.34,
      "learning_rate": 8.842748091603054e-05,
      "loss": 0.1553,
      "step": 384
    },
    {
      "epoch": 2.35,
      "learning_rate": 8.83969465648855e-05,
      "loss": 0.2581,
      "step": 385
    },
    {
      "epoch": 2.35,
      "learning_rate": 8.836641221374045e-05,
      "loss": 0.2513,
      "step": 386
    },
    {
      "epoch": 2.36,
      "learning_rate": 8.833587786259543e-05,
      "loss": 0.2566,
      "step": 387
    },
    {
      "epoch": 2.37,
      "learning_rate": 8.83053435114504e-05,
      "loss": 0.258,
      "step": 388
    },
    {
      "epoch": 2.37,
      "learning_rate": 8.827480916030534e-05,
      "loss": 0.2188,
      "step": 389
    },
    {
      "epoch": 2.38,
      "learning_rate": 8.824427480916031e-05,
      "loss": 0.1876,
      "step": 390
    },
    {
      "epoch": 2.38,
      "learning_rate": 8.821374045801527e-05,
      "loss": 0.2914,
      "step": 391
    },
    {
      "epoch": 2.39,
      "learning_rate": 8.818320610687022e-05,
      "loss": 0.2699,
      "step": 392
    },
    {
      "epoch": 2.4,
      "learning_rate": 8.815267175572519e-05,
      "loss": 0.2909,
      "step": 393
    },
    {
      "epoch": 2.4,
      "learning_rate": 8.812213740458015e-05,
      "loss": 0.2052,
      "step": 394
    },
    {
      "epoch": 2.41,
      "learning_rate": 8.809160305343512e-05,
      "loss": 0.2693,
      "step": 395
    },
    {
      "epoch": 2.41,
      "learning_rate": 8.806106870229008e-05,
      "loss": 0.2288,
      "step": 396
    },
    {
      "epoch": 2.42,
      "learning_rate": 8.803053435114505e-05,
      "loss": 0.2341,
      "step": 397
    },
    {
      "epoch": 2.43,
      "learning_rate": 8.800000000000001e-05,
      "loss": 0.2062,
      "step": 398
    },
    {
      "epoch": 2.43,
      "learning_rate": 8.796946564885496e-05,
      "loss": 0.3259,
      "step": 399
    },
    {
      "epoch": 2.44,
      "learning_rate": 8.793893129770993e-05,
      "loss": 0.198,
      "step": 400
    },
    {
      "epoch": 2.45,
      "learning_rate": 8.790839694656489e-05,
      "loss": 0.2691,
      "step": 401
    },
    {
      "epoch": 2.45,
      "learning_rate": 8.787786259541984e-05,
      "loss": 0.3033,
      "step": 402
    },
    {
      "epoch": 2.46,
      "learning_rate": 8.784732824427482e-05,
      "loss": 0.1691,
      "step": 403
    },
    {
      "epoch": 2.46,
      "learning_rate": 8.781679389312978e-05,
      "loss": 0.2174,
      "step": 404
    },
    {
      "epoch": 2.47,
      "learning_rate": 8.778625954198474e-05,
      "loss": 0.2557,
      "step": 405
    },
    {
      "epoch": 2.48,
      "learning_rate": 8.77557251908397e-05,
      "loss": 0.2545,
      "step": 406
    },
    {
      "epoch": 2.48,
      "learning_rate": 8.772519083969466e-05,
      "loss": 0.3322,
      "step": 407
    },
    {
      "epoch": 2.49,
      "learning_rate": 8.769465648854962e-05,
      "loss": 0.2399,
      "step": 408
    },
    {
      "epoch": 2.49,
      "learning_rate": 8.766412213740458e-05,
      "loss": 0.2669,
      "step": 409
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.763358778625954e-05,
      "loss": 0.3598,
      "step": 410
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.760305343511451e-05,
      "loss": 0.246,
      "step": 411
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.757251908396947e-05,
      "loss": 0.2945,
      "step": 412
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.754198473282444e-05,
      "loss": 0.2345,
      "step": 413
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.751145038167939e-05,
      "loss": 0.209,
      "step": 414
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.748091603053435e-05,
      "loss": 0.2816,
      "step": 415
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.745038167938932e-05,
      "loss": 0.17,
      "step": 416
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.741984732824427e-05,
      "loss": 0.279,
      "step": 417
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.738931297709923e-05,
      "loss": 0.2916,
      "step": 418
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.735877862595421e-05,
      "loss": 0.2809,
      "step": 419
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.732824427480916e-05,
      "loss": 0.2261,
      "step": 420
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.729770992366413e-05,
      "loss": 0.2473,
      "step": 421
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.726717557251909e-05,
      "loss": 0.2096,
      "step": 422
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.723664122137404e-05,
      "loss": 0.2658,
      "step": 423
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.720610687022901e-05,
      "loss": 0.234,
      "step": 424
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.717557251908397e-05,
      "loss": 0.3053,
      "step": 425
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.714503816793894e-05,
      "loss": 0.2429,
      "step": 426
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.71145038167939e-05,
      "loss": 0.2079,
      "step": 427
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.708396946564887e-05,
      "loss": 0.3142,
      "step": 428
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.705343511450383e-05,
      "loss": 0.2435,
      "step": 429
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.702290076335878e-05,
      "loss": 0.2399,
      "step": 430
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.699236641221375e-05,
      "loss": 0.2412,
      "step": 431
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.696183206106871e-05,
      "loss": 0.21,
      "step": 432
    },
    {
      "epoch": 2.64,
      "learning_rate": 8.693129770992366e-05,
      "loss": 0.2897,
      "step": 433
    },
    {
      "epoch": 2.65,
      "learning_rate": 8.690076335877862e-05,
      "loss": 0.2416,
      "step": 434
    },
    {
      "epoch": 2.65,
      "learning_rate": 8.68702290076336e-05,
      "loss": 0.2614,
      "step": 435
    },
    {
      "epoch": 2.66,
      "learning_rate": 8.683969465648855e-05,
      "loss": 0.1989,
      "step": 436
    },
    {
      "epoch": 2.66,
      "learning_rate": 8.680916030534352e-05,
      "loss": 0.274,
      "step": 437
    },
    {
      "epoch": 2.67,
      "learning_rate": 8.677862595419848e-05,
      "loss": 0.2354,
      "step": 438
    },
    {
      "epoch": 2.68,
      "learning_rate": 8.674809160305343e-05,
      "loss": 0.1938,
      "step": 439
    },
    {
      "epoch": 2.68,
      "learning_rate": 8.67175572519084e-05,
      "loss": 0.2172,
      "step": 440
    },
    {
      "epoch": 2.69,
      "learning_rate": 8.668702290076336e-05,
      "loss": 0.2853,
      "step": 441
    },
    {
      "epoch": 2.7,
      "learning_rate": 8.665648854961831e-05,
      "loss": 0.2652,
      "step": 442
    },
    {
      "epoch": 2.7,
      "learning_rate": 8.662595419847329e-05,
      "loss": 0.2447,
      "step": 443
    },
    {
      "epoch": 2.71,
      "learning_rate": 8.659541984732826e-05,
      "loss": 0.2776,
      "step": 444
    },
    {
      "epoch": 2.71,
      "learning_rate": 8.656488549618321e-05,
      "loss": 0.3307,
      "step": 445
    },
    {
      "epoch": 2.72,
      "learning_rate": 8.653435114503817e-05,
      "loss": 0.2604,
      "step": 446
    },
    {
      "epoch": 2.73,
      "learning_rate": 8.650381679389314e-05,
      "loss": 0.2863,
      "step": 447
    },
    {
      "epoch": 2.73,
      "learning_rate": 8.647328244274809e-05,
      "loss": 0.1823,
      "step": 448
    },
    {
      "epoch": 2.74,
      "learning_rate": 8.644274809160305e-05,
      "loss": 0.3117,
      "step": 449
    },
    {
      "epoch": 2.74,
      "learning_rate": 8.641221374045802e-05,
      "loss": 0.253,
      "step": 450
    },
    {
      "epoch": 2.75,
      "learning_rate": 8.638167938931298e-05,
      "loss": 0.2414,
      "step": 451
    },
    {
      "epoch": 2.76,
      "learning_rate": 8.635114503816795e-05,
      "loss": 0.2821,
      "step": 452
    },
    {
      "epoch": 2.76,
      "learning_rate": 8.632061068702291e-05,
      "loss": 0.2559,
      "step": 453
    },
    {
      "epoch": 2.77,
      "learning_rate": 8.629007633587786e-05,
      "loss": 0.2269,
      "step": 454
    },
    {
      "epoch": 2.77,
      "learning_rate": 8.625954198473283e-05,
      "loss": 0.345,
      "step": 455
    },
    {
      "epoch": 2.78,
      "learning_rate": 8.622900763358779e-05,
      "loss": 0.275,
      "step": 456
    },
    {
      "epoch": 2.79,
      "learning_rate": 8.619847328244275e-05,
      "loss": 0.2764,
      "step": 457
    },
    {
      "epoch": 2.79,
      "learning_rate": 8.61679389312977e-05,
      "loss": 0.2579,
      "step": 458
    },
    {
      "epoch": 2.8,
      "learning_rate": 8.613740458015268e-05,
      "loss": 0.2322,
      "step": 459
    },
    {
      "epoch": 2.8,
      "learning_rate": 8.610687022900765e-05,
      "loss": 0.2384,
      "step": 460
    },
    {
      "epoch": 2.81,
      "learning_rate": 8.60763358778626e-05,
      "loss": 0.2218,
      "step": 461
    },
    {
      "epoch": 2.82,
      "learning_rate": 8.604580152671756e-05,
      "loss": 0.3349,
      "step": 462
    },
    {
      "epoch": 2.82,
      "learning_rate": 8.601526717557253e-05,
      "loss": 0.3369,
      "step": 463
    },
    {
      "epoch": 2.83,
      "learning_rate": 8.598473282442748e-05,
      "loss": 0.3096,
      "step": 464
    },
    {
      "epoch": 2.84,
      "learning_rate": 8.595419847328244e-05,
      "loss": 0.2151,
      "step": 465
    },
    {
      "epoch": 2.84,
      "learning_rate": 8.592366412213741e-05,
      "loss": 0.2439,
      "step": 466
    },
    {
      "epoch": 2.85,
      "learning_rate": 8.589312977099237e-05,
      "loss": 0.2586,
      "step": 467
    },
    {
      "epoch": 2.85,
      "learning_rate": 8.586259541984734e-05,
      "loss": 0.226,
      "step": 468
    },
    {
      "epoch": 2.86,
      "learning_rate": 8.58320610687023e-05,
      "loss": 0.2862,
      "step": 469
    },
    {
      "epoch": 2.87,
      "learning_rate": 8.580152671755725e-05,
      "loss": 0.2419,
      "step": 470
    },
    {
      "epoch": 2.87,
      "learning_rate": 8.577099236641222e-05,
      "loss": 0.1894,
      "step": 471
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.574045801526718e-05,
      "loss": 0.2373,
      "step": 472
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.570992366412213e-05,
      "loss": 0.2734,
      "step": 473
    },
    {
      "epoch": 2.89,
      "learning_rate": 8.56793893129771e-05,
      "loss": 0.2147,
      "step": 474
    },
    {
      "epoch": 2.9,
      "learning_rate": 8.564885496183207e-05,
      "loss": 0.2406,
      "step": 475
    },
    {
      "epoch": 2.9,
      "learning_rate": 8.561832061068703e-05,
      "loss": 0.2034,
      "step": 476
    },
    {
      "epoch": 2.91,
      "learning_rate": 8.558778625954199e-05,
      "loss": 0.307,
      "step": 477
    },
    {
      "epoch": 2.91,
      "learning_rate": 8.555725190839695e-05,
      "loss": 0.2994,
      "step": 478
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.55267175572519e-05,
      "loss": 0.198,
      "step": 479
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.549618320610687e-05,
      "loss": 0.2183,
      "step": 480
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.546564885496183e-05,
      "loss": 0.3347,
      "step": 481
    },
    {
      "epoch": 2.94,
      "learning_rate": 8.54351145038168e-05,
      "loss": 0.3664,
      "step": 482
    },
    {
      "epoch": 2.95,
      "learning_rate": 8.540458015267176e-05,
      "loss": 0.1762,
      "step": 483
    },
    {
      "epoch": 2.95,
      "learning_rate": 8.537404580152673e-05,
      "loss": 0.2786,
      "step": 484
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.534351145038168e-05,
      "loss": 0.2122,
      "step": 485
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.531297709923664e-05,
      "loss": 0.2194,
      "step": 486
    },
    {
      "epoch": 2.97,
      "learning_rate": 8.528244274809161e-05,
      "loss": 0.2681,
      "step": 487
    },
    {
      "epoch": 2.98,
      "learning_rate": 8.525190839694657e-05,
      "loss": 0.3039,
      "step": 488
    },
    {
      "epoch": 2.98,
      "learning_rate": 8.522137404580152e-05,
      "loss": 0.2666,
      "step": 489
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.519083969465649e-05,
      "loss": 0.3254,
      "step": 490
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.516030534351147e-05,
      "loss": 0.2658,
      "step": 491
    },
    {
      "epoch": 3.0,
      "learning_rate": 8.512977099236642e-05,
      "loss": 0.2529,
      "step": 492
    },
    {
      "epoch": 3.01,
      "learning_rate": 8.509923664122138e-05,
      "loss": 0.188,
      "step": 493
    },
    {
      "epoch": 3.01,
      "learning_rate": 8.506870229007635e-05,
      "loss": 0.1928,
      "step": 494
    },
    {
      "epoch": 3.02,
      "learning_rate": 8.50381679389313e-05,
      "loss": 0.1784,
      "step": 495
    },
    {
      "epoch": 3.02,
      "learning_rate": 8.500763358778626e-05,
      "loss": 0.1277,
      "step": 496
    },
    {
      "epoch": 3.03,
      "learning_rate": 8.497709923664123e-05,
      "loss": 0.1697,
      "step": 497
    },
    {
      "epoch": 3.04,
      "learning_rate": 8.494656488549618e-05,
      "loss": 0.1116,
      "step": 498
    },
    {
      "epoch": 3.04,
      "learning_rate": 8.491603053435116e-05,
      "loss": 0.1836,
      "step": 499
    },
    {
      "epoch": 3.05,
      "learning_rate": 8.488549618320612e-05,
      "loss": 0.228,
      "step": 500
    },
    {
      "epoch": 3.05,
      "learning_rate": 8.485496183206107e-05,
      "loss": 0.1418,
      "step": 501
    },
    {
      "epoch": 3.06,
      "learning_rate": 8.482442748091604e-05,
      "loss": 0.2189,
      "step": 502
    },
    {
      "epoch": 3.07,
      "learning_rate": 8.4793893129771e-05,
      "loss": 0.1194,
      "step": 503
    },
    {
      "epoch": 3.07,
      "learning_rate": 8.476335877862595e-05,
      "loss": 0.1397,
      "step": 504
    },
    {
      "epoch": 3.08,
      "learning_rate": 8.473282442748092e-05,
      "loss": 0.1877,
      "step": 505
    },
    {
      "epoch": 3.09,
      "learning_rate": 8.470229007633588e-05,
      "loss": 0.1286,
      "step": 506
    },
    {
      "epoch": 3.09,
      "learning_rate": 8.467175572519084e-05,
      "loss": 0.1678,
      "step": 507
    },
    {
      "epoch": 3.1,
      "learning_rate": 8.464122137404581e-05,
      "loss": 0.2278,
      "step": 508
    },
    {
      "epoch": 3.1,
      "learning_rate": 8.461068702290077e-05,
      "loss": 0.1551,
      "step": 509
    },
    {
      "epoch": 3.11,
      "learning_rate": 8.458015267175572e-05,
      "loss": 0.1838,
      "step": 510
    },
    {
      "epoch": 3.12,
      "learning_rate": 8.454961832061069e-05,
      "loss": 0.1639,
      "step": 511
    },
    {
      "epoch": 3.12,
      "learning_rate": 8.451908396946565e-05,
      "loss": 0.1917,
      "step": 512
    },
    {
      "epoch": 3.13,
      "learning_rate": 8.448854961832062e-05,
      "loss": 0.2001,
      "step": 513
    },
    {
      "epoch": 3.13,
      "learning_rate": 8.445801526717557e-05,
      "loss": 0.2411,
      "step": 514
    },
    {
      "epoch": 3.14,
      "learning_rate": 8.442748091603055e-05,
      "loss": 0.1653,
      "step": 515
    },
    {
      "epoch": 3.15,
      "learning_rate": 8.43969465648855e-05,
      "loss": 0.1688,
      "step": 516
    },
    {
      "epoch": 3.15,
      "learning_rate": 8.436641221374046e-05,
      "loss": 0.1549,
      "step": 517
    },
    {
      "epoch": 3.16,
      "learning_rate": 8.433587786259543e-05,
      "loss": 0.2109,
      "step": 518
    },
    {
      "epoch": 3.16,
      "learning_rate": 8.430534351145039e-05,
      "loss": 0.1737,
      "step": 519
    },
    {
      "epoch": 3.17,
      "learning_rate": 8.427480916030534e-05,
      "loss": 0.1697,
      "step": 520
    },
    {
      "epoch": 3.18,
      "learning_rate": 8.42442748091603e-05,
      "loss": 0.1802,
      "step": 521
    },
    {
      "epoch": 3.18,
      "learning_rate": 8.421374045801527e-05,
      "loss": 0.1386,
      "step": 522
    },
    {
      "epoch": 3.19,
      "learning_rate": 8.418320610687024e-05,
      "loss": 0.1198,
      "step": 523
    },
    {
      "epoch": 3.2,
      "learning_rate": 8.41526717557252e-05,
      "loss": 0.1115,
      "step": 524
    },
    {
      "epoch": 3.2,
      "learning_rate": 8.412213740458016e-05,
      "loss": 0.163,
      "step": 525
    },
    {
      "epoch": 3.21,
      "learning_rate": 8.409160305343512e-05,
      "loss": 0.1632,
      "step": 526
    },
    {
      "epoch": 3.21,
      "learning_rate": 8.406106870229008e-05,
      "loss": 0.1269,
      "step": 527
    },
    {
      "epoch": 3.22,
      "learning_rate": 8.403053435114504e-05,
      "loss": 0.1381,
      "step": 528
    },
    {
      "epoch": 3.23,
      "learning_rate": 8.4e-05,
      "loss": 0.1259,
      "step": 529
    },
    {
      "epoch": 3.23,
      "learning_rate": 8.396946564885496e-05,
      "loss": 0.1762,
      "step": 530
    },
    {
      "epoch": 3.24,
      "learning_rate": 8.393893129770994e-05,
      "loss": 0.2014,
      "step": 531
    },
    {
      "epoch": 3.24,
      "learning_rate": 8.390839694656489e-05,
      "loss": 0.1223,
      "step": 532
    },
    {
      "epoch": 3.25,
      "learning_rate": 8.387786259541985e-05,
      "loss": 0.1958,
      "step": 533
    },
    {
      "epoch": 3.26,
      "learning_rate": 8.384732824427482e-05,
      "loss": 0.1429,
      "step": 534
    },
    {
      "epoch": 3.26,
      "learning_rate": 8.381679389312977e-05,
      "loss": 0.1024,
      "step": 535
    },
    {
      "epoch": 3.27,
      "learning_rate": 8.378625954198473e-05,
      "loss": 0.1527,
      "step": 536
    },
    {
      "epoch": 3.27,
      "learning_rate": 8.37557251908397e-05,
      "loss": 0.1488,
      "step": 537
    },
    {
      "epoch": 3.28,
      "learning_rate": 8.372519083969466e-05,
      "loss": 0.1871,
      "step": 538
    },
    {
      "epoch": 3.29,
      "learning_rate": 8.369465648854963e-05,
      "loss": 0.1104,
      "step": 539
    },
    {
      "epoch": 3.29,
      "learning_rate": 8.366412213740459e-05,
      "loss": 0.1814,
      "step": 540
    },
    {
      "epoch": 3.3,
      "learning_rate": 8.363358778625954e-05,
      "loss": 0.1227,
      "step": 541
    },
    {
      "epoch": 3.3,
      "learning_rate": 8.360305343511451e-05,
      "loss": 0.2913,
      "step": 542
    },
    {
      "epoch": 3.31,
      "learning_rate": 8.357251908396947e-05,
      "loss": 0.1841,
      "step": 543
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.354198473282444e-05,
      "loss": 0.1623,
      "step": 544
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.351145038167939e-05,
      "loss": 0.2268,
      "step": 545
    },
    {
      "epoch": 3.33,
      "learning_rate": 8.348091603053435e-05,
      "loss": 0.1895,
      "step": 546
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.345038167938932e-05,
      "loss": 0.136,
      "step": 547
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.341984732824428e-05,
      "loss": 0.1987,
      "step": 548
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.338931297709924e-05,
      "loss": 0.1911,
      "step": 549
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.335877862595421e-05,
      "loss": 0.1391,
      "step": 550
    },
    {
      "epoch": 3.36,
      "learning_rate": 8.332824427480916e-05,
      "loss": 0.1749,
      "step": 551
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.329770992366412e-05,
      "loss": 0.2244,
      "step": 552
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.326717557251909e-05,
      "loss": 0.1935,
      "step": 553
    },
    {
      "epoch": 3.38,
      "learning_rate": 8.323664122137404e-05,
      "loss": 0.1558,
      "step": 554
    },
    {
      "epoch": 3.38,
      "learning_rate": 8.320610687022902e-05,
      "loss": 0.2371,
      "step": 555
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.317557251908398e-05,
      "loss": 0.1506,
      "step": 556
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.314503816793893e-05,
      "loss": 0.1628,
      "step": 557
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.31145038167939e-05,
      "loss": 0.1364,
      "step": 558
    },
    {
      "epoch": 3.41,
      "learning_rate": 8.308396946564886e-05,
      "loss": 0.1417,
      "step": 559
    },
    {
      "epoch": 3.41,
      "learning_rate": 8.305343511450381e-05,
      "loss": 0.1774,
      "step": 560
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.302290076335878e-05,
      "loss": 0.122,
      "step": 561
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.299236641221374e-05,
      "loss": 0.1767,
      "step": 562
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.296183206106871e-05,
      "loss": 0.1469,
      "step": 563
    },
    {
      "epoch": 3.44,
      "learning_rate": 8.293129770992367e-05,
      "loss": 0.1451,
      "step": 564
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.290076335877864e-05,
      "loss": 0.1772,
      "step": 565
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.287022900763359e-05,
      "loss": 0.2016,
      "step": 566
    },
    {
      "epoch": 3.46,
      "learning_rate": 8.283969465648855e-05,
      "loss": 0.1881,
      "step": 567
    },
    {
      "epoch": 3.46,
      "learning_rate": 8.280916030534352e-05,
      "loss": 0.1281,
      "step": 568
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.277862595419848e-05,
      "loss": 0.2067,
      "step": 569
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.274809160305343e-05,
      "loss": 0.1653,
      "step": 570
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.271755725190841e-05,
      "loss": 0.1952,
      "step": 571
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.268702290076336e-05,
      "loss": 0.1803,
      "step": 572
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.265648854961833e-05,
      "loss": 0.1754,
      "step": 573
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.262595419847329e-05,
      "loss": 0.1684,
      "step": 574
    },
    {
      "epoch": 3.51,
      "learning_rate": 8.259541984732825e-05,
      "loss": 0.2083,
      "step": 575
    },
    {
      "epoch": 3.51,
      "learning_rate": 8.25648854961832e-05,
      "loss": 0.1767,
      "step": 576
    },
    {
      "epoch": 3.52,
      "learning_rate": 8.253435114503817e-05,
      "loss": 0.1875,
      "step": 577
    },
    {
      "epoch": 3.52,
      "learning_rate": 8.250381679389313e-05,
      "loss": 0.1876,
      "step": 578
    },
    {
      "epoch": 3.53,
      "learning_rate": 8.24732824427481e-05,
      "loss": 0.1593,
      "step": 579
    },
    {
      "epoch": 3.54,
      "learning_rate": 8.244274809160306e-05,
      "loss": 0.1535,
      "step": 580
    },
    {
      "epoch": 3.54,
      "learning_rate": 8.241221374045803e-05,
      "loss": 0.2212,
      "step": 581
    },
    {
      "epoch": 3.55,
      "learning_rate": 8.238167938931298e-05,
      "loss": 0.2082,
      "step": 582
    },
    {
      "epoch": 3.55,
      "learning_rate": 8.235114503816794e-05,
      "loss": 0.1695,
      "step": 583
    },
    {
      "epoch": 3.56,
      "learning_rate": 8.232061068702291e-05,
      "loss": 0.2336,
      "step": 584
    },
    {
      "epoch": 3.57,
      "learning_rate": 8.229007633587786e-05,
      "loss": 0.1853,
      "step": 585
    },
    {
      "epoch": 3.57,
      "learning_rate": 8.225954198473282e-05,
      "loss": 0.1577,
      "step": 586
    },
    {
      "epoch": 3.58,
      "learning_rate": 8.22290076335878e-05,
      "loss": 0.1487,
      "step": 587
    },
    {
      "epoch": 3.59,
      "learning_rate": 8.219847328244275e-05,
      "loss": 0.1859,
      "step": 588
    },
    {
      "epoch": 3.59,
      "learning_rate": 8.216793893129772e-05,
      "loss": 0.152,
      "step": 589
    },
    {
      "epoch": 3.6,
      "learning_rate": 8.213740458015268e-05,
      "loss": 0.1558,
      "step": 590
    },
    {
      "epoch": 3.6,
      "learning_rate": 8.210687022900763e-05,
      "loss": 0.2258,
      "step": 591
    },
    {
      "epoch": 3.61,
      "learning_rate": 8.20763358778626e-05,
      "loss": 0.1949,
      "step": 592
    },
    {
      "epoch": 3.62,
      "learning_rate": 8.204580152671756e-05,
      "loss": 0.1727,
      "step": 593
    },
    {
      "epoch": 3.62,
      "learning_rate": 8.201526717557253e-05,
      "loss": 0.1742,
      "step": 594
    },
    {
      "epoch": 3.63,
      "learning_rate": 8.198473282442749e-05,
      "loss": 0.1865,
      "step": 595
    },
    {
      "epoch": 3.63,
      "learning_rate": 8.195419847328245e-05,
      "loss": 0.1916,
      "step": 596
    },
    {
      "epoch": 3.64,
      "learning_rate": 8.19236641221374e-05,
      "loss": 0.1637,
      "step": 597
    },
    {
      "epoch": 3.65,
      "learning_rate": 8.189312977099237e-05,
      "loss": 0.1857,
      "step": 598
    },
    {
      "epoch": 3.65,
      "learning_rate": 8.186259541984733e-05,
      "loss": 0.1766,
      "step": 599
    },
    {
      "epoch": 3.66,
      "learning_rate": 8.18320610687023e-05,
      "loss": 0.1551,
      "step": 600
    },
    {
      "epoch": 3.66,
      "learning_rate": 8.180152671755725e-05,
      "loss": 0.2247,
      "step": 601
    },
    {
      "epoch": 3.67,
      "learning_rate": 8.177099236641221e-05,
      "loss": 0.1889,
      "step": 602
    },
    {
      "epoch": 3.68,
      "learning_rate": 8.174045801526718e-05,
      "loss": 0.1638,
      "step": 603
    },
    {
      "epoch": 3.68,
      "learning_rate": 8.170992366412214e-05,
      "loss": 0.2203,
      "step": 604
    },
    {
      "epoch": 3.69,
      "learning_rate": 8.167938931297711e-05,
      "loss": 0.1717,
      "step": 605
    },
    {
      "epoch": 3.7,
      "learning_rate": 8.164885496183207e-05,
      "loss": 0.1622,
      "step": 606
    },
    {
      "epoch": 3.7,
      "learning_rate": 8.161832061068702e-05,
      "loss": 0.1036,
      "step": 607
    },
    {
      "epoch": 3.71,
      "learning_rate": 8.158778625954199e-05,
      "loss": 0.221,
      "step": 608
    },
    {
      "epoch": 3.71,
      "learning_rate": 8.155725190839695e-05,
      "loss": 0.1926,
      "step": 609
    },
    {
      "epoch": 3.72,
      "learning_rate": 8.15267175572519e-05,
      "loss": 0.1622,
      "step": 610
    },
    {
      "epoch": 3.73,
      "learning_rate": 8.149618320610687e-05,
      "loss": 0.1322,
      "step": 611
    },
    {
      "epoch": 3.73,
      "learning_rate": 8.146564885496185e-05,
      "loss": 0.1905,
      "step": 612
    },
    {
      "epoch": 3.74,
      "learning_rate": 8.14351145038168e-05,
      "loss": 0.1996,
      "step": 613
    },
    {
      "epoch": 3.74,
      "learning_rate": 8.140458015267176e-05,
      "loss": 0.2482,
      "step": 614
    },
    {
      "epoch": 3.75,
      "learning_rate": 8.137404580152673e-05,
      "loss": 0.1584,
      "step": 615
    },
    {
      "epoch": 3.76,
      "learning_rate": 8.134351145038168e-05,
      "loss": 0.1914,
      "step": 616
    },
    {
      "epoch": 3.76,
      "learning_rate": 8.131297709923664e-05,
      "loss": 0.1667,
      "step": 617
    },
    {
      "epoch": 3.77,
      "learning_rate": 8.12824427480916e-05,
      "loss": 0.1395,
      "step": 618
    },
    {
      "epoch": 3.77,
      "learning_rate": 8.125190839694656e-05,
      "loss": 0.2358,
      "step": 619
    },
    {
      "epoch": 3.78,
      "learning_rate": 8.122137404580153e-05,
      "loss": 0.1436,
      "step": 620
    },
    {
      "epoch": 3.79,
      "learning_rate": 8.11908396946565e-05,
      "loss": 0.1794,
      "step": 621
    },
    {
      "epoch": 3.79,
      "learning_rate": 8.116030534351145e-05,
      "loss": 0.1929,
      "step": 622
    },
    {
      "epoch": 3.8,
      "learning_rate": 8.112977099236641e-05,
      "loss": 0.1958,
      "step": 623
    },
    {
      "epoch": 3.8,
      "learning_rate": 8.109923664122138e-05,
      "loss": 0.1785,
      "step": 624
    },
    {
      "epoch": 3.81,
      "learning_rate": 8.106870229007634e-05,
      "loss": 0.1648,
      "step": 625
    },
    {
      "epoch": 3.82,
      "learning_rate": 8.10381679389313e-05,
      "loss": 0.1407,
      "step": 626
    },
    {
      "epoch": 3.82,
      "learning_rate": 8.100763358778626e-05,
      "loss": 0.1669,
      "step": 627
    },
    {
      "epoch": 3.83,
      "learning_rate": 8.097709923664122e-05,
      "loss": 0.2007,
      "step": 628
    },
    {
      "epoch": 3.84,
      "learning_rate": 8.094656488549619e-05,
      "loss": 0.1838,
      "step": 629
    },
    {
      "epoch": 3.84,
      "learning_rate": 8.091603053435115e-05,
      "loss": 0.188,
      "step": 630
    },
    {
      "epoch": 3.85,
      "learning_rate": 8.088549618320612e-05,
      "loss": 0.2075,
      "step": 631
    },
    {
      "epoch": 3.85,
      "learning_rate": 8.085496183206107e-05,
      "loss": 0.1737,
      "step": 632
    },
    {
      "epoch": 3.86,
      "learning_rate": 8.082442748091603e-05,
      "loss": 0.1765,
      "step": 633
    },
    {
      "epoch": 3.87,
      "learning_rate": 8.0793893129771e-05,
      "loss": 0.2095,
      "step": 634
    },
    {
      "epoch": 3.87,
      "learning_rate": 8.076335877862595e-05,
      "loss": 0.152,
      "step": 635
    },
    {
      "epoch": 3.88,
      "learning_rate": 8.073282442748093e-05,
      "loss": 0.1996,
      "step": 636
    },
    {
      "epoch": 3.88,
      "learning_rate": 8.070229007633589e-05,
      "loss": 0.1395,
      "step": 637
    },
    {
      "epoch": 3.89,
      "learning_rate": 8.067175572519084e-05,
      "loss": 0.1855,
      "step": 638
    },
    {
      "epoch": 3.9,
      "learning_rate": 8.06412213740458e-05,
      "loss": 0.1644,
      "step": 639
    },
    {
      "epoch": 3.9,
      "learning_rate": 8.061068702290077e-05,
      "loss": 0.1212,
      "step": 640
    },
    {
      "epoch": 3.91,
      "learning_rate": 8.058015267175572e-05,
      "loss": 0.1906,
      "step": 641
    },
    {
      "epoch": 3.91,
      "learning_rate": 8.054961832061069e-05,
      "loss": 0.1962,
      "step": 642
    },
    {
      "epoch": 3.92,
      "learning_rate": 8.051908396946565e-05,
      "loss": 0.2514,
      "step": 643
    },
    {
      "epoch": 3.93,
      "learning_rate": 8.048854961832062e-05,
      "loss": 0.2318,
      "step": 644
    },
    {
      "epoch": 3.93,
      "learning_rate": 8.045801526717558e-05,
      "loss": 0.1278,
      "step": 645
    },
    {
      "epoch": 3.94,
      "learning_rate": 8.042748091603054e-05,
      "loss": 0.2125,
      "step": 646
    },
    {
      "epoch": 3.95,
      "learning_rate": 8.03969465648855e-05,
      "loss": 0.2389,
      "step": 647
    },
    {
      "epoch": 3.95,
      "learning_rate": 8.036641221374046e-05,
      "loss": 0.1755,
      "step": 648
    },
    {
      "epoch": 3.96,
      "learning_rate": 8.033587786259542e-05,
      "loss": 0.2302,
      "step": 649
    },
    {
      "epoch": 3.96,
      "learning_rate": 8.030534351145038e-05,
      "loss": 0.197,
      "step": 650
    },
    {
      "epoch": 3.97,
      "learning_rate": 8.027480916030534e-05,
      "loss": 0.2151,
      "step": 651
    },
    {
      "epoch": 3.98,
      "learning_rate": 8.024427480916032e-05,
      "loss": 0.1842,
      "step": 652
    },
    {
      "epoch": 3.98,
      "learning_rate": 8.021374045801527e-05,
      "loss": 0.1801,
      "step": 653
    },
    {
      "epoch": 3.99,
      "learning_rate": 8.018320610687023e-05,
      "loss": 0.1773,
      "step": 654
    },
    {
      "epoch": 3.99,
      "learning_rate": 8.01526717557252e-05,
      "loss": 0.1767,
      "step": 655
    },
    {
      "epoch": 4.0,
      "learning_rate": 8.012213740458016e-05,
      "loss": 0.1385,
      "step": 656
    },
    {
      "epoch": 4.01,
      "learning_rate": 8.009160305343511e-05,
      "loss": 0.1492,
      "step": 657
    },
    {
      "epoch": 4.01,
      "learning_rate": 8.006106870229008e-05,
      "loss": 0.1354,
      "step": 658
    },
    {
      "epoch": 4.02,
      "learning_rate": 8.003053435114504e-05,
      "loss": 0.1457,
      "step": 659
    },
    {
      "epoch": 4.02,
      "learning_rate": 8e-05,
      "loss": 0.1314,
      "step": 660
    },
    {
      "epoch": 4.03,
      "learning_rate": 7.996946564885497e-05,
      "loss": 0.1279,
      "step": 661
    },
    {
      "epoch": 4.04,
      "learning_rate": 7.993893129770994e-05,
      "loss": 0.0943,
      "step": 662
    },
    {
      "epoch": 4.04,
      "learning_rate": 7.990839694656489e-05,
      "loss": 0.1233,
      "step": 663
    },
    {
      "epoch": 4.05,
      "learning_rate": 7.987786259541985e-05,
      "loss": 0.1192,
      "step": 664
    },
    {
      "epoch": 4.05,
      "learning_rate": 7.984732824427482e-05,
      "loss": 0.1128,
      "step": 665
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.981679389312977e-05,
      "loss": 0.1316,
      "step": 666
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.978625954198473e-05,
      "loss": 0.1417,
      "step": 667
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.975572519083971e-05,
      "loss": 0.1149,
      "step": 668
    },
    {
      "epoch": 4.08,
      "learning_rate": 7.972519083969466e-05,
      "loss": 0.1172,
      "step": 669
    },
    {
      "epoch": 4.09,
      "learning_rate": 7.969465648854962e-05,
      "loss": 0.0864,
      "step": 670
    },
    {
      "epoch": 4.09,
      "learning_rate": 7.966412213740459e-05,
      "loss": 0.1772,
      "step": 671
    },
    {
      "epoch": 4.1,
      "learning_rate": 7.963358778625954e-05,
      "loss": 0.0974,
      "step": 672
    },
    {
      "epoch": 4.1,
      "learning_rate": 7.96030534351145e-05,
      "loss": 0.1119,
      "step": 673
    },
    {
      "epoch": 4.11,
      "learning_rate": 7.957251908396947e-05,
      "loss": 0.1296,
      "step": 674
    },
    {
      "epoch": 4.12,
      "learning_rate": 7.954198473282442e-05,
      "loss": 0.1457,
      "step": 675
    },
    {
      "epoch": 4.12,
      "learning_rate": 7.95114503816794e-05,
      "loss": 0.126,
      "step": 676
    },
    {
      "epoch": 4.13,
      "learning_rate": 7.948091603053436e-05,
      "loss": 0.1135,
      "step": 677
    },
    {
      "epoch": 4.13,
      "learning_rate": 7.945038167938931e-05,
      "loss": 0.1146,
      "step": 678
    },
    {
      "epoch": 4.14,
      "learning_rate": 7.941984732824428e-05,
      "loss": 0.0828,
      "step": 679
    },
    {
      "epoch": 4.15,
      "learning_rate": 7.938931297709924e-05,
      "loss": 0.1494,
      "step": 680
    },
    {
      "epoch": 4.15,
      "learning_rate": 7.93587786259542e-05,
      "loss": 0.1849,
      "step": 681
    },
    {
      "epoch": 4.16,
      "learning_rate": 7.932824427480916e-05,
      "loss": 0.13,
      "step": 682
    },
    {
      "epoch": 4.16,
      "learning_rate": 7.929770992366412e-05,
      "loss": 0.1666,
      "step": 683
    },
    {
      "epoch": 4.17,
      "learning_rate": 7.926717557251909e-05,
      "loss": 0.1555,
      "step": 684
    },
    {
      "epoch": 4.18,
      "learning_rate": 7.923664122137405e-05,
      "loss": 0.1081,
      "step": 685
    },
    {
      "epoch": 4.18,
      "learning_rate": 7.920610687022902e-05,
      "loss": 0.0998,
      "step": 686
    },
    {
      "epoch": 4.19,
      "learning_rate": 7.917557251908398e-05,
      "loss": 0.1293,
      "step": 687
    },
    {
      "epoch": 4.2,
      "learning_rate": 7.914503816793893e-05,
      "loss": 0.1435,
      "step": 688
    },
    {
      "epoch": 4.2,
      "learning_rate": 7.91145038167939e-05,
      "loss": 0.0983,
      "step": 689
    },
    {
      "epoch": 4.21,
      "learning_rate": 7.908396946564886e-05,
      "loss": 0.1419,
      "step": 690
    },
    {
      "epoch": 4.21,
      "learning_rate": 7.905343511450381e-05,
      "loss": 0.1198,
      "step": 691
    },
    {
      "epoch": 4.22,
      "learning_rate": 7.902290076335879e-05,
      "loss": 0.1289,
      "step": 692
    },
    {
      "epoch": 4.23,
      "learning_rate": 7.899236641221375e-05,
      "loss": 0.1452,
      "step": 693
    },
    {
      "epoch": 4.23,
      "learning_rate": 7.89618320610687e-05,
      "loss": 0.1004,
      "step": 694
    },
    {
      "epoch": 4.24,
      "learning_rate": 7.893129770992367e-05,
      "loss": 0.1463,
      "step": 695
    },
    {
      "epoch": 4.24,
      "learning_rate": 7.890076335877863e-05,
      "loss": 0.1273,
      "step": 696
    },
    {
      "epoch": 4.25,
      "learning_rate": 7.887022900763358e-05,
      "loss": 0.1115,
      "step": 697
    },
    {
      "epoch": 4.26,
      "learning_rate": 7.883969465648855e-05,
      "loss": 0.1244,
      "step": 698
    },
    {
      "epoch": 4.26,
      "learning_rate": 7.880916030534351e-05,
      "loss": 0.1201,
      "step": 699
    },
    {
      "epoch": 4.27,
      "learning_rate": 7.877862595419848e-05,
      "loss": 0.1693,
      "step": 700
    },
    {
      "epoch": 4.27,
      "learning_rate": 7.874809160305344e-05,
      "loss": 0.1272,
      "step": 701
    },
    {
      "epoch": 4.28,
      "learning_rate": 7.871755725190841e-05,
      "loss": 0.1261,
      "step": 702
    },
    {
      "epoch": 4.29,
      "learning_rate": 7.868702290076336e-05,
      "loss": 0.1728,
      "step": 703
    },
    {
      "epoch": 4.29,
      "learning_rate": 7.865648854961832e-05,
      "loss": 0.1058,
      "step": 704
    },
    {
      "epoch": 4.3,
      "learning_rate": 7.862595419847329e-05,
      "loss": 0.1553,
      "step": 705
    },
    {
      "epoch": 4.3,
      "learning_rate": 7.859541984732824e-05,
      "loss": 0.1786,
      "step": 706
    },
    {
      "epoch": 4.31,
      "learning_rate": 7.85648854961832e-05,
      "loss": 0.155,
      "step": 707
    },
    {
      "epoch": 4.32,
      "learning_rate": 7.853435114503818e-05,
      "loss": 0.1347,
      "step": 708
    },
    {
      "epoch": 4.32,
      "learning_rate": 7.850381679389313e-05,
      "loss": 0.1217,
      "step": 709
    },
    {
      "epoch": 4.33,
      "learning_rate": 7.84732824427481e-05,
      "loss": 0.1319,
      "step": 710
    },
    {
      "epoch": 4.34,
      "learning_rate": 7.844274809160306e-05,
      "loss": 0.121,
      "step": 711
    },
    {
      "epoch": 4.34,
      "learning_rate": 7.841221374045801e-05,
      "loss": 0.1674,
      "step": 712
    },
    {
      "epoch": 4.35,
      "learning_rate": 7.838167938931298e-05,
      "loss": 0.1254,
      "step": 713
    },
    {
      "epoch": 4.35,
      "learning_rate": 7.835114503816794e-05,
      "loss": 0.1017,
      "step": 714
    },
    {
      "epoch": 4.36,
      "learning_rate": 7.83206106870229e-05,
      "loss": 0.1224,
      "step": 715
    },
    {
      "epoch": 4.37,
      "learning_rate": 7.829007633587787e-05,
      "loss": 0.12,
      "step": 716
    },
    {
      "epoch": 4.37,
      "learning_rate": 7.825954198473283e-05,
      "loss": 0.1433,
      "step": 717
    },
    {
      "epoch": 4.38,
      "learning_rate": 7.82290076335878e-05,
      "loss": 0.1232,
      "step": 718
    },
    {
      "epoch": 4.38,
      "learning_rate": 7.819847328244275e-05,
      "loss": 0.1414,
      "step": 719
    },
    {
      "epoch": 4.39,
      "learning_rate": 7.816793893129771e-05,
      "loss": 0.1532,
      "step": 720
    },
    {
      "epoch": 4.4,
      "learning_rate": 7.813740458015268e-05,
      "loss": 0.1113,
      "step": 721
    },
    {
      "epoch": 4.4,
      "learning_rate": 7.810687022900763e-05,
      "loss": 0.1376,
      "step": 722
    },
    {
      "epoch": 4.41,
      "learning_rate": 7.80763358778626e-05,
      "loss": 0.123,
      "step": 723
    },
    {
      "epoch": 4.41,
      "learning_rate": 7.804580152671757e-05,
      "loss": 0.147,
      "step": 724
    },
    {
      "epoch": 4.42,
      "learning_rate": 7.801526717557252e-05,
      "loss": 0.1066,
      "step": 725
    },
    {
      "epoch": 4.43,
      "learning_rate": 7.798473282442749e-05,
      "loss": 0.1478,
      "step": 726
    },
    {
      "epoch": 4.43,
      "learning_rate": 7.795419847328245e-05,
      "loss": 0.2081,
      "step": 727
    },
    {
      "epoch": 4.44,
      "learning_rate": 7.79236641221374e-05,
      "loss": 0.1319,
      "step": 728
    },
    {
      "epoch": 4.45,
      "learning_rate": 7.789312977099237e-05,
      "loss": 0.116,
      "step": 729
    },
    {
      "epoch": 4.45,
      "learning_rate": 7.786259541984733e-05,
      "loss": 0.1346,
      "step": 730
    },
    {
      "epoch": 4.46,
      "learning_rate": 7.783206106870228e-05,
      "loss": 0.1235,
      "step": 731
    },
    {
      "epoch": 4.46,
      "learning_rate": 7.780152671755726e-05,
      "loss": 0.1798,
      "step": 732
    },
    {
      "epoch": 4.47,
      "learning_rate": 7.777099236641223e-05,
      "loss": 0.1263,
      "step": 733
    },
    {
      "epoch": 4.48,
      "learning_rate": 7.774045801526718e-05,
      "loss": 0.1119,
      "step": 734
    },
    {
      "epoch": 4.48,
      "learning_rate": 7.770992366412214e-05,
      "loss": 0.1237,
      "step": 735
    },
    {
      "epoch": 4.49,
      "learning_rate": 7.76793893129771e-05,
      "loss": 0.1956,
      "step": 736
    },
    {
      "epoch": 4.49,
      "learning_rate": 7.764885496183206e-05,
      "loss": 0.1219,
      "step": 737
    },
    {
      "epoch": 4.5,
      "learning_rate": 7.761832061068702e-05,
      "loss": 0.1291,
      "step": 738
    },
    {
      "epoch": 4.51,
      "learning_rate": 7.758778625954199e-05,
      "loss": 0.1209,
      "step": 739
    },
    {
      "epoch": 4.51,
      "learning_rate": 7.755725190839695e-05,
      "loss": 0.1194,
      "step": 740
    },
    {
      "epoch": 4.52,
      "learning_rate": 7.752671755725191e-05,
      "loss": 0.1327,
      "step": 741
    },
    {
      "epoch": 4.52,
      "learning_rate": 7.749618320610688e-05,
      "loss": 0.0975,
      "step": 742
    },
    {
      "epoch": 4.53,
      "learning_rate": 7.746564885496183e-05,
      "loss": 0.1299,
      "step": 743
    },
    {
      "epoch": 4.54,
      "learning_rate": 7.74351145038168e-05,
      "loss": 0.1283,
      "step": 744
    },
    {
      "epoch": 4.54,
      "learning_rate": 7.740458015267176e-05,
      "loss": 0.1333,
      "step": 745
    },
    {
      "epoch": 4.55,
      "learning_rate": 7.737404580152672e-05,
      "loss": 0.1523,
      "step": 746
    },
    {
      "epoch": 4.55,
      "learning_rate": 7.734351145038167e-05,
      "loss": 0.1216,
      "step": 747
    },
    {
      "epoch": 4.56,
      "learning_rate": 7.731297709923665e-05,
      "loss": 0.1559,
      "step": 748
    },
    {
      "epoch": 4.57,
      "learning_rate": 7.728244274809162e-05,
      "loss": 0.1158,
      "step": 749
    },
    {
      "epoch": 4.57,
      "learning_rate": 7.725190839694657e-05,
      "loss": 0.1447,
      "step": 750
    },
    {
      "epoch": 4.58,
      "learning_rate": 7.722137404580153e-05,
      "loss": 0.1392,
      "step": 751
    },
    {
      "epoch": 4.59,
      "learning_rate": 7.71908396946565e-05,
      "loss": 0.152,
      "step": 752
    },
    {
      "epoch": 4.59,
      "learning_rate": 7.716030534351145e-05,
      "loss": 0.1689,
      "step": 753
    },
    {
      "epoch": 4.6,
      "learning_rate": 7.712977099236641e-05,
      "loss": 0.1439,
      "step": 754
    },
    {
      "epoch": 4.6,
      "learning_rate": 7.709923664122138e-05,
      "loss": 0.1261,
      "step": 755
    },
    {
      "epoch": 4.61,
      "learning_rate": 7.706870229007634e-05,
      "loss": 0.1282,
      "step": 756
    },
    {
      "epoch": 4.62,
      "learning_rate": 7.70381679389313e-05,
      "loss": 0.1108,
      "step": 757
    },
    {
      "epoch": 4.62,
      "learning_rate": 7.700763358778627e-05,
      "loss": 0.1155,
      "step": 758
    },
    {
      "epoch": 4.63,
      "learning_rate": 7.697709923664122e-05,
      "loss": 0.1546,
      "step": 759
    },
    {
      "epoch": 4.63,
      "learning_rate": 7.694656488549619e-05,
      "loss": 0.1171,
      "step": 760
    },
    {
      "epoch": 4.64,
      "learning_rate": 7.691603053435115e-05,
      "loss": 0.142,
      "step": 761
    },
    {
      "epoch": 4.65,
      "learning_rate": 7.68854961832061e-05,
      "loss": 0.1198,
      "step": 762
    },
    {
      "epoch": 4.65,
      "learning_rate": 7.685496183206107e-05,
      "loss": 0.1094,
      "step": 763
    },
    {
      "epoch": 4.66,
      "learning_rate": 7.682442748091604e-05,
      "loss": 0.1169,
      "step": 764
    },
    {
      "epoch": 4.66,
      "learning_rate": 7.6793893129771e-05,
      "loss": 0.1479,
      "step": 765
    },
    {
      "epoch": 4.67,
      "learning_rate": 7.676335877862596e-05,
      "loss": 0.1495,
      "step": 766
    },
    {
      "epoch": 4.68,
      "learning_rate": 7.673282442748092e-05,
      "loss": 0.0969,
      "step": 767
    },
    {
      "epoch": 4.68,
      "learning_rate": 7.670229007633587e-05,
      "loss": 0.1505,
      "step": 768
    },
    {
      "epoch": 4.69,
      "learning_rate": 7.667175572519084e-05,
      "loss": 0.1917,
      "step": 769
    },
    {
      "epoch": 4.7,
      "learning_rate": 7.66412213740458e-05,
      "loss": 0.1726,
      "step": 770
    },
    {
      "epoch": 4.7,
      "learning_rate": 7.661068702290077e-05,
      "loss": 0.0947,
      "step": 771
    },
    {
      "epoch": 4.71,
      "learning_rate": 7.658015267175573e-05,
      "loss": 0.1261,
      "step": 772
    },
    {
      "epoch": 4.71,
      "learning_rate": 7.65496183206107e-05,
      "loss": 0.1311,
      "step": 773
    },
    {
      "epoch": 4.72,
      "learning_rate": 7.651908396946565e-05,
      "loss": 0.1348,
      "step": 774
    },
    {
      "epoch": 4.73,
      "learning_rate": 7.648854961832061e-05,
      "loss": 0.1402,
      "step": 775
    },
    {
      "epoch": 4.73,
      "learning_rate": 7.645801526717558e-05,
      "loss": 0.1131,
      "step": 776
    },
    {
      "epoch": 4.74,
      "learning_rate": 7.642748091603054e-05,
      "loss": 0.1529,
      "step": 777
    },
    {
      "epoch": 4.74,
      "learning_rate": 7.639694656488549e-05,
      "loss": 0.1258,
      "step": 778
    },
    {
      "epoch": 4.75,
      "learning_rate": 7.636641221374046e-05,
      "loss": 0.1269,
      "step": 779
    },
    {
      "epoch": 4.76,
      "learning_rate": 7.633587786259542e-05,
      "loss": 0.1792,
      "step": 780
    },
    {
      "epoch": 4.76,
      "learning_rate": 7.630534351145039e-05,
      "loss": 0.1196,
      "step": 781
    },
    {
      "epoch": 4.77,
      "learning_rate": 7.627480916030535e-05,
      "loss": 0.1113,
      "step": 782
    },
    {
      "epoch": 4.77,
      "learning_rate": 7.624427480916032e-05,
      "loss": 0.1413,
      "step": 783
    },
    {
      "epoch": 4.78,
      "learning_rate": 7.621374045801527e-05,
      "loss": 0.1178,
      "step": 784
    },
    {
      "epoch": 4.79,
      "learning_rate": 7.618320610687023e-05,
      "loss": 0.112,
      "step": 785
    },
    {
      "epoch": 4.79,
      "learning_rate": 7.61526717557252e-05,
      "loss": 0.1456,
      "step": 786
    },
    {
      "epoch": 4.8,
      "learning_rate": 7.612213740458015e-05,
      "loss": 0.1319,
      "step": 787
    },
    {
      "epoch": 4.8,
      "learning_rate": 7.609160305343512e-05,
      "loss": 0.1416,
      "step": 788
    },
    {
      "epoch": 4.81,
      "learning_rate": 7.606106870229009e-05,
      "loss": 0.115,
      "step": 789
    },
    {
      "epoch": 4.82,
      "learning_rate": 7.603053435114504e-05,
      "loss": 0.1395,
      "step": 790
    },
    {
      "epoch": 4.82,
      "learning_rate": 7.6e-05,
      "loss": 0.1527,
      "step": 791
    },
    {
      "epoch": 4.83,
      "learning_rate": 7.596946564885497e-05,
      "loss": 0.1326,
      "step": 792
    },
    {
      "epoch": 4.84,
      "learning_rate": 7.593893129770992e-05,
      "loss": 0.1326,
      "step": 793
    },
    {
      "epoch": 4.84,
      "learning_rate": 7.590839694656488e-05,
      "loss": 0.1449,
      "step": 794
    },
    {
      "epoch": 4.85,
      "learning_rate": 7.587786259541985e-05,
      "loss": 0.1399,
      "step": 795
    },
    {
      "epoch": 4.85,
      "learning_rate": 7.584732824427481e-05,
      "loss": 0.1303,
      "step": 796
    },
    {
      "epoch": 4.86,
      "learning_rate": 7.581679389312978e-05,
      "loss": 0.1407,
      "step": 797
    },
    {
      "epoch": 4.87,
      "learning_rate": 7.578625954198474e-05,
      "loss": 0.1497,
      "step": 798
    },
    {
      "epoch": 4.87,
      "learning_rate": 7.575572519083969e-05,
      "loss": 0.1138,
      "step": 799
    },
    {
      "epoch": 4.88,
      "learning_rate": 7.572519083969466e-05,
      "loss": 0.1423,
      "step": 800
    },
    {
      "epoch": 4.88,
      "learning_rate": 7.569465648854962e-05,
      "loss": 0.1821,
      "step": 801
    },
    {
      "epoch": 4.89,
      "learning_rate": 7.566412213740459e-05,
      "loss": 0.1461,
      "step": 802
    },
    {
      "epoch": 4.9,
      "learning_rate": 7.563358778625954e-05,
      "loss": 0.159,
      "step": 803
    },
    {
      "epoch": 4.9,
      "learning_rate": 7.560305343511452e-05,
      "loss": 0.1557,
      "step": 804
    },
    {
      "epoch": 4.91,
      "learning_rate": 7.557251908396947e-05,
      "loss": 0.1186,
      "step": 805
    },
    {
      "epoch": 4.91,
      "learning_rate": 7.554198473282443e-05,
      "loss": 0.1109,
      "step": 806
    },
    {
      "epoch": 4.92,
      "learning_rate": 7.55114503816794e-05,
      "loss": 0.1296,
      "step": 807
    },
    {
      "epoch": 4.93,
      "learning_rate": 7.548091603053436e-05,
      "loss": 0.153,
      "step": 808
    },
    {
      "epoch": 4.93,
      "learning_rate": 7.545038167938931e-05,
      "loss": 0.1287,
      "step": 809
    },
    {
      "epoch": 4.94,
      "learning_rate": 7.541984732824428e-05,
      "loss": 0.1172,
      "step": 810
    },
    {
      "epoch": 4.95,
      "learning_rate": 7.538931297709924e-05,
      "loss": 0.1316,
      "step": 811
    },
    {
      "epoch": 4.95,
      "learning_rate": 7.53587786259542e-05,
      "loss": 0.1374,
      "step": 812
    },
    {
      "epoch": 4.96,
      "learning_rate": 7.532824427480917e-05,
      "loss": 0.1411,
      "step": 813
    },
    {
      "epoch": 4.96,
      "learning_rate": 7.529770992366413e-05,
      "loss": 0.1164,
      "step": 814
    },
    {
      "epoch": 4.97,
      "learning_rate": 7.526717557251908e-05,
      "loss": 0.1306,
      "step": 815
    },
    {
      "epoch": 4.98,
      "learning_rate": 7.523664122137405e-05,
      "loss": 0.1533,
      "step": 816
    },
    {
      "epoch": 4.98,
      "learning_rate": 7.520610687022901e-05,
      "loss": 0.1259,
      "step": 817
    },
    {
      "epoch": 4.99,
      "learning_rate": 7.517557251908396e-05,
      "loss": 0.1246,
      "step": 818
    },
    {
      "epoch": 4.99,
      "learning_rate": 7.514503816793893e-05,
      "loss": 0.1923,
      "step": 819
    },
    {
      "epoch": 5.0,
      "learning_rate": 7.511450381679391e-05,
      "loss": 0.1909,
      "step": 820
    },
    {
      "epoch": 5.01,
      "learning_rate": 7.508396946564886e-05,
      "loss": 0.0819,
      "step": 821
    },
    {
      "epoch": 5.01,
      "learning_rate": 7.505343511450382e-05,
      "loss": 0.1049,
      "step": 822
    },
    {
      "epoch": 5.02,
      "learning_rate": 7.502290076335879e-05,
      "loss": 0.1344,
      "step": 823
    },
    {
      "epoch": 5.02,
      "learning_rate": 7.499236641221374e-05,
      "loss": 0.0956,
      "step": 824
    },
    {
      "epoch": 5.03,
      "learning_rate": 7.49618320610687e-05,
      "loss": 0.0726,
      "step": 825
    },
    {
      "epoch": 5.04,
      "learning_rate": 7.493129770992367e-05,
      "loss": 0.0972,
      "step": 826
    },
    {
      "epoch": 5.04,
      "learning_rate": 7.490076335877863e-05,
      "loss": 0.1555,
      "step": 827
    },
    {
      "epoch": 5.05,
      "learning_rate": 7.48702290076336e-05,
      "loss": 0.0828,
      "step": 828
    },
    {
      "epoch": 5.05,
      "learning_rate": 7.483969465648856e-05,
      "loss": 0.1236,
      "step": 829
    },
    {
      "epoch": 5.06,
      "learning_rate": 7.480916030534351e-05,
      "loss": 0.082,
      "step": 830
    },
    {
      "epoch": 5.07,
      "learning_rate": 7.477862595419848e-05,
      "loss": 0.0909,
      "step": 831
    },
    {
      "epoch": 5.07,
      "learning_rate": 7.474809160305344e-05,
      "loss": 0.0745,
      "step": 832
    },
    {
      "epoch": 5.08,
      "learning_rate": 7.47175572519084e-05,
      "loss": 0.0916,
      "step": 833
    },
    {
      "epoch": 5.09,
      "learning_rate": 7.468702290076336e-05,
      "loss": 0.1289,
      "step": 834
    },
    {
      "epoch": 5.09,
      "learning_rate": 7.465648854961832e-05,
      "loss": 0.1063,
      "step": 835
    },
    {
      "epoch": 5.1,
      "learning_rate": 7.462595419847328e-05,
      "loss": 0.1162,
      "step": 836
    },
    {
      "epoch": 5.1,
      "learning_rate": 7.459541984732825e-05,
      "loss": 0.0862,
      "step": 837
    },
    {
      "epoch": 5.11,
      "learning_rate": 7.456488549618321e-05,
      "loss": 0.1055,
      "step": 838
    },
    {
      "epoch": 5.12,
      "learning_rate": 7.453435114503818e-05,
      "loss": 0.1112,
      "step": 839
    },
    {
      "epoch": 5.12,
      "learning_rate": 7.450381679389313e-05,
      "loss": 0.1385,
      "step": 840
    },
    {
      "epoch": 5.13,
      "learning_rate": 7.44732824427481e-05,
      "loss": 0.0831,
      "step": 841
    },
    {
      "epoch": 5.13,
      "learning_rate": 7.444274809160306e-05,
      "loss": 0.1491,
      "step": 842
    },
    {
      "epoch": 5.14,
      "learning_rate": 7.441221374045801e-05,
      "loss": 0.1084,
      "step": 843
    },
    {
      "epoch": 5.15,
      "learning_rate": 7.438167938931299e-05,
      "loss": 0.1316,
      "step": 844
    },
    {
      "epoch": 5.15,
      "learning_rate": 7.435114503816795e-05,
      "loss": 0.08,
      "step": 845
    },
    {
      "epoch": 5.16,
      "learning_rate": 7.43206106870229e-05,
      "loss": 0.0912,
      "step": 846
    },
    {
      "epoch": 5.16,
      "learning_rate": 7.429007633587787e-05,
      "loss": 0.0889,
      "step": 847
    },
    {
      "epoch": 5.17,
      "learning_rate": 7.425954198473283e-05,
      "loss": 0.1079,
      "step": 848
    },
    {
      "epoch": 5.18,
      "learning_rate": 7.422900763358778e-05,
      "loss": 0.1327,
      "step": 849
    },
    {
      "epoch": 5.18,
      "learning_rate": 7.419847328244275e-05,
      "loss": 0.0906,
      "step": 850
    },
    {
      "epoch": 5.19,
      "learning_rate": 7.416793893129771e-05,
      "loss": 0.0928,
      "step": 851
    },
    {
      "epoch": 5.2,
      "learning_rate": 7.413740458015268e-05,
      "loss": 0.096,
      "step": 852
    },
    {
      "epoch": 5.2,
      "learning_rate": 7.410687022900764e-05,
      "loss": 0.1385,
      "step": 853
    },
    {
      "epoch": 5.21,
      "learning_rate": 7.40763358778626e-05,
      "loss": 0.0942,
      "step": 854
    },
    {
      "epoch": 5.21,
      "learning_rate": 7.404580152671756e-05,
      "loss": 0.1354,
      "step": 855
    },
    {
      "epoch": 5.22,
      "learning_rate": 7.401526717557252e-05,
      "loss": 0.1258,
      "step": 856
    },
    {
      "epoch": 5.23,
      "learning_rate": 7.398473282442749e-05,
      "loss": 0.0992,
      "step": 857
    },
    {
      "epoch": 5.23,
      "learning_rate": 7.395419847328245e-05,
      "loss": 0.1213,
      "step": 858
    },
    {
      "epoch": 5.24,
      "learning_rate": 7.39236641221374e-05,
      "loss": 0.1254,
      "step": 859
    },
    {
      "epoch": 5.24,
      "learning_rate": 7.389312977099238e-05,
      "loss": 0.0863,
      "step": 860
    },
    {
      "epoch": 5.25,
      "learning_rate": 7.386259541984733e-05,
      "loss": 0.0891,
      "step": 861
    },
    {
      "epoch": 5.26,
      "learning_rate": 7.38320610687023e-05,
      "loss": 0.0944,
      "step": 862
    },
    {
      "epoch": 5.26,
      "learning_rate": 7.380152671755726e-05,
      "loss": 0.0702,
      "step": 863
    },
    {
      "epoch": 5.27,
      "learning_rate": 7.377099236641222e-05,
      "loss": 0.1155,
      "step": 864
    },
    {
      "epoch": 5.27,
      "learning_rate": 7.374045801526717e-05,
      "loss": 0.1366,
      "step": 865
    },
    {
      "epoch": 5.28,
      "learning_rate": 7.370992366412214e-05,
      "loss": 0.0856,
      "step": 866
    },
    {
      "epoch": 5.29,
      "learning_rate": 7.36793893129771e-05,
      "loss": 0.0825,
      "step": 867
    },
    {
      "epoch": 5.29,
      "learning_rate": 7.364885496183207e-05,
      "loss": 0.1175,
      "step": 868
    },
    {
      "epoch": 5.3,
      "learning_rate": 7.361832061068703e-05,
      "loss": 0.1063,
      "step": 869
    },
    {
      "epoch": 5.3,
      "learning_rate": 7.3587786259542e-05,
      "loss": 0.0768,
      "step": 870
    },
    {
      "epoch": 5.31,
      "learning_rate": 7.355725190839695e-05,
      "loss": 0.1288,
      "step": 871
    },
    {
      "epoch": 5.32,
      "learning_rate": 7.352671755725191e-05,
      "loss": 0.0953,
      "step": 872
    },
    {
      "epoch": 5.32,
      "learning_rate": 7.349618320610688e-05,
      "loss": 0.1084,
      "step": 873
    },
    {
      "epoch": 5.33,
      "learning_rate": 7.346564885496183e-05,
      "loss": 0.0828,
      "step": 874
    },
    {
      "epoch": 5.34,
      "learning_rate": 7.343511450381679e-05,
      "loss": 0.1042,
      "step": 875
    },
    {
      "epoch": 5.34,
      "learning_rate": 7.340458015267177e-05,
      "loss": 0.1112,
      "step": 876
    },
    {
      "epoch": 5.35,
      "learning_rate": 7.337404580152672e-05,
      "loss": 0.0883,
      "step": 877
    },
    {
      "epoch": 5.35,
      "learning_rate": 7.334351145038169e-05,
      "loss": 0.1547,
      "step": 878
    },
    {
      "epoch": 5.36,
      "learning_rate": 7.331297709923665e-05,
      "loss": 0.0988,
      "step": 879
    },
    {
      "epoch": 5.37,
      "learning_rate": 7.32824427480916e-05,
      "loss": 0.1587,
      "step": 880
    },
    {
      "epoch": 5.37,
      "learning_rate": 7.325190839694657e-05,
      "loss": 0.1241,
      "step": 881
    },
    {
      "epoch": 5.38,
      "learning_rate": 7.322137404580153e-05,
      "loss": 0.0916,
      "step": 882
    },
    {
      "epoch": 5.38,
      "learning_rate": 7.31908396946565e-05,
      "loss": 0.0753,
      "step": 883
    },
    {
      "epoch": 5.39,
      "learning_rate": 7.316030534351146e-05,
      "loss": 0.1248,
      "step": 884
    },
    {
      "epoch": 5.4,
      "learning_rate": 7.312977099236642e-05,
      "loss": 0.0985,
      "step": 885
    },
    {
      "epoch": 5.4,
      "learning_rate": 7.309923664122137e-05,
      "loss": 0.1049,
      "step": 886
    },
    {
      "epoch": 5.41,
      "learning_rate": 7.306870229007634e-05,
      "loss": 0.1167,
      "step": 887
    },
    {
      "epoch": 5.41,
      "learning_rate": 7.30381679389313e-05,
      "loss": 0.1632,
      "step": 888
    },
    {
      "epoch": 5.42,
      "learning_rate": 7.300763358778627e-05,
      "loss": 0.0925,
      "step": 889
    },
    {
      "epoch": 5.43,
      "learning_rate": 7.297709923664122e-05,
      "loss": 0.0952,
      "step": 890
    },
    {
      "epoch": 5.43,
      "learning_rate": 7.294656488549618e-05,
      "loss": 0.0876,
      "step": 891
    },
    {
      "epoch": 5.44,
      "learning_rate": 7.291603053435115e-05,
      "loss": 0.1315,
      "step": 892
    },
    {
      "epoch": 5.45,
      "learning_rate": 7.288549618320611e-05,
      "loss": 0.0943,
      "step": 893
    },
    {
      "epoch": 5.45,
      "learning_rate": 7.285496183206108e-05,
      "loss": 0.1081,
      "step": 894
    },
    {
      "epoch": 5.46,
      "learning_rate": 7.282442748091604e-05,
      "loss": 0.0865,
      "step": 895
    },
    {
      "epoch": 5.46,
      "learning_rate": 7.279389312977099e-05,
      "loss": 0.1096,
      "step": 896
    },
    {
      "epoch": 5.47,
      "learning_rate": 7.276335877862596e-05,
      "loss": 0.1161,
      "step": 897
    },
    {
      "epoch": 5.48,
      "learning_rate": 7.273282442748092e-05,
      "loss": 0.1111,
      "step": 898
    },
    {
      "epoch": 5.48,
      "learning_rate": 7.270229007633587e-05,
      "loss": 0.088,
      "step": 899
    },
    {
      "epoch": 5.49,
      "learning_rate": 7.267175572519084e-05,
      "loss": 0.119,
      "step": 900
    },
    {
      "epoch": 5.49,
      "learning_rate": 7.264122137404582e-05,
      "loss": 0.1121,
      "step": 901
    },
    {
      "epoch": 5.5,
      "learning_rate": 7.261068702290077e-05,
      "loss": 0.0804,
      "step": 902
    },
    {
      "epoch": 5.51,
      "learning_rate": 7.258015267175573e-05,
      "loss": 0.0927,
      "step": 903
    },
    {
      "epoch": 5.51,
      "learning_rate": 7.25496183206107e-05,
      "loss": 0.1183,
      "step": 904
    },
    {
      "epoch": 5.52,
      "learning_rate": 7.251908396946565e-05,
      "loss": 0.0907,
      "step": 905
    },
    {
      "epoch": 5.52,
      "learning_rate": 7.248854961832061e-05,
      "loss": 0.1035,
      "step": 906
    },
    {
      "epoch": 5.53,
      "learning_rate": 7.245801526717557e-05,
      "loss": 0.1297,
      "step": 907
    },
    {
      "epoch": 5.54,
      "learning_rate": 7.242748091603053e-05,
      "loss": 0.1023,
      "step": 908
    },
    {
      "epoch": 5.54,
      "learning_rate": 7.23969465648855e-05,
      "loss": 0.129,
      "step": 909
    },
    {
      "epoch": 5.55,
      "learning_rate": 7.236641221374047e-05,
      "loss": 0.1109,
      "step": 910
    },
    {
      "epoch": 5.55,
      "learning_rate": 7.233587786259542e-05,
      "loss": 0.1515,
      "step": 911
    },
    {
      "epoch": 5.56,
      "learning_rate": 7.230534351145038e-05,
      "loss": 0.112,
      "step": 912
    },
    {
      "epoch": 5.57,
      "learning_rate": 7.227480916030535e-05,
      "loss": 0.1213,
      "step": 913
    },
    {
      "epoch": 5.57,
      "learning_rate": 7.22442748091603e-05,
      "loss": 0.1067,
      "step": 914
    },
    {
      "epoch": 5.58,
      "learning_rate": 7.221374045801526e-05,
      "loss": 0.0884,
      "step": 915
    },
    {
      "epoch": 5.59,
      "learning_rate": 7.218320610687023e-05,
      "loss": 0.1239,
      "step": 916
    },
    {
      "epoch": 5.59,
      "learning_rate": 7.215267175572519e-05,
      "loss": 0.0735,
      "step": 917
    },
    {
      "epoch": 5.6,
      "learning_rate": 7.212213740458016e-05,
      "loss": 0.1659,
      "step": 918
    },
    {
      "epoch": 5.6,
      "learning_rate": 7.209160305343512e-05,
      "loss": 0.1021,
      "step": 919
    },
    {
      "epoch": 5.61,
      "learning_rate": 7.206106870229009e-05,
      "loss": 0.1378,
      "step": 920
    },
    {
      "epoch": 5.62,
      "learning_rate": 7.203053435114504e-05,
      "loss": 0.0973,
      "step": 921
    },
    {
      "epoch": 5.62,
      "learning_rate": 7.2e-05,
      "loss": 0.0937,
      "step": 922
    },
    {
      "epoch": 5.63,
      "learning_rate": 7.196946564885497e-05,
      "loss": 0.0998,
      "step": 923
    },
    {
      "epoch": 5.63,
      "learning_rate": 7.193893129770992e-05,
      "loss": 0.127,
      "step": 924
    },
    {
      "epoch": 5.64,
      "learning_rate": 7.19083969465649e-05,
      "loss": 0.0971,
      "step": 925
    },
    {
      "epoch": 5.65,
      "learning_rate": 7.187786259541986e-05,
      "loss": 0.1076,
      "step": 926
    },
    {
      "epoch": 5.65,
      "learning_rate": 7.184732824427481e-05,
      "loss": 0.1307,
      "step": 927
    },
    {
      "epoch": 5.66,
      "learning_rate": 7.181679389312978e-05,
      "loss": 0.0962,
      "step": 928
    },
    {
      "epoch": 5.66,
      "learning_rate": 7.178625954198474e-05,
      "loss": 0.1198,
      "step": 929
    },
    {
      "epoch": 5.67,
      "learning_rate": 7.175572519083969e-05,
      "loss": 0.1035,
      "step": 930
    },
    {
      "epoch": 5.68,
      "learning_rate": 7.172519083969466e-05,
      "loss": 0.1162,
      "step": 931
    },
    {
      "epoch": 5.68,
      "learning_rate": 7.169465648854962e-05,
      "loss": 0.111,
      "step": 932
    },
    {
      "epoch": 5.69,
      "learning_rate": 7.166412213740458e-05,
      "loss": 0.1019,
      "step": 933
    },
    {
      "epoch": 5.7,
      "learning_rate": 7.163358778625955e-05,
      "loss": 0.0922,
      "step": 934
    },
    {
      "epoch": 5.7,
      "learning_rate": 7.160305343511451e-05,
      "loss": 0.1112,
      "step": 935
    },
    {
      "epoch": 5.71,
      "learning_rate": 7.157251908396946e-05,
      "loss": 0.0834,
      "step": 936
    },
    {
      "epoch": 5.71,
      "learning_rate": 7.154198473282443e-05,
      "loss": 0.0943,
      "step": 937
    },
    {
      "epoch": 5.72,
      "learning_rate": 7.15114503816794e-05,
      "loss": 0.1189,
      "step": 938
    },
    {
      "epoch": 5.73,
      "learning_rate": 7.148091603053434e-05,
      "loss": 0.1038,
      "step": 939
    },
    {
      "epoch": 5.73,
      "learning_rate": 7.145038167938931e-05,
      "loss": 0.0876,
      "step": 940
    },
    {
      "epoch": 5.74,
      "learning_rate": 7.141984732824429e-05,
      "loss": 0.1093,
      "step": 941
    },
    {
      "epoch": 5.74,
      "learning_rate": 7.138931297709924e-05,
      "loss": 0.1378,
      "step": 942
    },
    {
      "epoch": 5.75,
      "learning_rate": 7.13587786259542e-05,
      "loss": 0.1205,
      "step": 943
    },
    {
      "epoch": 5.76,
      "learning_rate": 7.132824427480917e-05,
      "loss": 0.0961,
      "step": 944
    },
    {
      "epoch": 5.76,
      "learning_rate": 7.129770992366412e-05,
      "loss": 0.1192,
      "step": 945
    },
    {
      "epoch": 5.77,
      "learning_rate": 7.126717557251908e-05,
      "loss": 0.1077,
      "step": 946
    },
    {
      "epoch": 5.77,
      "learning_rate": 7.123664122137405e-05,
      "loss": 0.136,
      "step": 947
    },
    {
      "epoch": 5.78,
      "learning_rate": 7.120610687022901e-05,
      "loss": 0.1121,
      "step": 948
    },
    {
      "epoch": 5.79,
      "learning_rate": 7.117557251908398e-05,
      "loss": 0.1309,
      "step": 949
    },
    {
      "epoch": 5.79,
      "learning_rate": 7.114503816793894e-05,
      "loss": 0.0983,
      "step": 950
    },
    {
      "epoch": 5.8,
      "learning_rate": 7.11145038167939e-05,
      "loss": 0.1186,
      "step": 951
    },
    {
      "epoch": 5.8,
      "learning_rate": 7.108396946564886e-05,
      "loss": 0.0993,
      "step": 952
    },
    {
      "epoch": 5.81,
      "learning_rate": 7.105343511450382e-05,
      "loss": 0.102,
      "step": 953
    },
    {
      "epoch": 5.82,
      "learning_rate": 7.102290076335878e-05,
      "loss": 0.1127,
      "step": 954
    },
    {
      "epoch": 5.82,
      "learning_rate": 7.099236641221374e-05,
      "loss": 0.1399,
      "step": 955
    },
    {
      "epoch": 5.83,
      "learning_rate": 7.09618320610687e-05,
      "loss": 0.0982,
      "step": 956
    },
    {
      "epoch": 5.84,
      "learning_rate": 7.093129770992368e-05,
      "loss": 0.0938,
      "step": 957
    },
    {
      "epoch": 5.84,
      "learning_rate": 7.090076335877863e-05,
      "loss": 0.0811,
      "step": 958
    },
    {
      "epoch": 5.85,
      "learning_rate": 7.08702290076336e-05,
      "loss": 0.0969,
      "step": 959
    },
    {
      "epoch": 5.85,
      "learning_rate": 7.083969465648856e-05,
      "loss": 0.1141,
      "step": 960
    },
    {
      "epoch": 5.86,
      "learning_rate": 7.080916030534351e-05,
      "loss": 0.1441,
      "step": 961
    },
    {
      "epoch": 5.87,
      "learning_rate": 7.077862595419847e-05,
      "loss": 0.0799,
      "step": 962
    },
    {
      "epoch": 5.87,
      "learning_rate": 7.074809160305344e-05,
      "loss": 0.1521,
      "step": 963
    },
    {
      "epoch": 5.88,
      "learning_rate": 7.071755725190839e-05,
      "loss": 0.1032,
      "step": 964
    },
    {
      "epoch": 5.88,
      "learning_rate": 7.068702290076337e-05,
      "loss": 0.0923,
      "step": 965
    },
    {
      "epoch": 5.89,
      "learning_rate": 7.065648854961833e-05,
      "loss": 0.0798,
      "step": 966
    },
    {
      "epoch": 5.9,
      "learning_rate": 7.062595419847328e-05,
      "loss": 0.093,
      "step": 967
    },
    {
      "epoch": 5.9,
      "learning_rate": 7.059541984732825e-05,
      "loss": 0.0877,
      "step": 968
    },
    {
      "epoch": 5.91,
      "learning_rate": 7.056488549618321e-05,
      "loss": 0.135,
      "step": 969
    },
    {
      "epoch": 5.91,
      "learning_rate": 7.053435114503816e-05,
      "loss": 0.0833,
      "step": 970
    },
    {
      "epoch": 5.92,
      "learning_rate": 7.050381679389313e-05,
      "loss": 0.1073,
      "step": 971
    },
    {
      "epoch": 5.93,
      "learning_rate": 7.047328244274809e-05,
      "loss": 0.0905,
      "step": 972
    },
    {
      "epoch": 5.93,
      "learning_rate": 7.044274809160306e-05,
      "loss": 0.1121,
      "step": 973
    },
    {
      "epoch": 5.94,
      "learning_rate": 7.041221374045802e-05,
      "loss": 0.1359,
      "step": 974
    },
    {
      "epoch": 5.95,
      "learning_rate": 7.038167938931299e-05,
      "loss": 0.1312,
      "step": 975
    },
    {
      "epoch": 5.95,
      "learning_rate": 7.035114503816794e-05,
      "loss": 0.1261,
      "step": 976
    },
    {
      "epoch": 5.96,
      "learning_rate": 7.03206106870229e-05,
      "loss": 0.1203,
      "step": 977
    },
    {
      "epoch": 5.96,
      "learning_rate": 7.029007633587786e-05,
      "loss": 0.0899,
      "step": 978
    },
    {
      "epoch": 5.97,
      "learning_rate": 7.025954198473283e-05,
      "loss": 0.1452,
      "step": 979
    },
    {
      "epoch": 5.98,
      "learning_rate": 7.022900763358778e-05,
      "loss": 0.1006,
      "step": 980
    },
    {
      "epoch": 5.98,
      "learning_rate": 7.019847328244276e-05,
      "loss": 0.1183,
      "step": 981
    },
    {
      "epoch": 5.99,
      "learning_rate": 7.016793893129772e-05,
      "loss": 0.0917,
      "step": 982
    },
    {
      "epoch": 5.99,
      "learning_rate": 7.013740458015267e-05,
      "loss": 0.0932,
      "step": 983
    },
    {
      "epoch": 6.0,
      "learning_rate": 7.010687022900764e-05,
      "loss": 0.0653,
      "step": 984
    },
    {
      "epoch": 6.01,
      "learning_rate": 7.00763358778626e-05,
      "loss": 0.1097,
      "step": 985
    },
    {
      "epoch": 6.01,
      "learning_rate": 7.004580152671755e-05,
      "loss": 0.0858,
      "step": 986
    },
    {
      "epoch": 6.02,
      "learning_rate": 7.001526717557252e-05,
      "loss": 0.0853,
      "step": 987
    },
    {
      "epoch": 6.02,
      "learning_rate": 6.998473282442748e-05,
      "loss": 0.075,
      "step": 988
    },
    {
      "epoch": 6.03,
      "learning_rate": 6.995419847328245e-05,
      "loss": 0.1118,
      "step": 989
    },
    {
      "epoch": 6.04,
      "learning_rate": 6.992366412213741e-05,
      "loss": 0.0788,
      "step": 990
    },
    {
      "epoch": 6.04,
      "learning_rate": 6.989312977099238e-05,
      "loss": 0.0784,
      "step": 991
    },
    {
      "epoch": 6.05,
      "learning_rate": 6.986259541984733e-05,
      "loss": 0.0808,
      "step": 992
    },
    {
      "epoch": 6.05,
      "learning_rate": 6.983206106870229e-05,
      "loss": 0.0646,
      "step": 993
    },
    {
      "epoch": 6.06,
      "learning_rate": 6.980152671755726e-05,
      "loss": 0.0604,
      "step": 994
    },
    {
      "epoch": 6.07,
      "learning_rate": 6.977099236641221e-05,
      "loss": 0.06,
      "step": 995
    },
    {
      "epoch": 6.07,
      "learning_rate": 6.974045801526717e-05,
      "loss": 0.0894,
      "step": 996
    },
    {
      "epoch": 6.08,
      "learning_rate": 6.970992366412215e-05,
      "loss": 0.0658,
      "step": 997
    },
    {
      "epoch": 6.09,
      "learning_rate": 6.96793893129771e-05,
      "loss": 0.077,
      "step": 998
    },
    {
      "epoch": 6.09,
      "learning_rate": 6.964885496183207e-05,
      "loss": 0.1324,
      "step": 999
    },
    {
      "epoch": 6.1,
      "learning_rate": 6.961832061068703e-05,
      "loss": 0.1029,
      "step": 1000
    },
    {
      "epoch": 6.1,
      "learning_rate": 6.958778625954198e-05,
      "loss": 0.0899,
      "step": 1001
    },
    {
      "epoch": 6.11,
      "learning_rate": 6.955725190839695e-05,
      "loss": 0.0989,
      "step": 1002
    },
    {
      "epoch": 6.12,
      "learning_rate": 6.952671755725191e-05,
      "loss": 0.0834,
      "step": 1003
    },
    {
      "epoch": 6.12,
      "learning_rate": 6.949618320610687e-05,
      "loss": 0.0833,
      "step": 1004
    },
    {
      "epoch": 6.13,
      "learning_rate": 6.946564885496184e-05,
      "loss": 0.0875,
      "step": 1005
    },
    {
      "epoch": 6.13,
      "learning_rate": 6.94351145038168e-05,
      "loss": 0.0683,
      "step": 1006
    },
    {
      "epoch": 6.14,
      "learning_rate": 6.940458015267175e-05,
      "loss": 0.0766,
      "step": 1007
    },
    {
      "epoch": 6.15,
      "learning_rate": 6.937404580152672e-05,
      "loss": 0.0913,
      "step": 1008
    },
    {
      "epoch": 6.15,
      "learning_rate": 6.934351145038168e-05,
      "loss": 0.0774,
      "step": 1009
    },
    {
      "epoch": 6.16,
      "learning_rate": 6.931297709923665e-05,
      "loss": 0.0669,
      "step": 1010
    },
    {
      "epoch": 6.16,
      "learning_rate": 6.92824427480916e-05,
      "loss": 0.1113,
      "step": 1011
    },
    {
      "epoch": 6.17,
      "learning_rate": 6.925190839694656e-05,
      "loss": 0.0882,
      "step": 1012
    },
    {
      "epoch": 6.18,
      "learning_rate": 6.922137404580154e-05,
      "loss": 0.0729,
      "step": 1013
    },
    {
      "epoch": 6.18,
      "learning_rate": 6.919083969465649e-05,
      "loss": 0.0563,
      "step": 1014
    },
    {
      "epoch": 6.19,
      "learning_rate": 6.916030534351146e-05,
      "loss": 0.0758,
      "step": 1015
    },
    {
      "epoch": 6.2,
      "learning_rate": 6.912977099236642e-05,
      "loss": 0.0765,
      "step": 1016
    },
    {
      "epoch": 6.2,
      "learning_rate": 6.909923664122137e-05,
      "loss": 0.106,
      "step": 1017
    },
    {
      "epoch": 6.21,
      "learning_rate": 6.906870229007634e-05,
      "loss": 0.0644,
      "step": 1018
    },
    {
      "epoch": 6.21,
      "learning_rate": 6.90381679389313e-05,
      "loss": 0.1185,
      "step": 1019
    },
    {
      "epoch": 6.22,
      "learning_rate": 6.900763358778625e-05,
      "loss": 0.093,
      "step": 1020
    },
    {
      "epoch": 6.23,
      "learning_rate": 6.897709923664123e-05,
      "loss": 0.151,
      "step": 1021
    },
    {
      "epoch": 6.23,
      "learning_rate": 6.89465648854962e-05,
      "loss": 0.1239,
      "step": 1022
    },
    {
      "epoch": 6.24,
      "learning_rate": 6.891603053435115e-05,
      "loss": 0.1231,
      "step": 1023
    },
    {
      "epoch": 6.24,
      "learning_rate": 6.888549618320611e-05,
      "loss": 0.0779,
      "step": 1024
    },
    {
      "epoch": 6.25,
      "learning_rate": 6.885496183206107e-05,
      "loss": 0.07,
      "step": 1025
    },
    {
      "epoch": 6.26,
      "learning_rate": 6.882442748091603e-05,
      "loss": 0.0895,
      "step": 1026
    },
    {
      "epoch": 6.26,
      "learning_rate": 6.879389312977099e-05,
      "loss": 0.0852,
      "step": 1027
    },
    {
      "epoch": 6.27,
      "learning_rate": 6.876335877862595e-05,
      "loss": 0.102,
      "step": 1028
    },
    {
      "epoch": 6.27,
      "learning_rate": 6.873282442748092e-05,
      "loss": 0.0809,
      "step": 1029
    },
    {
      "epoch": 6.28,
      "learning_rate": 6.870229007633588e-05,
      "loss": 0.0978,
      "step": 1030
    },
    {
      "epoch": 6.29,
      "learning_rate": 6.867175572519085e-05,
      "loss": 0.0783,
      "step": 1031
    },
    {
      "epoch": 6.29,
      "learning_rate": 6.86412213740458e-05,
      "loss": 0.0885,
      "step": 1032
    },
    {
      "epoch": 6.3,
      "learning_rate": 6.861068702290076e-05,
      "loss": 0.0868,
      "step": 1033
    },
    {
      "epoch": 6.3,
      "learning_rate": 6.858015267175573e-05,
      "loss": 0.0753,
      "step": 1034
    },
    {
      "epoch": 6.31,
      "learning_rate": 6.854961832061069e-05,
      "loss": 0.0816,
      "step": 1035
    },
    {
      "epoch": 6.32,
      "learning_rate": 6.851908396946564e-05,
      "loss": 0.0864,
      "step": 1036
    },
    {
      "epoch": 6.32,
      "learning_rate": 6.848854961832062e-05,
      "loss": 0.0811,
      "step": 1037
    },
    {
      "epoch": 6.33,
      "learning_rate": 6.845801526717557e-05,
      "loss": 0.0989,
      "step": 1038
    },
    {
      "epoch": 6.34,
      "learning_rate": 6.842748091603054e-05,
      "loss": 0.0617,
      "step": 1039
    },
    {
      "epoch": 6.34,
      "learning_rate": 6.83969465648855e-05,
      "loss": 0.0843,
      "step": 1040
    },
    {
      "epoch": 6.35,
      "learning_rate": 6.836641221374047e-05,
      "loss": 0.0952,
      "step": 1041
    },
    {
      "epoch": 6.35,
      "learning_rate": 6.833587786259542e-05,
      "loss": 0.0907,
      "step": 1042
    },
    {
      "epoch": 6.36,
      "learning_rate": 6.830534351145038e-05,
      "loss": 0.1021,
      "step": 1043
    },
    {
      "epoch": 6.37,
      "learning_rate": 6.827480916030535e-05,
      "loss": 0.091,
      "step": 1044
    },
    {
      "epoch": 6.37,
      "learning_rate": 6.824427480916031e-05,
      "loss": 0.1007,
      "step": 1045
    },
    {
      "epoch": 6.38,
      "learning_rate": 6.821374045801528e-05,
      "loss": 0.0634,
      "step": 1046
    },
    {
      "epoch": 6.38,
      "learning_rate": 6.818320610687024e-05,
      "loss": 0.1118,
      "step": 1047
    },
    {
      "epoch": 6.39,
      "learning_rate": 6.815267175572519e-05,
      "loss": 0.0862,
      "step": 1048
    },
    {
      "epoch": 6.4,
      "learning_rate": 6.812213740458016e-05,
      "loss": 0.0805,
      "step": 1049
    },
    {
      "epoch": 6.4,
      "learning_rate": 6.809160305343512e-05,
      "loss": 0.0865,
      "step": 1050
    },
    {
      "epoch": 6.41,
      "learning_rate": 6.806106870229007e-05,
      "loss": 0.1031,
      "step": 1051
    },
    {
      "epoch": 6.41,
      "learning_rate": 6.803053435114503e-05,
      "loss": 0.0865,
      "step": 1052
    },
    {
      "epoch": 6.42,
      "learning_rate": 6.800000000000001e-05,
      "loss": 0.1419,
      "step": 1053
    },
    {
      "epoch": 6.43,
      "learning_rate": 6.796946564885496e-05,
      "loss": 0.0808,
      "step": 1054
    },
    {
      "epoch": 6.43,
      "learning_rate": 6.793893129770993e-05,
      "loss": 0.0689,
      "step": 1055
    },
    {
      "epoch": 6.44,
      "learning_rate": 6.790839694656489e-05,
      "loss": 0.0804,
      "step": 1056
    },
    {
      "epoch": 6.45,
      "learning_rate": 6.787786259541984e-05,
      "loss": 0.1268,
      "step": 1057
    },
    {
      "epoch": 6.45,
      "learning_rate": 6.784732824427481e-05,
      "loss": 0.0851,
      "step": 1058
    },
    {
      "epoch": 6.46,
      "learning_rate": 6.781679389312977e-05,
      "loss": 0.1143,
      "step": 1059
    },
    {
      "epoch": 6.46,
      "learning_rate": 6.778625954198474e-05,
      "loss": 0.066,
      "step": 1060
    },
    {
      "epoch": 6.47,
      "learning_rate": 6.77557251908397e-05,
      "loss": 0.0882,
      "step": 1061
    },
    {
      "epoch": 6.48,
      "learning_rate": 6.772519083969467e-05,
      "loss": 0.0764,
      "step": 1062
    },
    {
      "epoch": 6.48,
      "learning_rate": 6.769465648854962e-05,
      "loss": 0.0662,
      "step": 1063
    },
    {
      "epoch": 6.49,
      "learning_rate": 6.766412213740458e-05,
      "loss": 0.0966,
      "step": 1064
    },
    {
      "epoch": 6.49,
      "learning_rate": 6.763358778625955e-05,
      "loss": 0.095,
      "step": 1065
    },
    {
      "epoch": 6.5,
      "learning_rate": 6.760305343511451e-05,
      "loss": 0.0797,
      "step": 1066
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.757251908396946e-05,
      "loss": 0.1145,
      "step": 1067
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.754198473282443e-05,
      "loss": 0.0672,
      "step": 1068
    },
    {
      "epoch": 6.52,
      "learning_rate": 6.751145038167939e-05,
      "loss": 0.1368,
      "step": 1069
    },
    {
      "epoch": 6.52,
      "learning_rate": 6.748091603053436e-05,
      "loss": 0.1092,
      "step": 1070
    },
    {
      "epoch": 6.53,
      "learning_rate": 6.745038167938932e-05,
      "loss": 0.0832,
      "step": 1071
    },
    {
      "epoch": 6.54,
      "learning_rate": 6.741984732824428e-05,
      "loss": 0.074,
      "step": 1072
    },
    {
      "epoch": 6.54,
      "learning_rate": 6.738931297709924e-05,
      "loss": 0.0929,
      "step": 1073
    },
    {
      "epoch": 6.55,
      "learning_rate": 6.73587786259542e-05,
      "loss": 0.0966,
      "step": 1074
    },
    {
      "epoch": 6.55,
      "learning_rate": 6.732824427480916e-05,
      "loss": 0.0825,
      "step": 1075
    },
    {
      "epoch": 6.56,
      "learning_rate": 6.729770992366412e-05,
      "loss": 0.0649,
      "step": 1076
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.72671755725191e-05,
      "loss": 0.0893,
      "step": 1077
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.723664122137406e-05,
      "loss": 0.0649,
      "step": 1078
    },
    {
      "epoch": 6.58,
      "learning_rate": 6.720610687022901e-05,
      "loss": 0.0858,
      "step": 1079
    },
    {
      "epoch": 6.59,
      "learning_rate": 6.717557251908397e-05,
      "loss": 0.0839,
      "step": 1080
    },
    {
      "epoch": 6.59,
      "learning_rate": 6.714503816793894e-05,
      "loss": 0.0941,
      "step": 1081
    },
    {
      "epoch": 6.6,
      "learning_rate": 6.711450381679389e-05,
      "loss": 0.1106,
      "step": 1082
    },
    {
      "epoch": 6.6,
      "learning_rate": 6.708396946564885e-05,
      "loss": 0.0655,
      "step": 1083
    },
    {
      "epoch": 6.61,
      "learning_rate": 6.705343511450382e-05,
      "loss": 0.0989,
      "step": 1084
    },
    {
      "epoch": 6.62,
      "learning_rate": 6.702290076335878e-05,
      "loss": 0.0809,
      "step": 1085
    },
    {
      "epoch": 6.62,
      "learning_rate": 6.699236641221375e-05,
      "loss": 0.0909,
      "step": 1086
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.696183206106871e-05,
      "loss": 0.0782,
      "step": 1087
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.693129770992366e-05,
      "loss": 0.087,
      "step": 1088
    },
    {
      "epoch": 6.64,
      "learning_rate": 6.690076335877863e-05,
      "loss": 0.1032,
      "step": 1089
    },
    {
      "epoch": 6.65,
      "learning_rate": 6.687022900763359e-05,
      "loss": 0.0979,
      "step": 1090
    },
    {
      "epoch": 6.65,
      "learning_rate": 6.683969465648856e-05,
      "loss": 0.0849,
      "step": 1091
    },
    {
      "epoch": 6.66,
      "learning_rate": 6.68091603053435e-05,
      "loss": 0.0794,
      "step": 1092
    },
    {
      "epoch": 6.66,
      "learning_rate": 6.677862595419848e-05,
      "loss": 0.0971,
      "step": 1093
    },
    {
      "epoch": 6.67,
      "learning_rate": 6.674809160305344e-05,
      "loss": 0.1152,
      "step": 1094
    },
    {
      "epoch": 6.68,
      "learning_rate": 6.67175572519084e-05,
      "loss": 0.0833,
      "step": 1095
    },
    {
      "epoch": 6.68,
      "learning_rate": 6.668702290076336e-05,
      "loss": 0.1035,
      "step": 1096
    },
    {
      "epoch": 6.69,
      "learning_rate": 6.665648854961833e-05,
      "loss": 0.0858,
      "step": 1097
    },
    {
      "epoch": 6.7,
      "learning_rate": 6.662595419847328e-05,
      "loss": 0.0896,
      "step": 1098
    },
    {
      "epoch": 6.7,
      "learning_rate": 6.659541984732824e-05,
      "loss": 0.0831,
      "step": 1099
    },
    {
      "epoch": 6.71,
      "learning_rate": 6.656488549618321e-05,
      "loss": 0.1279,
      "step": 1100
    },
    {
      "epoch": 6.71,
      "learning_rate": 6.653435114503817e-05,
      "loss": 0.0719,
      "step": 1101
    },
    {
      "epoch": 6.72,
      "learning_rate": 6.650381679389314e-05,
      "loss": 0.1141,
      "step": 1102
    },
    {
      "epoch": 6.73,
      "learning_rate": 6.64732824427481e-05,
      "loss": 0.078,
      "step": 1103
    },
    {
      "epoch": 6.73,
      "learning_rate": 6.644274809160305e-05,
      "loss": 0.0725,
      "step": 1104
    },
    {
      "epoch": 6.74,
      "learning_rate": 6.641221374045802e-05,
      "loss": 0.1037,
      "step": 1105
    },
    {
      "epoch": 6.74,
      "learning_rate": 6.638167938931298e-05,
      "loss": 0.0789,
      "step": 1106
    },
    {
      "epoch": 6.75,
      "learning_rate": 6.635114503816793e-05,
      "loss": 0.0869,
      "step": 1107
    },
    {
      "epoch": 6.76,
      "learning_rate": 6.63206106870229e-05,
      "loss": 0.1058,
      "step": 1108
    },
    {
      "epoch": 6.76,
      "learning_rate": 6.629007633587788e-05,
      "loss": 0.0858,
      "step": 1109
    },
    {
      "epoch": 6.77,
      "learning_rate": 6.625954198473283e-05,
      "loss": 0.1004,
      "step": 1110
    },
    {
      "epoch": 6.77,
      "learning_rate": 6.622900763358779e-05,
      "loss": 0.0792,
      "step": 1111
    },
    {
      "epoch": 6.78,
      "learning_rate": 6.619847328244276e-05,
      "loss": 0.0969,
      "step": 1112
    },
    {
      "epoch": 6.79,
      "learning_rate": 6.616793893129771e-05,
      "loss": 0.0776,
      "step": 1113
    },
    {
      "epoch": 6.79,
      "learning_rate": 6.613740458015267e-05,
      "loss": 0.082,
      "step": 1114
    },
    {
      "epoch": 6.8,
      "learning_rate": 6.610687022900764e-05,
      "loss": 0.0764,
      "step": 1115
    },
    {
      "epoch": 6.8,
      "learning_rate": 6.60763358778626e-05,
      "loss": 0.0716,
      "step": 1116
    },
    {
      "epoch": 6.81,
      "learning_rate": 6.604580152671757e-05,
      "loss": 0.0752,
      "step": 1117
    },
    {
      "epoch": 6.82,
      "learning_rate": 6.601526717557253e-05,
      "loss": 0.0926,
      "step": 1118
    },
    {
      "epoch": 6.82,
      "learning_rate": 6.598473282442748e-05,
      "loss": 0.1161,
      "step": 1119
    },
    {
      "epoch": 6.83,
      "learning_rate": 6.595419847328245e-05,
      "loss": 0.0794,
      "step": 1120
    },
    {
      "epoch": 6.84,
      "learning_rate": 6.592366412213741e-05,
      "loss": 0.155,
      "step": 1121
    },
    {
      "epoch": 6.84,
      "learning_rate": 6.589312977099237e-05,
      "loss": 0.0868,
      "step": 1122
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.586259541984732e-05,
      "loss": 0.1169,
      "step": 1123
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.583206106870229e-05,
      "loss": 0.0839,
      "step": 1124
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.580152671755725e-05,
      "loss": 0.102,
      "step": 1125
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.577099236641222e-05,
      "loss": 0.0573,
      "step": 1126
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.574045801526718e-05,
      "loss": 0.0985,
      "step": 1127
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.570992366412215e-05,
      "loss": 0.1003,
      "step": 1128
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.56793893129771e-05,
      "loss": 0.0916,
      "step": 1129
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.564885496183206e-05,
      "loss": 0.0699,
      "step": 1130
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.561832061068703e-05,
      "loss": 0.0883,
      "step": 1131
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.558778625954198e-05,
      "loss": 0.0741,
      "step": 1132
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.555725190839696e-05,
      "loss": 0.0811,
      "step": 1133
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.552671755725192e-05,
      "loss": 0.0764,
      "step": 1134
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.549618320610687e-05,
      "loss": 0.1017,
      "step": 1135
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.546564885496184e-05,
      "loss": 0.1179,
      "step": 1136
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.54351145038168e-05,
      "loss": 0.0887,
      "step": 1137
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.540458015267175e-05,
      "loss": 0.0783,
      "step": 1138
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.537404580152672e-05,
      "loss": 0.0884,
      "step": 1139
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.534351145038168e-05,
      "loss": 0.0819,
      "step": 1140
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.531297709923665e-05,
      "loss": 0.11,
      "step": 1141
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.528244274809161e-05,
      "loss": 0.076,
      "step": 1142
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.525190839694657e-05,
      "loss": 0.0971,
      "step": 1143
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.522137404580153e-05,
      "loss": 0.1057,
      "step": 1144
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.519083969465649e-05,
      "loss": 0.0865,
      "step": 1145
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.516030534351145e-05,
      "loss": 0.0965,
      "step": 1146
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.512977099236642e-05,
      "loss": 0.1045,
      "step": 1147
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.509923664122137e-05,
      "loss": 0.0749,
      "step": 1148
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.506870229007635e-05,
      "loss": 0.0823,
      "step": 1149
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.50381679389313e-05,
      "loss": 0.0707,
      "step": 1150
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.500763358778626e-05,
      "loss": 0.0614,
      "step": 1151
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.497709923664123e-05,
      "loss": 0.0625,
      "step": 1152
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.494656488549619e-05,
      "loss": 0.1018,
      "step": 1153
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.491603053435114e-05,
      "loss": 0.0854,
      "step": 1154
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.488549618320611e-05,
      "loss": 0.071,
      "step": 1155
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.485496183206107e-05,
      "loss": 0.0766,
      "step": 1156
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.482442748091604e-05,
      "loss": 0.09,
      "step": 1157
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.4793893129771e-05,
      "loss": 0.0943,
      "step": 1158
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.476335877862597e-05,
      "loss": 0.0763,
      "step": 1159
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.473282442748092e-05,
      "loss": 0.0704,
      "step": 1160
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.470229007633588e-05,
      "loss": 0.0597,
      "step": 1161
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.467175572519085e-05,
      "loss": 0.0732,
      "step": 1162
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.46412213740458e-05,
      "loss": 0.0568,
      "step": 1163
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.461068702290076e-05,
      "loss": 0.054,
      "step": 1164
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.458015267175574e-05,
      "loss": 0.0875,
      "step": 1165
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.454961832061069e-05,
      "loss": 0.0692,
      "step": 1166
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.451908396946565e-05,
      "loss": 0.0636,
      "step": 1167
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.448854961832062e-05,
      "loss": 0.0776,
      "step": 1168
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.445801526717557e-05,
      "loss": 0.0979,
      "step": 1169
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.442748091603053e-05,
      "loss": 0.0626,
      "step": 1170
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.43969465648855e-05,
      "loss": 0.0686,
      "step": 1171
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.436641221374045e-05,
      "loss": 0.0944,
      "step": 1172
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.433587786259543e-05,
      "loss": 0.095,
      "step": 1173
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.430534351145039e-05,
      "loss": 0.0625,
      "step": 1174
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.427480916030534e-05,
      "loss": 0.0784,
      "step": 1175
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.424427480916031e-05,
      "loss": 0.0682,
      "step": 1176
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.421374045801527e-05,
      "loss": 0.0652,
      "step": 1177
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.418320610687024e-05,
      "loss": 0.0837,
      "step": 1178
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.415267175572519e-05,
      "loss": 0.0715,
      "step": 1179
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.412213740458015e-05,
      "loss": 0.0813,
      "step": 1180
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.409160305343512e-05,
      "loss": 0.0929,
      "step": 1181
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.406106870229008e-05,
      "loss": 0.0588,
      "step": 1182
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.403053435114505e-05,
      "loss": 0.0878,
      "step": 1183
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.400000000000001e-05,
      "loss": 0.0737,
      "step": 1184
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.396946564885496e-05,
      "loss": 0.0798,
      "step": 1185
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.393893129770993e-05,
      "loss": 0.072,
      "step": 1186
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.390839694656489e-05,
      "loss": 0.0955,
      "step": 1187
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.387786259541984e-05,
      "loss": 0.0609,
      "step": 1188
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.384732824427482e-05,
      "loss": 0.0936,
      "step": 1189
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.381679389312978e-05,
      "loss": 0.0723,
      "step": 1190
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.378625954198474e-05,
      "loss": 0.0644,
      "step": 1191
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.37557251908397e-05,
      "loss": 0.0825,
      "step": 1192
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.372519083969466e-05,
      "loss": 0.0938,
      "step": 1193
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.369465648854962e-05,
      "loss": 0.0551,
      "step": 1194
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.366412213740458e-05,
      "loss": 0.0993,
      "step": 1195
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.363358778625954e-05,
      "loss": 0.1125,
      "step": 1196
    },
    {
      "epoch": 7.3,
      "learning_rate": 6.36030534351145e-05,
      "loss": 0.0867,
      "step": 1197
    },
    {
      "epoch": 7.3,
      "learning_rate": 6.357251908396947e-05,
      "loss": 0.0691,
      "step": 1198
    },
    {
      "epoch": 7.31,
      "learning_rate": 6.354198473282444e-05,
      "loss": 0.0686,
      "step": 1199
    },
    {
      "epoch": 7.32,
      "learning_rate": 6.351145038167939e-05,
      "loss": 0.0719,
      "step": 1200
    },
    {
      "epoch": 7.32,
      "learning_rate": 6.348091603053435e-05,
      "loss": 0.0666,
      "step": 1201
    },
    {
      "epoch": 7.33,
      "learning_rate": 6.345038167938932e-05,
      "loss": 0.081,
      "step": 1202
    },
    {
      "epoch": 7.34,
      "learning_rate": 6.341984732824427e-05,
      "loss": 0.0712,
      "step": 1203
    },
    {
      "epoch": 7.34,
      "learning_rate": 6.338931297709923e-05,
      "loss": 0.0771,
      "step": 1204
    },
    {
      "epoch": 7.35,
      "learning_rate": 6.33587786259542e-05,
      "loss": 0.0679,
      "step": 1205
    },
    {
      "epoch": 7.35,
      "learning_rate": 6.332824427480916e-05,
      "loss": 0.0821,
      "step": 1206
    },
    {
      "epoch": 7.36,
      "learning_rate": 6.329770992366413e-05,
      "loss": 0.0815,
      "step": 1207
    },
    {
      "epoch": 7.37,
      "learning_rate": 6.326717557251909e-05,
      "loss": 0.0648,
      "step": 1208
    },
    {
      "epoch": 7.37,
      "learning_rate": 6.323664122137406e-05,
      "loss": 0.0511,
      "step": 1209
    },
    {
      "epoch": 7.38,
      "learning_rate": 6.3206106870229e-05,
      "loss": 0.0652,
      "step": 1210
    },
    {
      "epoch": 7.38,
      "learning_rate": 6.317557251908397e-05,
      "loss": 0.0742,
      "step": 1211
    },
    {
      "epoch": 7.39,
      "learning_rate": 6.314503816793894e-05,
      "loss": 0.0639,
      "step": 1212
    },
    {
      "epoch": 7.4,
      "learning_rate": 6.311450381679389e-05,
      "loss": 0.0749,
      "step": 1213
    },
    {
      "epoch": 7.4,
      "learning_rate": 6.308396946564886e-05,
      "loss": 0.0767,
      "step": 1214
    },
    {
      "epoch": 7.41,
      "learning_rate": 6.305343511450383e-05,
      "loss": 0.0604,
      "step": 1215
    },
    {
      "epoch": 7.41,
      "learning_rate": 6.302290076335878e-05,
      "loss": 0.0737,
      "step": 1216
    },
    {
      "epoch": 7.42,
      "learning_rate": 6.299236641221374e-05,
      "loss": 0.0797,
      "step": 1217
    },
    {
      "epoch": 7.43,
      "learning_rate": 6.296183206106871e-05,
      "loss": 0.0641,
      "step": 1218
    },
    {
      "epoch": 7.43,
      "learning_rate": 6.293129770992366e-05,
      "loss": 0.0859,
      "step": 1219
    },
    {
      "epoch": 7.44,
      "learning_rate": 6.290076335877862e-05,
      "loss": 0.0714,
      "step": 1220
    },
    {
      "epoch": 7.45,
      "learning_rate": 6.287022900763359e-05,
      "loss": 0.0793,
      "step": 1221
    },
    {
      "epoch": 7.45,
      "learning_rate": 6.283969465648855e-05,
      "loss": 0.0666,
      "step": 1222
    },
    {
      "epoch": 7.46,
      "learning_rate": 6.280916030534352e-05,
      "loss": 0.0781,
      "step": 1223
    },
    {
      "epoch": 7.46,
      "learning_rate": 6.277862595419848e-05,
      "loss": 0.0898,
      "step": 1224
    },
    {
      "epoch": 7.47,
      "learning_rate": 6.274809160305343e-05,
      "loss": 0.057,
      "step": 1225
    },
    {
      "epoch": 7.48,
      "learning_rate": 6.27175572519084e-05,
      "loss": 0.078,
      "step": 1226
    },
    {
      "epoch": 7.48,
      "learning_rate": 6.268702290076336e-05,
      "loss": 0.077,
      "step": 1227
    },
    {
      "epoch": 7.49,
      "learning_rate": 6.265648854961831e-05,
      "loss": 0.0808,
      "step": 1228
    },
    {
      "epoch": 7.49,
      "learning_rate": 6.262595419847328e-05,
      "loss": 0.104,
      "step": 1229
    },
    {
      "epoch": 7.5,
      "learning_rate": 6.259541984732826e-05,
      "loss": 0.0718,
      "step": 1230
    },
    {
      "epoch": 7.51,
      "learning_rate": 6.256488549618321e-05,
      "loss": 0.0781,
      "step": 1231
    },
    {
      "epoch": 7.51,
      "learning_rate": 6.253435114503817e-05,
      "loss": 0.1017,
      "step": 1232
    },
    {
      "epoch": 7.52,
      "learning_rate": 6.250381679389314e-05,
      "loss": 0.0785,
      "step": 1233
    },
    {
      "epoch": 7.52,
      "learning_rate": 6.247328244274809e-05,
      "loss": 0.0814,
      "step": 1234
    },
    {
      "epoch": 7.53,
      "learning_rate": 6.244274809160305e-05,
      "loss": 0.0919,
      "step": 1235
    },
    {
      "epoch": 7.54,
      "learning_rate": 6.241221374045802e-05,
      "loss": 0.0738,
      "step": 1236
    },
    {
      "epoch": 7.54,
      "learning_rate": 6.238167938931298e-05,
      "loss": 0.0661,
      "step": 1237
    },
    {
      "epoch": 7.55,
      "learning_rate": 6.235114503816794e-05,
      "loss": 0.0697,
      "step": 1238
    },
    {
      "epoch": 7.55,
      "learning_rate": 6.232061068702291e-05,
      "loss": 0.0956,
      "step": 1239
    },
    {
      "epoch": 7.56,
      "learning_rate": 6.229007633587787e-05,
      "loss": 0.0647,
      "step": 1240
    },
    {
      "epoch": 7.57,
      "learning_rate": 6.225954198473282e-05,
      "loss": 0.0673,
      "step": 1241
    },
    {
      "epoch": 7.57,
      "learning_rate": 6.222900763358779e-05,
      "loss": 0.0658,
      "step": 1242
    },
    {
      "epoch": 7.58,
      "learning_rate": 6.219847328244275e-05,
      "loss": 0.0723,
      "step": 1243
    },
    {
      "epoch": 7.59,
      "learning_rate": 6.21679389312977e-05,
      "loss": 0.0841,
      "step": 1244
    },
    {
      "epoch": 7.59,
      "learning_rate": 6.213740458015267e-05,
      "loss": 0.0687,
      "step": 1245
    },
    {
      "epoch": 7.6,
      "learning_rate": 6.210687022900765e-05,
      "loss": 0.0797,
      "step": 1246
    },
    {
      "epoch": 7.6,
      "learning_rate": 6.20763358778626e-05,
      "loss": 0.0994,
      "step": 1247
    },
    {
      "epoch": 7.61,
      "learning_rate": 6.204580152671756e-05,
      "loss": 0.0604,
      "step": 1248
    },
    {
      "epoch": 7.62,
      "learning_rate": 6.201526717557253e-05,
      "loss": 0.0832,
      "step": 1249
    },
    {
      "epoch": 7.62,
      "learning_rate": 6.198473282442748e-05,
      "loss": 0.0698,
      "step": 1250
    },
    {
      "epoch": 7.63,
      "learning_rate": 6.195419847328244e-05,
      "loss": 0.092,
      "step": 1251
    },
    {
      "epoch": 7.63,
      "learning_rate": 6.192366412213741e-05,
      "loss": 0.0704,
      "step": 1252
    },
    {
      "epoch": 7.64,
      "learning_rate": 6.189312977099236e-05,
      "loss": 0.0891,
      "step": 1253
    },
    {
      "epoch": 7.65,
      "learning_rate": 6.186259541984734e-05,
      "loss": 0.0685,
      "step": 1254
    },
    {
      "epoch": 7.65,
      "learning_rate": 6.18320610687023e-05,
      "loss": 0.0824,
      "step": 1255
    },
    {
      "epoch": 7.66,
      "learning_rate": 6.180152671755725e-05,
      "loss": 0.078,
      "step": 1256
    },
    {
      "epoch": 7.66,
      "learning_rate": 6.177099236641222e-05,
      "loss": 0.081,
      "step": 1257
    },
    {
      "epoch": 7.67,
      "learning_rate": 6.174045801526718e-05,
      "loss": 0.1037,
      "step": 1258
    },
    {
      "epoch": 7.68,
      "learning_rate": 6.170992366412213e-05,
      "loss": 0.0661,
      "step": 1259
    },
    {
      "epoch": 7.68,
      "learning_rate": 6.16793893129771e-05,
      "loss": 0.0921,
      "step": 1260
    },
    {
      "epoch": 7.69,
      "learning_rate": 6.164885496183206e-05,
      "loss": 0.0958,
      "step": 1261
    },
    {
      "epoch": 7.7,
      "learning_rate": 6.161832061068703e-05,
      "loss": 0.0863,
      "step": 1262
    },
    {
      "epoch": 7.7,
      "learning_rate": 6.158778625954199e-05,
      "loss": 0.08,
      "step": 1263
    },
    {
      "epoch": 7.71,
      "learning_rate": 6.155725190839695e-05,
      "loss": 0.087,
      "step": 1264
    },
    {
      "epoch": 7.71,
      "learning_rate": 6.15267175572519e-05,
      "loss": 0.0728,
      "step": 1265
    },
    {
      "epoch": 7.72,
      "learning_rate": 6.149618320610687e-05,
      "loss": 0.0643,
      "step": 1266
    },
    {
      "epoch": 7.73,
      "learning_rate": 6.146564885496183e-05,
      "loss": 0.0663,
      "step": 1267
    },
    {
      "epoch": 7.73,
      "learning_rate": 6.14351145038168e-05,
      "loss": 0.0918,
      "step": 1268
    },
    {
      "epoch": 7.74,
      "learning_rate": 6.140458015267175e-05,
      "loss": 0.0736,
      "step": 1269
    },
    {
      "epoch": 7.74,
      "learning_rate": 6.137404580152673e-05,
      "loss": 0.0696,
      "step": 1270
    },
    {
      "epoch": 7.75,
      "learning_rate": 6.134351145038169e-05,
      "loss": 0.061,
      "step": 1271
    },
    {
      "epoch": 7.76,
      "learning_rate": 6.131297709923664e-05,
      "loss": 0.0631,
      "step": 1272
    },
    {
      "epoch": 7.76,
      "learning_rate": 6.128244274809161e-05,
      "loss": 0.097,
      "step": 1273
    },
    {
      "epoch": 7.77,
      "learning_rate": 6.125190839694657e-05,
      "loss": 0.0636,
      "step": 1274
    },
    {
      "epoch": 7.77,
      "learning_rate": 6.122137404580152e-05,
      "loss": 0.0703,
      "step": 1275
    },
    {
      "epoch": 7.78,
      "learning_rate": 6.119083969465649e-05,
      "loss": 0.0653,
      "step": 1276
    },
    {
      "epoch": 7.79,
      "learning_rate": 6.116030534351145e-05,
      "loss": 0.0786,
      "step": 1277
    },
    {
      "epoch": 7.79,
      "learning_rate": 6.112977099236642e-05,
      "loss": 0.0742,
      "step": 1278
    },
    {
      "epoch": 7.8,
      "learning_rate": 6.109923664122138e-05,
      "loss": 0.0696,
      "step": 1279
    },
    {
      "epoch": 7.8,
      "learning_rate": 6.106870229007635e-05,
      "loss": 0.078,
      "step": 1280
    },
    {
      "epoch": 7.81,
      "learning_rate": 6.10381679389313e-05,
      "loss": 0.0924,
      "step": 1281
    },
    {
      "epoch": 7.82,
      "learning_rate": 6.100763358778626e-05,
      "loss": 0.0732,
      "step": 1282
    },
    {
      "epoch": 7.82,
      "learning_rate": 6.097709923664122e-05,
      "loss": 0.0733,
      "step": 1283
    },
    {
      "epoch": 7.83,
      "learning_rate": 6.094656488549618e-05,
      "loss": 0.0808,
      "step": 1284
    },
    {
      "epoch": 7.84,
      "learning_rate": 6.091603053435114e-05,
      "loss": 0.0866,
      "step": 1285
    },
    {
      "epoch": 7.84,
      "learning_rate": 6.088549618320611e-05,
      "loss": 0.0813,
      "step": 1286
    },
    {
      "epoch": 7.85,
      "learning_rate": 6.085496183206108e-05,
      "loss": 0.076,
      "step": 1287
    },
    {
      "epoch": 7.85,
      "learning_rate": 6.0824427480916034e-05,
      "loss": 0.0877,
      "step": 1288
    },
    {
      "epoch": 7.86,
      "learning_rate": 6.0793893129771e-05,
      "loss": 0.0575,
      "step": 1289
    },
    {
      "epoch": 7.87,
      "learning_rate": 6.076335877862596e-05,
      "loss": 0.0702,
      "step": 1290
    },
    {
      "epoch": 7.87,
      "learning_rate": 6.0732824427480914e-05,
      "loss": 0.0756,
      "step": 1291
    },
    {
      "epoch": 7.88,
      "learning_rate": 6.070229007633588e-05,
      "loss": 0.1205,
      "step": 1292
    },
    {
      "epoch": 7.88,
      "learning_rate": 6.067175572519084e-05,
      "loss": 0.0679,
      "step": 1293
    },
    {
      "epoch": 7.89,
      "learning_rate": 6.064122137404581e-05,
      "loss": 0.0674,
      "step": 1294
    },
    {
      "epoch": 7.9,
      "learning_rate": 6.061068702290077e-05,
      "loss": 0.0809,
      "step": 1295
    },
    {
      "epoch": 7.9,
      "learning_rate": 6.058015267175573e-05,
      "loss": 0.0765,
      "step": 1296
    },
    {
      "epoch": 7.91,
      "learning_rate": 6.054961832061069e-05,
      "loss": 0.0701,
      "step": 1297
    },
    {
      "epoch": 7.91,
      "learning_rate": 6.051908396946565e-05,
      "loss": 0.0689,
      "step": 1298
    },
    {
      "epoch": 7.92,
      "learning_rate": 6.048854961832061e-05,
      "loss": 0.0827,
      "step": 1299
    },
    {
      "epoch": 7.93,
      "learning_rate": 6.0458015267175575e-05,
      "loss": 0.0679,
      "step": 1300
    },
    {
      "epoch": 7.93,
      "learning_rate": 6.042748091603053e-05,
      "loss": 0.0687,
      "step": 1301
    },
    {
      "epoch": 7.94,
      "learning_rate": 6.0396946564885504e-05,
      "loss": 0.0958,
      "step": 1302
    },
    {
      "epoch": 7.95,
      "learning_rate": 6.036641221374046e-05,
      "loss": 0.0771,
      "step": 1303
    },
    {
      "epoch": 7.95,
      "learning_rate": 6.0335877862595426e-05,
      "loss": 0.0835,
      "step": 1304
    },
    {
      "epoch": 7.96,
      "learning_rate": 6.0305343511450384e-05,
      "loss": 0.0731,
      "step": 1305
    },
    {
      "epoch": 7.96,
      "learning_rate": 6.027480916030535e-05,
      "loss": 0.0634,
      "step": 1306
    },
    {
      "epoch": 7.97,
      "learning_rate": 6.0244274809160306e-05,
      "loss": 0.0958,
      "step": 1307
    },
    {
      "epoch": 7.98,
      "learning_rate": 6.0213740458015264e-05,
      "loss": 0.0905,
      "step": 1308
    },
    {
      "epoch": 7.98,
      "learning_rate": 6.018320610687023e-05,
      "loss": 0.0744,
      "step": 1309
    },
    {
      "epoch": 7.99,
      "learning_rate": 6.01526717557252e-05,
      "loss": 0.1056,
      "step": 1310
    },
    {
      "epoch": 7.99,
      "learning_rate": 6.012213740458016e-05,
      "loss": 0.0924,
      "step": 1311
    },
    {
      "epoch": 8.0,
      "learning_rate": 6.009160305343512e-05,
      "loss": 0.0825,
      "step": 1312
    },
    {
      "epoch": 8.01,
      "learning_rate": 6.006106870229008e-05,
      "loss": 0.0757,
      "step": 1313
    },
    {
      "epoch": 8.01,
      "learning_rate": 6.003053435114504e-05,
      "loss": 0.0773,
      "step": 1314
    },
    {
      "epoch": 8.02,
      "learning_rate": 6e-05,
      "loss": 0.0781,
      "step": 1315
    },
    {
      "epoch": 8.02,
      "learning_rate": 5.996946564885496e-05,
      "loss": 0.067,
      "step": 1316
    },
    {
      "epoch": 8.03,
      "learning_rate": 5.9938931297709924e-05,
      "loss": 0.0679,
      "step": 1317
    },
    {
      "epoch": 8.04,
      "learning_rate": 5.9908396946564895e-05,
      "loss": 0.0614,
      "step": 1318
    },
    {
      "epoch": 8.04,
      "learning_rate": 5.987786259541985e-05,
      "loss": 0.0596,
      "step": 1319
    },
    {
      "epoch": 8.05,
      "learning_rate": 5.984732824427482e-05,
      "loss": 0.0638,
      "step": 1320
    },
    {
      "epoch": 8.05,
      "learning_rate": 5.9816793893129775e-05,
      "loss": 0.0744,
      "step": 1321
    },
    {
      "epoch": 8.06,
      "learning_rate": 5.978625954198473e-05,
      "loss": 0.0788,
      "step": 1322
    },
    {
      "epoch": 8.07,
      "learning_rate": 5.97557251908397e-05,
      "loss": 0.0571,
      "step": 1323
    },
    {
      "epoch": 8.07,
      "learning_rate": 5.9725190839694655e-05,
      "loss": 0.0541,
      "step": 1324
    },
    {
      "epoch": 8.08,
      "learning_rate": 5.969465648854962e-05,
      "loss": 0.0915,
      "step": 1325
    },
    {
      "epoch": 8.09,
      "learning_rate": 5.966412213740459e-05,
      "loss": 0.062,
      "step": 1326
    },
    {
      "epoch": 8.09,
      "learning_rate": 5.963358778625955e-05,
      "loss": 0.0727,
      "step": 1327
    },
    {
      "epoch": 8.1,
      "learning_rate": 5.9603053435114506e-05,
      "loss": 0.0626,
      "step": 1328
    },
    {
      "epoch": 8.1,
      "learning_rate": 5.957251908396947e-05,
      "loss": 0.0605,
      "step": 1329
    },
    {
      "epoch": 8.11,
      "learning_rate": 5.954198473282443e-05,
      "loss": 0.0843,
      "step": 1330
    },
    {
      "epoch": 8.12,
      "learning_rate": 5.951145038167939e-05,
      "loss": 0.0767,
      "step": 1331
    },
    {
      "epoch": 8.12,
      "learning_rate": 5.948091603053435e-05,
      "loss": 0.0592,
      "step": 1332
    },
    {
      "epoch": 8.13,
      "learning_rate": 5.945038167938931e-05,
      "loss": 0.072,
      "step": 1333
    },
    {
      "epoch": 8.13,
      "learning_rate": 5.941984732824428e-05,
      "loss": 0.0673,
      "step": 1334
    },
    {
      "epoch": 8.14,
      "learning_rate": 5.9389312977099244e-05,
      "loss": 0.0584,
      "step": 1335
    },
    {
      "epoch": 8.15,
      "learning_rate": 5.93587786259542e-05,
      "loss": 0.0654,
      "step": 1336
    },
    {
      "epoch": 8.15,
      "learning_rate": 5.9328244274809166e-05,
      "loss": 0.0503,
      "step": 1337
    },
    {
      "epoch": 8.16,
      "learning_rate": 5.9297709923664124e-05,
      "loss": 0.0711,
      "step": 1338
    },
    {
      "epoch": 8.16,
      "learning_rate": 5.926717557251908e-05,
      "loss": 0.0607,
      "step": 1339
    },
    {
      "epoch": 8.17,
      "learning_rate": 5.9236641221374046e-05,
      "loss": 0.0681,
      "step": 1340
    },
    {
      "epoch": 8.18,
      "learning_rate": 5.9206106870229004e-05,
      "loss": 0.0731,
      "step": 1341
    },
    {
      "epoch": 8.18,
      "learning_rate": 5.9175572519083975e-05,
      "loss": 0.0784,
      "step": 1342
    },
    {
      "epoch": 8.19,
      "learning_rate": 5.914503816793894e-05,
      "loss": 0.0636,
      "step": 1343
    },
    {
      "epoch": 8.2,
      "learning_rate": 5.91145038167939e-05,
      "loss": 0.0705,
      "step": 1344
    },
    {
      "epoch": 8.2,
      "learning_rate": 5.9083969465648855e-05,
      "loss": 0.0789,
      "step": 1345
    },
    {
      "epoch": 8.21,
      "learning_rate": 5.905343511450382e-05,
      "loss": 0.0713,
      "step": 1346
    },
    {
      "epoch": 8.21,
      "learning_rate": 5.902290076335878e-05,
      "loss": 0.0541,
      "step": 1347
    },
    {
      "epoch": 8.22,
      "learning_rate": 5.899236641221374e-05,
      "loss": 0.0573,
      "step": 1348
    },
    {
      "epoch": 8.23,
      "learning_rate": 5.89618320610687e-05,
      "loss": 0.0611,
      "step": 1349
    },
    {
      "epoch": 8.23,
      "learning_rate": 5.893129770992367e-05,
      "loss": 0.07,
      "step": 1350
    },
    {
      "epoch": 8.24,
      "learning_rate": 5.8900763358778636e-05,
      "loss": 0.0645,
      "step": 1351
    },
    {
      "epoch": 8.24,
      "learning_rate": 5.887022900763359e-05,
      "loss": 0.0638,
      "step": 1352
    },
    {
      "epoch": 8.25,
      "learning_rate": 5.883969465648855e-05,
      "loss": 0.0581,
      "step": 1353
    },
    {
      "epoch": 8.26,
      "learning_rate": 5.8809160305343516e-05,
      "loss": 0.0805,
      "step": 1354
    },
    {
      "epoch": 8.26,
      "learning_rate": 5.877862595419847e-05,
      "loss": 0.0534,
      "step": 1355
    },
    {
      "epoch": 8.27,
      "learning_rate": 5.874809160305344e-05,
      "loss": 0.07,
      "step": 1356
    },
    {
      "epoch": 8.27,
      "learning_rate": 5.8717557251908395e-05,
      "loss": 0.0543,
      "step": 1357
    },
    {
      "epoch": 8.28,
      "learning_rate": 5.868702290076337e-05,
      "loss": 0.0703,
      "step": 1358
    },
    {
      "epoch": 8.29,
      "learning_rate": 5.8656488549618324e-05,
      "loss": 0.0606,
      "step": 1359
    },
    {
      "epoch": 8.29,
      "learning_rate": 5.862595419847329e-05,
      "loss": 0.0705,
      "step": 1360
    },
    {
      "epoch": 8.3,
      "learning_rate": 5.859541984732825e-05,
      "loss": 0.0565,
      "step": 1361
    },
    {
      "epoch": 8.3,
      "learning_rate": 5.856488549618321e-05,
      "loss": 0.0674,
      "step": 1362
    },
    {
      "epoch": 8.31,
      "learning_rate": 5.853435114503817e-05,
      "loss": 0.0657,
      "step": 1363
    },
    {
      "epoch": 8.32,
      "learning_rate": 5.850381679389313e-05,
      "loss": 0.0575,
      "step": 1364
    },
    {
      "epoch": 8.32,
      "learning_rate": 5.847328244274809e-05,
      "loss": 0.0676,
      "step": 1365
    },
    {
      "epoch": 8.33,
      "learning_rate": 5.844274809160306e-05,
      "loss": 0.0571,
      "step": 1366
    },
    {
      "epoch": 8.34,
      "learning_rate": 5.841221374045802e-05,
      "loss": 0.0514,
      "step": 1367
    },
    {
      "epoch": 8.34,
      "learning_rate": 5.8381679389312985e-05,
      "loss": 0.0691,
      "step": 1368
    },
    {
      "epoch": 8.35,
      "learning_rate": 5.835114503816794e-05,
      "loss": 0.0749,
      "step": 1369
    },
    {
      "epoch": 8.35,
      "learning_rate": 5.83206106870229e-05,
      "loss": 0.0666,
      "step": 1370
    },
    {
      "epoch": 8.36,
      "learning_rate": 5.8290076335877865e-05,
      "loss": 0.1009,
      "step": 1371
    },
    {
      "epoch": 8.37,
      "learning_rate": 5.825954198473282e-05,
      "loss": 0.0756,
      "step": 1372
    },
    {
      "epoch": 8.37,
      "learning_rate": 5.822900763358779e-05,
      "loss": 0.1077,
      "step": 1373
    },
    {
      "epoch": 8.38,
      "learning_rate": 5.819847328244276e-05,
      "loss": 0.0537,
      "step": 1374
    },
    {
      "epoch": 8.38,
      "learning_rate": 5.8167938931297716e-05,
      "loss": 0.0751,
      "step": 1375
    },
    {
      "epoch": 8.39,
      "learning_rate": 5.8137404580152674e-05,
      "loss": 0.0704,
      "step": 1376
    },
    {
      "epoch": 8.4,
      "learning_rate": 5.810687022900764e-05,
      "loss": 0.0731,
      "step": 1377
    },
    {
      "epoch": 8.4,
      "learning_rate": 5.8076335877862596e-05,
      "loss": 0.0796,
      "step": 1378
    },
    {
      "epoch": 8.41,
      "learning_rate": 5.804580152671756e-05,
      "loss": 0.0677,
      "step": 1379
    },
    {
      "epoch": 8.41,
      "learning_rate": 5.801526717557252e-05,
      "loss": 0.0738,
      "step": 1380
    },
    {
      "epoch": 8.42,
      "learning_rate": 5.7984732824427476e-05,
      "loss": 0.0506,
      "step": 1381
    },
    {
      "epoch": 8.43,
      "learning_rate": 5.7954198473282454e-05,
      "loss": 0.0574,
      "step": 1382
    },
    {
      "epoch": 8.43,
      "learning_rate": 5.792366412213741e-05,
      "loss": 0.0987,
      "step": 1383
    },
    {
      "epoch": 8.44,
      "learning_rate": 5.789312977099237e-05,
      "loss": 0.0675,
      "step": 1384
    },
    {
      "epoch": 8.45,
      "learning_rate": 5.7862595419847334e-05,
      "loss": 0.0668,
      "step": 1385
    },
    {
      "epoch": 8.45,
      "learning_rate": 5.783206106870229e-05,
      "loss": 0.0673,
      "step": 1386
    },
    {
      "epoch": 8.46,
      "learning_rate": 5.7801526717557256e-05,
      "loss": 0.0596,
      "step": 1387
    },
    {
      "epoch": 8.46,
      "learning_rate": 5.7770992366412214e-05,
      "loss": 0.0649,
      "step": 1388
    },
    {
      "epoch": 8.47,
      "learning_rate": 5.774045801526717e-05,
      "loss": 0.0653,
      "step": 1389
    },
    {
      "epoch": 8.48,
      "learning_rate": 5.770992366412214e-05,
      "loss": 0.0519,
      "step": 1390
    },
    {
      "epoch": 8.48,
      "learning_rate": 5.767938931297711e-05,
      "loss": 0.0599,
      "step": 1391
    },
    {
      "epoch": 8.49,
      "learning_rate": 5.7648854961832065e-05,
      "loss": 0.0711,
      "step": 1392
    },
    {
      "epoch": 8.49,
      "learning_rate": 5.761832061068703e-05,
      "loss": 0.0569,
      "step": 1393
    },
    {
      "epoch": 8.5,
      "learning_rate": 5.758778625954199e-05,
      "loss": 0.0736,
      "step": 1394
    },
    {
      "epoch": 8.51,
      "learning_rate": 5.7557251908396945e-05,
      "loss": 0.0593,
      "step": 1395
    },
    {
      "epoch": 8.51,
      "learning_rate": 5.752671755725191e-05,
      "loss": 0.0699,
      "step": 1396
    },
    {
      "epoch": 8.52,
      "learning_rate": 5.749618320610687e-05,
      "loss": 0.0782,
      "step": 1397
    },
    {
      "epoch": 8.52,
      "learning_rate": 5.746564885496184e-05,
      "loss": 0.0788,
      "step": 1398
    },
    {
      "epoch": 8.53,
      "learning_rate": 5.74351145038168e-05,
      "loss": 0.0626,
      "step": 1399
    },
    {
      "epoch": 8.54,
      "learning_rate": 5.740458015267176e-05,
      "loss": 0.0736,
      "step": 1400
    },
    {
      "epoch": 8.54,
      "learning_rate": 5.737404580152672e-05,
      "loss": 0.0816,
      "step": 1401
    },
    {
      "epoch": 8.55,
      "learning_rate": 5.734351145038168e-05,
      "loss": 0.0641,
      "step": 1402
    },
    {
      "epoch": 8.55,
      "learning_rate": 5.731297709923664e-05,
      "loss": 0.0614,
      "step": 1403
    },
    {
      "epoch": 8.56,
      "learning_rate": 5.7282442748091605e-05,
      "loss": 0.0739,
      "step": 1404
    },
    {
      "epoch": 8.57,
      "learning_rate": 5.725190839694656e-05,
      "loss": 0.0708,
      "step": 1405
    },
    {
      "epoch": 8.57,
      "learning_rate": 5.7221374045801534e-05,
      "loss": 0.0735,
      "step": 1406
    },
    {
      "epoch": 8.58,
      "learning_rate": 5.719083969465649e-05,
      "loss": 0.0747,
      "step": 1407
    },
    {
      "epoch": 8.59,
      "learning_rate": 5.7160305343511456e-05,
      "loss": 0.0845,
      "step": 1408
    },
    {
      "epoch": 8.59,
      "learning_rate": 5.7129770992366414e-05,
      "loss": 0.0533,
      "step": 1409
    },
    {
      "epoch": 8.6,
      "learning_rate": 5.709923664122138e-05,
      "loss": 0.0809,
      "step": 1410
    },
    {
      "epoch": 8.6,
      "learning_rate": 5.7068702290076336e-05,
      "loss": 0.0674,
      "step": 1411
    },
    {
      "epoch": 8.61,
      "learning_rate": 5.7038167938931294e-05,
      "loss": 0.0681,
      "step": 1412
    },
    {
      "epoch": 8.62,
      "learning_rate": 5.700763358778626e-05,
      "loss": 0.067,
      "step": 1413
    },
    {
      "epoch": 8.62,
      "learning_rate": 5.697709923664123e-05,
      "loss": 0.0661,
      "step": 1414
    },
    {
      "epoch": 8.63,
      "learning_rate": 5.694656488549619e-05,
      "loss": 0.0783,
      "step": 1415
    },
    {
      "epoch": 8.63,
      "learning_rate": 5.691603053435115e-05,
      "loss": 0.0764,
      "step": 1416
    },
    {
      "epoch": 8.64,
      "learning_rate": 5.688549618320611e-05,
      "loss": 0.0719,
      "step": 1417
    },
    {
      "epoch": 8.65,
      "learning_rate": 5.6854961832061074e-05,
      "loss": 0.0768,
      "step": 1418
    },
    {
      "epoch": 8.65,
      "learning_rate": 5.682442748091603e-05,
      "loss": 0.0713,
      "step": 1419
    },
    {
      "epoch": 8.66,
      "learning_rate": 5.679389312977099e-05,
      "loss": 0.0693,
      "step": 1420
    },
    {
      "epoch": 8.66,
      "learning_rate": 5.6763358778625954e-05,
      "loss": 0.0635,
      "step": 1421
    },
    {
      "epoch": 8.67,
      "learning_rate": 5.6732824427480926e-05,
      "loss": 0.0646,
      "step": 1422
    },
    {
      "epoch": 8.68,
      "learning_rate": 5.670229007633588e-05,
      "loss": 0.0641,
      "step": 1423
    },
    {
      "epoch": 8.68,
      "learning_rate": 5.667175572519085e-05,
      "loss": 0.062,
      "step": 1424
    },
    {
      "epoch": 8.69,
      "learning_rate": 5.6641221374045806e-05,
      "loss": 0.0766,
      "step": 1425
    },
    {
      "epoch": 8.7,
      "learning_rate": 5.661068702290076e-05,
      "loss": 0.081,
      "step": 1426
    },
    {
      "epoch": 8.7,
      "learning_rate": 5.658015267175573e-05,
      "loss": 0.0686,
      "step": 1427
    },
    {
      "epoch": 8.71,
      "learning_rate": 5.6549618320610686e-05,
      "loss": 0.0544,
      "step": 1428
    },
    {
      "epoch": 8.71,
      "learning_rate": 5.651908396946565e-05,
      "loss": 0.0512,
      "step": 1429
    },
    {
      "epoch": 8.72,
      "learning_rate": 5.648854961832062e-05,
      "loss": 0.0656,
      "step": 1430
    },
    {
      "epoch": 8.73,
      "learning_rate": 5.645801526717558e-05,
      "loss": 0.0765,
      "step": 1431
    },
    {
      "epoch": 8.73,
      "learning_rate": 5.642748091603054e-05,
      "loss": 0.0722,
      "step": 1432
    },
    {
      "epoch": 8.74,
      "learning_rate": 5.63969465648855e-05,
      "loss": 0.0694,
      "step": 1433
    },
    {
      "epoch": 8.74,
      "learning_rate": 5.636641221374046e-05,
      "loss": 0.0695,
      "step": 1434
    },
    {
      "epoch": 8.75,
      "learning_rate": 5.6335877862595423e-05,
      "loss": 0.0693,
      "step": 1435
    },
    {
      "epoch": 8.76,
      "learning_rate": 5.630534351145038e-05,
      "loss": 0.0735,
      "step": 1436
    },
    {
      "epoch": 8.76,
      "learning_rate": 5.627480916030534e-05,
      "loss": 0.0549,
      "step": 1437
    },
    {
      "epoch": 8.77,
      "learning_rate": 5.624427480916031e-05,
      "loss": 0.0547,
      "step": 1438
    },
    {
      "epoch": 8.77,
      "learning_rate": 5.6213740458015275e-05,
      "loss": 0.075,
      "step": 1439
    },
    {
      "epoch": 8.78,
      "learning_rate": 5.618320610687023e-05,
      "loss": 0.0676,
      "step": 1440
    },
    {
      "epoch": 8.79,
      "learning_rate": 5.61526717557252e-05,
      "loss": 0.065,
      "step": 1441
    },
    {
      "epoch": 8.79,
      "learning_rate": 5.6122137404580155e-05,
      "loss": 0.0814,
      "step": 1442
    },
    {
      "epoch": 8.8,
      "learning_rate": 5.609160305343511e-05,
      "loss": 0.0838,
      "step": 1443
    },
    {
      "epoch": 8.8,
      "learning_rate": 5.606106870229008e-05,
      "loss": 0.0788,
      "step": 1444
    },
    {
      "epoch": 8.81,
      "learning_rate": 5.6030534351145035e-05,
      "loss": 0.0716,
      "step": 1445
    },
    {
      "epoch": 8.82,
      "learning_rate": 5.6000000000000006e-05,
      "loss": 0.0861,
      "step": 1446
    },
    {
      "epoch": 8.82,
      "learning_rate": 5.596946564885497e-05,
      "loss": 0.096,
      "step": 1447
    },
    {
      "epoch": 8.83,
      "learning_rate": 5.593893129770993e-05,
      "loss": 0.0765,
      "step": 1448
    },
    {
      "epoch": 8.84,
      "learning_rate": 5.590839694656489e-05,
      "loss": 0.0578,
      "step": 1449
    },
    {
      "epoch": 8.84,
      "learning_rate": 5.587786259541985e-05,
      "loss": 0.0681,
      "step": 1450
    },
    {
      "epoch": 8.85,
      "learning_rate": 5.584732824427481e-05,
      "loss": 0.0747,
      "step": 1451
    },
    {
      "epoch": 8.85,
      "learning_rate": 5.581679389312977e-05,
      "loss": 0.07,
      "step": 1452
    },
    {
      "epoch": 8.86,
      "learning_rate": 5.578625954198473e-05,
      "loss": 0.0783,
      "step": 1453
    },
    {
      "epoch": 8.87,
      "learning_rate": 5.57557251908397e-05,
      "loss": 0.0709,
      "step": 1454
    },
    {
      "epoch": 8.87,
      "learning_rate": 5.5725190839694666e-05,
      "loss": 0.0698,
      "step": 1455
    },
    {
      "epoch": 8.88,
      "learning_rate": 5.5694656488549624e-05,
      "loss": 0.0795,
      "step": 1456
    },
    {
      "epoch": 8.88,
      "learning_rate": 5.566412213740458e-05,
      "loss": 0.0632,
      "step": 1457
    },
    {
      "epoch": 8.89,
      "learning_rate": 5.5633587786259546e-05,
      "loss": 0.0567,
      "step": 1458
    },
    {
      "epoch": 8.9,
      "learning_rate": 5.5603053435114504e-05,
      "loss": 0.0694,
      "step": 1459
    },
    {
      "epoch": 8.9,
      "learning_rate": 5.557251908396947e-05,
      "loss": 0.0633,
      "step": 1460
    },
    {
      "epoch": 8.91,
      "learning_rate": 5.5541984732824426e-05,
      "loss": 0.0766,
      "step": 1461
    },
    {
      "epoch": 8.91,
      "learning_rate": 5.55114503816794e-05,
      "loss": 0.0696,
      "step": 1462
    },
    {
      "epoch": 8.92,
      "learning_rate": 5.5480916030534355e-05,
      "loss": 0.0592,
      "step": 1463
    },
    {
      "epoch": 8.93,
      "learning_rate": 5.545038167938932e-05,
      "loss": 0.0705,
      "step": 1464
    },
    {
      "epoch": 8.93,
      "learning_rate": 5.541984732824428e-05,
      "loss": 0.1017,
      "step": 1465
    },
    {
      "epoch": 8.94,
      "learning_rate": 5.538931297709924e-05,
      "loss": 0.061,
      "step": 1466
    },
    {
      "epoch": 8.95,
      "learning_rate": 5.53587786259542e-05,
      "loss": 0.0701,
      "step": 1467
    },
    {
      "epoch": 8.95,
      "learning_rate": 5.532824427480916e-05,
      "loss": 0.089,
      "step": 1468
    },
    {
      "epoch": 8.96,
      "learning_rate": 5.529770992366412e-05,
      "loss": 0.0704,
      "step": 1469
    },
    {
      "epoch": 8.96,
      "learning_rate": 5.526717557251909e-05,
      "loss": 0.0667,
      "step": 1470
    },
    {
      "epoch": 8.97,
      "learning_rate": 5.523664122137405e-05,
      "loss": 0.0675,
      "step": 1471
    },
    {
      "epoch": 8.98,
      "learning_rate": 5.5206106870229015e-05,
      "loss": 0.0739,
      "step": 1472
    },
    {
      "epoch": 8.98,
      "learning_rate": 5.517557251908397e-05,
      "loss": 0.0921,
      "step": 1473
    },
    {
      "epoch": 8.99,
      "learning_rate": 5.514503816793893e-05,
      "loss": 0.0707,
      "step": 1474
    },
    {
      "epoch": 8.99,
      "learning_rate": 5.5114503816793895e-05,
      "loss": 0.0639,
      "step": 1475
    },
    {
      "epoch": 9.0,
      "learning_rate": 5.508396946564885e-05,
      "loss": 0.0786,
      "step": 1476
    },
    {
      "epoch": 9.01,
      "learning_rate": 5.505343511450382e-05,
      "loss": 0.048,
      "step": 1477
    },
    {
      "epoch": 9.01,
      "learning_rate": 5.502290076335879e-05,
      "loss": 0.0635,
      "step": 1478
    },
    {
      "epoch": 9.02,
      "learning_rate": 5.4992366412213746e-05,
      "loss": 0.0449,
      "step": 1479
    },
    {
      "epoch": 9.02,
      "learning_rate": 5.496183206106871e-05,
      "loss": 0.0548,
      "step": 1480
    },
    {
      "epoch": 9.03,
      "learning_rate": 5.493129770992367e-05,
      "loss": 0.071,
      "step": 1481
    },
    {
      "epoch": 9.04,
      "learning_rate": 5.4900763358778626e-05,
      "loss": 0.0509,
      "step": 1482
    },
    {
      "epoch": 9.04,
      "learning_rate": 5.487022900763359e-05,
      "loss": 0.0566,
      "step": 1483
    },
    {
      "epoch": 9.05,
      "learning_rate": 5.483969465648855e-05,
      "loss": 0.0612,
      "step": 1484
    },
    {
      "epoch": 9.05,
      "learning_rate": 5.480916030534351e-05,
      "loss": 0.067,
      "step": 1485
    },
    {
      "epoch": 9.06,
      "learning_rate": 5.4778625954198484e-05,
      "loss": 0.0623,
      "step": 1486
    },
    {
      "epoch": 9.07,
      "learning_rate": 5.474809160305344e-05,
      "loss": 0.0612,
      "step": 1487
    },
    {
      "epoch": 9.07,
      "learning_rate": 5.47175572519084e-05,
      "loss": 0.0616,
      "step": 1488
    },
    {
      "epoch": 9.08,
      "learning_rate": 5.4687022900763364e-05,
      "loss": 0.0576,
      "step": 1489
    },
    {
      "epoch": 9.09,
      "learning_rate": 5.465648854961832e-05,
      "loss": 0.0786,
      "step": 1490
    },
    {
      "epoch": 9.09,
      "learning_rate": 5.462595419847329e-05,
      "loss": 0.0624,
      "step": 1491
    },
    {
      "epoch": 9.1,
      "learning_rate": 5.4595419847328244e-05,
      "loss": 0.0602,
      "step": 1492
    },
    {
      "epoch": 9.1,
      "learning_rate": 5.45648854961832e-05,
      "loss": 0.0573,
      "step": 1493
    },
    {
      "epoch": 9.11,
      "learning_rate": 5.4534351145038167e-05,
      "loss": 0.0562,
      "step": 1494
    },
    {
      "epoch": 9.12,
      "learning_rate": 5.450381679389314e-05,
      "loss": 0.0471,
      "step": 1495
    },
    {
      "epoch": 9.12,
      "learning_rate": 5.4473282442748096e-05,
      "loss": 0.0617,
      "step": 1496
    },
    {
      "epoch": 9.13,
      "learning_rate": 5.444274809160306e-05,
      "loss": 0.0675,
      "step": 1497
    },
    {
      "epoch": 9.13,
      "learning_rate": 5.441221374045802e-05,
      "loss": 0.0676,
      "step": 1498
    },
    {
      "epoch": 9.14,
      "learning_rate": 5.4381679389312976e-05,
      "loss": 0.0721,
      "step": 1499
    },
    {
      "epoch": 9.15,
      "learning_rate": 5.435114503816794e-05,
      "loss": 0.06,
      "step": 1500
    },
    {
      "epoch": 9.15,
      "learning_rate": 5.43206106870229e-05,
      "loss": 0.0525,
      "step": 1501
    },
    {
      "epoch": 9.16,
      "learning_rate": 5.429007633587786e-05,
      "loss": 0.0569,
      "step": 1502
    },
    {
      "epoch": 9.16,
      "learning_rate": 5.4259541984732834e-05,
      "loss": 0.0598,
      "step": 1503
    },
    {
      "epoch": 9.17,
      "learning_rate": 5.422900763358779e-05,
      "loss": 0.0618,
      "step": 1504
    },
    {
      "epoch": 9.18,
      "learning_rate": 5.419847328244275e-05,
      "loss": 0.063,
      "step": 1505
    },
    {
      "epoch": 9.18,
      "learning_rate": 5.4167938931297714e-05,
      "loss": 0.1083,
      "step": 1506
    },
    {
      "epoch": 9.19,
      "learning_rate": 5.413740458015267e-05,
      "loss": 0.0748,
      "step": 1507
    },
    {
      "epoch": 9.2,
      "learning_rate": 5.4106870229007636e-05,
      "loss": 0.059,
      "step": 1508
    },
    {
      "epoch": 9.2,
      "learning_rate": 5.4076335877862593e-05,
      "loss": 0.0497,
      "step": 1509
    },
    {
      "epoch": 9.21,
      "learning_rate": 5.404580152671755e-05,
      "loss": 0.0592,
      "step": 1510
    },
    {
      "epoch": 9.21,
      "learning_rate": 5.401526717557253e-05,
      "loss": 0.0666,
      "step": 1511
    },
    {
      "epoch": 9.22,
      "learning_rate": 5.398473282442749e-05,
      "loss": 0.0508,
      "step": 1512
    },
    {
      "epoch": 9.23,
      "learning_rate": 5.3954198473282445e-05,
      "loss": 0.0656,
      "step": 1513
    },
    {
      "epoch": 9.23,
      "learning_rate": 5.392366412213741e-05,
      "loss": 0.0569,
      "step": 1514
    },
    {
      "epoch": 9.24,
      "learning_rate": 5.389312977099237e-05,
      "loss": 0.0714,
      "step": 1515
    },
    {
      "epoch": 9.24,
      "learning_rate": 5.386259541984733e-05,
      "loss": 0.0709,
      "step": 1516
    },
    {
      "epoch": 9.25,
      "learning_rate": 5.383206106870229e-05,
      "loss": 0.0561,
      "step": 1517
    },
    {
      "epoch": 9.26,
      "learning_rate": 5.380152671755725e-05,
      "loss": 0.0619,
      "step": 1518
    },
    {
      "epoch": 9.26,
      "learning_rate": 5.377099236641222e-05,
      "loss": 0.0549,
      "step": 1519
    },
    {
      "epoch": 9.27,
      "learning_rate": 5.374045801526718e-05,
      "loss": 0.0819,
      "step": 1520
    },
    {
      "epoch": 9.27,
      "learning_rate": 5.370992366412214e-05,
      "loss": 0.0599,
      "step": 1521
    },
    {
      "epoch": 9.28,
      "learning_rate": 5.3679389312977105e-05,
      "loss": 0.0703,
      "step": 1522
    },
    {
      "epoch": 9.29,
      "learning_rate": 5.364885496183206e-05,
      "loss": 0.1142,
      "step": 1523
    },
    {
      "epoch": 9.29,
      "learning_rate": 5.361832061068702e-05,
      "loss": 0.0505,
      "step": 1524
    },
    {
      "epoch": 9.3,
      "learning_rate": 5.3587786259541985e-05,
      "loss": 0.0471,
      "step": 1525
    },
    {
      "epoch": 9.3,
      "learning_rate": 5.355725190839694e-05,
      "loss": 0.0716,
      "step": 1526
    },
    {
      "epoch": 9.31,
      "learning_rate": 5.3526717557251914e-05,
      "loss": 0.0551,
      "step": 1527
    },
    {
      "epoch": 9.32,
      "learning_rate": 5.349618320610688e-05,
      "loss": 0.0782,
      "step": 1528
    },
    {
      "epoch": 9.32,
      "learning_rate": 5.3465648854961836e-05,
      "loss": 0.0736,
      "step": 1529
    },
    {
      "epoch": 9.33,
      "learning_rate": 5.3435114503816794e-05,
      "loss": 0.0558,
      "step": 1530
    },
    {
      "epoch": 9.34,
      "learning_rate": 5.340458015267176e-05,
      "loss": 0.0602,
      "step": 1531
    },
    {
      "epoch": 9.34,
      "learning_rate": 5.3374045801526716e-05,
      "loss": 0.0801,
      "step": 1532
    },
    {
      "epoch": 9.35,
      "learning_rate": 5.334351145038168e-05,
      "loss": 0.0454,
      "step": 1533
    },
    {
      "epoch": 9.35,
      "learning_rate": 5.331297709923664e-05,
      "loss": 0.0613,
      "step": 1534
    },
    {
      "epoch": 9.36,
      "learning_rate": 5.328244274809161e-05,
      "loss": 0.0556,
      "step": 1535
    },
    {
      "epoch": 9.37,
      "learning_rate": 5.325190839694657e-05,
      "loss": 0.0678,
      "step": 1536
    },
    {
      "epoch": 9.37,
      "learning_rate": 5.322137404580153e-05,
      "loss": 0.0664,
      "step": 1537
    },
    {
      "epoch": 9.38,
      "learning_rate": 5.319083969465649e-05,
      "loss": 0.054,
      "step": 1538
    },
    {
      "epoch": 9.38,
      "learning_rate": 5.3160305343511454e-05,
      "loss": 0.0662,
      "step": 1539
    },
    {
      "epoch": 9.39,
      "learning_rate": 5.312977099236641e-05,
      "loss": 0.0721,
      "step": 1540
    },
    {
      "epoch": 9.4,
      "learning_rate": 5.309923664122137e-05,
      "loss": 0.0563,
      "step": 1541
    },
    {
      "epoch": 9.4,
      "learning_rate": 5.3068702290076334e-05,
      "loss": 0.0681,
      "step": 1542
    },
    {
      "epoch": 9.41,
      "learning_rate": 5.3038167938931305e-05,
      "loss": 0.0637,
      "step": 1543
    },
    {
      "epoch": 9.41,
      "learning_rate": 5.300763358778626e-05,
      "loss": 0.0751,
      "step": 1544
    },
    {
      "epoch": 9.42,
      "learning_rate": 5.297709923664123e-05,
      "loss": 0.0734,
      "step": 1545
    },
    {
      "epoch": 9.43,
      "learning_rate": 5.2946564885496185e-05,
      "loss": 0.0682,
      "step": 1546
    },
    {
      "epoch": 9.43,
      "learning_rate": 5.291603053435115e-05,
      "loss": 0.067,
      "step": 1547
    },
    {
      "epoch": 9.44,
      "learning_rate": 5.288549618320611e-05,
      "loss": 0.069,
      "step": 1548
    },
    {
      "epoch": 9.45,
      "learning_rate": 5.2854961832061065e-05,
      "loss": 0.055,
      "step": 1549
    },
    {
      "epoch": 9.45,
      "learning_rate": 5.282442748091603e-05,
      "loss": 0.0633,
      "step": 1550
    },
    {
      "epoch": 9.46,
      "learning_rate": 5.2793893129771e-05,
      "loss": 0.0581,
      "step": 1551
    },
    {
      "epoch": 9.46,
      "learning_rate": 5.276335877862596e-05,
      "loss": 0.0812,
      "step": 1552
    },
    {
      "epoch": 9.47,
      "learning_rate": 5.273282442748092e-05,
      "loss": 0.0605,
      "step": 1553
    },
    {
      "epoch": 9.48,
      "learning_rate": 5.270229007633588e-05,
      "loss": 0.0532,
      "step": 1554
    },
    {
      "epoch": 9.48,
      "learning_rate": 5.267175572519084e-05,
      "loss": 0.0738,
      "step": 1555
    },
    {
      "epoch": 9.49,
      "learning_rate": 5.26412213740458e-05,
      "loss": 0.0612,
      "step": 1556
    },
    {
      "epoch": 9.49,
      "learning_rate": 5.261068702290076e-05,
      "loss": 0.068,
      "step": 1557
    },
    {
      "epoch": 9.5,
      "learning_rate": 5.2580152671755725e-05,
      "loss": 0.0807,
      "step": 1558
    },
    {
      "epoch": 9.51,
      "learning_rate": 5.25496183206107e-05,
      "loss": 0.0512,
      "step": 1559
    },
    {
      "epoch": 9.51,
      "learning_rate": 5.2519083969465654e-05,
      "loss": 0.0475,
      "step": 1560
    },
    {
      "epoch": 9.52,
      "learning_rate": 5.248854961832061e-05,
      "loss": 0.0842,
      "step": 1561
    },
    {
      "epoch": 9.52,
      "learning_rate": 5.245801526717558e-05,
      "loss": 0.0758,
      "step": 1562
    },
    {
      "epoch": 9.53,
      "learning_rate": 5.2427480916030534e-05,
      "loss": 0.0714,
      "step": 1563
    },
    {
      "epoch": 9.54,
      "learning_rate": 5.23969465648855e-05,
      "loss": 0.055,
      "step": 1564
    },
    {
      "epoch": 9.54,
      "learning_rate": 5.236641221374046e-05,
      "loss": 0.0621,
      "step": 1565
    },
    {
      "epoch": 9.55,
      "learning_rate": 5.2335877862595414e-05,
      "loss": 0.0641,
      "step": 1566
    },
    {
      "epoch": 9.55,
      "learning_rate": 5.2305343511450386e-05,
      "loss": 0.0847,
      "step": 1567
    },
    {
      "epoch": 9.56,
      "learning_rate": 5.227480916030535e-05,
      "loss": 0.0596,
      "step": 1568
    },
    {
      "epoch": 9.57,
      "learning_rate": 5.224427480916031e-05,
      "loss": 0.0749,
      "step": 1569
    },
    {
      "epoch": 9.57,
      "learning_rate": 5.221374045801527e-05,
      "loss": 0.0571,
      "step": 1570
    },
    {
      "epoch": 9.58,
      "learning_rate": 5.218320610687023e-05,
      "loss": 0.0745,
      "step": 1571
    },
    {
      "epoch": 9.59,
      "learning_rate": 5.215267175572519e-05,
      "loss": 0.1129,
      "step": 1572
    },
    {
      "epoch": 9.59,
      "learning_rate": 5.212213740458015e-05,
      "loss": 0.0701,
      "step": 1573
    },
    {
      "epoch": 9.6,
      "learning_rate": 5.209160305343511e-05,
      "loss": 0.0745,
      "step": 1574
    },
    {
      "epoch": 9.6,
      "learning_rate": 5.206106870229008e-05,
      "loss": 0.0968,
      "step": 1575
    },
    {
      "epoch": 9.61,
      "learning_rate": 5.2030534351145046e-05,
      "loss": 0.0655,
      "step": 1576
    },
    {
      "epoch": 9.62,
      "learning_rate": 5.2000000000000004e-05,
      "loss": 0.0779,
      "step": 1577
    },
    {
      "epoch": 9.62,
      "learning_rate": 5.196946564885497e-05,
      "loss": 0.0602,
      "step": 1578
    },
    {
      "epoch": 9.63,
      "learning_rate": 5.1938931297709926e-05,
      "loss": 0.061,
      "step": 1579
    },
    {
      "epoch": 9.63,
      "learning_rate": 5.1908396946564884e-05,
      "loss": 0.0801,
      "step": 1580
    },
    {
      "epoch": 9.64,
      "learning_rate": 5.187786259541985e-05,
      "loss": 0.0727,
      "step": 1581
    },
    {
      "epoch": 9.65,
      "learning_rate": 5.1847328244274806e-05,
      "loss": 0.0587,
      "step": 1582
    },
    {
      "epoch": 9.65,
      "learning_rate": 5.181679389312978e-05,
      "loss": 0.0667,
      "step": 1583
    },
    {
      "epoch": 9.66,
      "learning_rate": 5.178625954198474e-05,
      "loss": 0.0758,
      "step": 1584
    },
    {
      "epoch": 9.66,
      "learning_rate": 5.17557251908397e-05,
      "loss": 0.0569,
      "step": 1585
    },
    {
      "epoch": 9.67,
      "learning_rate": 5.172519083969466e-05,
      "loss": 0.0827,
      "step": 1586
    },
    {
      "epoch": 9.68,
      "learning_rate": 5.169465648854962e-05,
      "loss": 0.0716,
      "step": 1587
    },
    {
      "epoch": 9.68,
      "learning_rate": 5.166412213740458e-05,
      "loss": 0.0736,
      "step": 1588
    },
    {
      "epoch": 9.69,
      "learning_rate": 5.1633587786259544e-05,
      "loss": 0.0746,
      "step": 1589
    },
    {
      "epoch": 9.7,
      "learning_rate": 5.16030534351145e-05,
      "loss": 0.0559,
      "step": 1590
    },
    {
      "epoch": 9.7,
      "learning_rate": 5.157251908396947e-05,
      "loss": 0.064,
      "step": 1591
    },
    {
      "epoch": 9.71,
      "learning_rate": 5.154198473282443e-05,
      "loss": 0.0663,
      "step": 1592
    },
    {
      "epoch": 9.71,
      "learning_rate": 5.1511450381679395e-05,
      "loss": 0.0685,
      "step": 1593
    },
    {
      "epoch": 9.72,
      "learning_rate": 5.148091603053435e-05,
      "loss": 0.0541,
      "step": 1594
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.145038167938932e-05,
      "loss": 0.0662,
      "step": 1595
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.1419847328244275e-05,
      "loss": 0.0745,
      "step": 1596
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.138931297709923e-05,
      "loss": 0.0531,
      "step": 1597
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.13587786259542e-05,
      "loss": 0.0651,
      "step": 1598
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.132824427480917e-05,
      "loss": 0.0616,
      "step": 1599
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.1297709923664126e-05,
      "loss": 0.0663,
      "step": 1600
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.126717557251909e-05,
      "loss": 0.0728,
      "step": 1601
    },
    {
      "epoch": 9.77,
      "learning_rate": 5.123664122137405e-05,
      "loss": 0.0569,
      "step": 1602
    },
    {
      "epoch": 9.77,
      "learning_rate": 5.1206106870229006e-05,
      "loss": 0.0559,
      "step": 1603
    },
    {
      "epoch": 9.78,
      "learning_rate": 5.117557251908397e-05,
      "loss": 0.0619,
      "step": 1604
    },
    {
      "epoch": 9.79,
      "learning_rate": 5.114503816793893e-05,
      "loss": 0.0725,
      "step": 1605
    },
    {
      "epoch": 9.79,
      "learning_rate": 5.111450381679389e-05,
      "loss": 0.0727,
      "step": 1606
    },
    {
      "epoch": 9.8,
      "learning_rate": 5.1083969465648864e-05,
      "loss": 0.092,
      "step": 1607
    },
    {
      "epoch": 9.8,
      "learning_rate": 5.105343511450382e-05,
      "loss": 0.0623,
      "step": 1608
    },
    {
      "epoch": 9.81,
      "learning_rate": 5.1022900763358786e-05,
      "loss": 0.0599,
      "step": 1609
    },
    {
      "epoch": 9.82,
      "learning_rate": 5.0992366412213744e-05,
      "loss": 0.0526,
      "step": 1610
    },
    {
      "epoch": 9.82,
      "learning_rate": 5.09618320610687e-05,
      "loss": 0.0667,
      "step": 1611
    },
    {
      "epoch": 9.83,
      "learning_rate": 5.0931297709923666e-05,
      "loss": 0.0554,
      "step": 1612
    },
    {
      "epoch": 9.84,
      "learning_rate": 5.0900763358778624e-05,
      "loss": 0.0513,
      "step": 1613
    },
    {
      "epoch": 9.84,
      "learning_rate": 5.087022900763359e-05,
      "loss": 0.0682,
      "step": 1614
    },
    {
      "epoch": 9.85,
      "learning_rate": 5.083969465648856e-05,
      "loss": 0.0859,
      "step": 1615
    },
    {
      "epoch": 9.85,
      "learning_rate": 5.080916030534352e-05,
      "loss": 0.0681,
      "step": 1616
    },
    {
      "epoch": 9.86,
      "learning_rate": 5.0778625954198475e-05,
      "loss": 0.0715,
      "step": 1617
    },
    {
      "epoch": 9.87,
      "learning_rate": 5.074809160305344e-05,
      "loss": 0.0708,
      "step": 1618
    },
    {
      "epoch": 9.87,
      "learning_rate": 5.07175572519084e-05,
      "loss": 0.065,
      "step": 1619
    },
    {
      "epoch": 9.88,
      "learning_rate": 5.068702290076336e-05,
      "loss": 0.062,
      "step": 1620
    },
    {
      "epoch": 9.88,
      "learning_rate": 5.065648854961832e-05,
      "loss": 0.0642,
      "step": 1621
    },
    {
      "epoch": 9.89,
      "learning_rate": 5.062595419847328e-05,
      "loss": 0.0679,
      "step": 1622
    },
    {
      "epoch": 9.9,
      "learning_rate": 5.059541984732825e-05,
      "loss": 0.0787,
      "step": 1623
    },
    {
      "epoch": 9.9,
      "learning_rate": 5.056488549618321e-05,
      "loss": 0.0842,
      "step": 1624
    },
    {
      "epoch": 9.91,
      "learning_rate": 5.053435114503817e-05,
      "loss": 0.0585,
      "step": 1625
    },
    {
      "epoch": 9.91,
      "learning_rate": 5.0503816793893136e-05,
      "loss": 0.0842,
      "step": 1626
    },
    {
      "epoch": 9.92,
      "learning_rate": 5.047328244274809e-05,
      "loss": 0.0694,
      "step": 1627
    },
    {
      "epoch": 9.93,
      "learning_rate": 5.044274809160305e-05,
      "loss": 0.0718,
      "step": 1628
    },
    {
      "epoch": 9.93,
      "learning_rate": 5.0412213740458015e-05,
      "loss": 0.0675,
      "step": 1629
    },
    {
      "epoch": 9.94,
      "learning_rate": 5.038167938931297e-05,
      "loss": 0.079,
      "step": 1630
    },
    {
      "epoch": 9.95,
      "learning_rate": 5.0351145038167944e-05,
      "loss": 0.0543,
      "step": 1631
    },
    {
      "epoch": 9.95,
      "learning_rate": 5.032061068702291e-05,
      "loss": 0.0618,
      "step": 1632
    },
    {
      "epoch": 9.96,
      "learning_rate": 5.029007633587787e-05,
      "loss": 0.0779,
      "step": 1633
    },
    {
      "epoch": 9.96,
      "learning_rate": 5.0259541984732824e-05,
      "loss": 0.0592,
      "step": 1634
    },
    {
      "epoch": 9.97,
      "learning_rate": 5.022900763358779e-05,
      "loss": 0.0852,
      "step": 1635
    },
    {
      "epoch": 9.98,
      "learning_rate": 5.019847328244275e-05,
      "loss": 0.0774,
      "step": 1636
    },
    {
      "epoch": 9.98,
      "learning_rate": 5.016793893129771e-05,
      "loss": 0.0702,
      "step": 1637
    },
    {
      "epoch": 9.99,
      "learning_rate": 5.013740458015267e-05,
      "loss": 0.0652,
      "step": 1638
    },
    {
      "epoch": 9.99,
      "learning_rate": 5.010687022900764e-05,
      "loss": 0.0578,
      "step": 1639
    },
    {
      "epoch": 10.0,
      "learning_rate": 5.00763358778626e-05,
      "loss": 0.0965,
      "step": 1640
    },
    {
      "epoch": 10.01,
      "learning_rate": 5.004580152671756e-05,
      "loss": 0.0601,
      "step": 1641
    },
    {
      "epoch": 10.01,
      "learning_rate": 5.001526717557252e-05,
      "loss": 0.0422,
      "step": 1642
    },
    {
      "epoch": 10.02,
      "learning_rate": 4.9984732824427485e-05,
      "loss": 0.0647,
      "step": 1643
    },
    {
      "epoch": 10.02,
      "learning_rate": 4.995419847328244e-05,
      "loss": 0.0433,
      "step": 1644
    },
    {
      "epoch": 10.03,
      "learning_rate": 4.992366412213741e-05,
      "loss": 0.0511,
      "step": 1645
    },
    {
      "epoch": 10.04,
      "learning_rate": 4.989312977099237e-05,
      "loss": 0.0488,
      "step": 1646
    },
    {
      "epoch": 10.04,
      "learning_rate": 4.986259541984733e-05,
      "loss": 0.0678,
      "step": 1647
    },
    {
      "epoch": 10.05,
      "learning_rate": 4.9832061068702294e-05,
      "loss": 0.0581,
      "step": 1648
    },
    {
      "epoch": 10.05,
      "learning_rate": 4.980152671755726e-05,
      "loss": 0.0585,
      "step": 1649
    },
    {
      "epoch": 10.06,
      "learning_rate": 4.9770992366412216e-05,
      "loss": 0.0712,
      "step": 1650
    },
    {
      "epoch": 10.07,
      "learning_rate": 4.974045801526718e-05,
      "loss": 0.0477,
      "step": 1651
    },
    {
      "epoch": 10.07,
      "learning_rate": 4.970992366412214e-05,
      "loss": 0.0585,
      "step": 1652
    },
    {
      "epoch": 10.08,
      "learning_rate": 4.96793893129771e-05,
      "loss": 0.0691,
      "step": 1653
    },
    {
      "epoch": 10.09,
      "learning_rate": 4.964885496183207e-05,
      "loss": 0.0559,
      "step": 1654
    },
    {
      "epoch": 10.09,
      "learning_rate": 4.9618320610687025e-05,
      "loss": 0.0554,
      "step": 1655
    },
    {
      "epoch": 10.1,
      "learning_rate": 4.958778625954198e-05,
      "loss": 0.0665,
      "step": 1656
    },
    {
      "epoch": 10.1,
      "learning_rate": 4.9557251908396954e-05,
      "loss": 0.0542,
      "step": 1657
    },
    {
      "epoch": 10.11,
      "learning_rate": 4.952671755725191e-05,
      "loss": 0.0624,
      "step": 1658
    },
    {
      "epoch": 10.12,
      "learning_rate": 4.949618320610687e-05,
      "loss": 0.0654,
      "step": 1659
    },
    {
      "epoch": 10.12,
      "learning_rate": 4.9465648854961834e-05,
      "loss": 0.0638,
      "step": 1660
    },
    {
      "epoch": 10.13,
      "learning_rate": 4.94351145038168e-05,
      "loss": 0.0634,
      "step": 1661
    },
    {
      "epoch": 10.13,
      "learning_rate": 4.9404580152671756e-05,
      "loss": 0.0639,
      "step": 1662
    },
    {
      "epoch": 10.14,
      "learning_rate": 4.937404580152672e-05,
      "loss": 0.0574,
      "step": 1663
    },
    {
      "epoch": 10.15,
      "learning_rate": 4.934351145038168e-05,
      "loss": 0.0662,
      "step": 1664
    },
    {
      "epoch": 10.15,
      "learning_rate": 4.931297709923664e-05,
      "loss": 0.0598,
      "step": 1665
    },
    {
      "epoch": 10.16,
      "learning_rate": 4.928244274809161e-05,
      "loss": 0.0707,
      "step": 1666
    },
    {
      "epoch": 10.16,
      "learning_rate": 4.9251908396946565e-05,
      "loss": 0.068,
      "step": 1667
    },
    {
      "epoch": 10.17,
      "learning_rate": 4.922137404580153e-05,
      "loss": 0.0691,
      "step": 1668
    },
    {
      "epoch": 10.18,
      "learning_rate": 4.9190839694656494e-05,
      "loss": 0.0519,
      "step": 1669
    },
    {
      "epoch": 10.18,
      "learning_rate": 4.916030534351145e-05,
      "loss": 0.0622,
      "step": 1670
    },
    {
      "epoch": 10.19,
      "learning_rate": 4.9129770992366416e-05,
      "loss": 0.0715,
      "step": 1671
    },
    {
      "epoch": 10.2,
      "learning_rate": 4.9099236641221374e-05,
      "loss": 0.0519,
      "step": 1672
    },
    {
      "epoch": 10.2,
      "learning_rate": 4.906870229007634e-05,
      "loss": 0.0571,
      "step": 1673
    },
    {
      "epoch": 10.21,
      "learning_rate": 4.90381679389313e-05,
      "loss": 0.0496,
      "step": 1674
    },
    {
      "epoch": 10.21,
      "learning_rate": 4.900763358778626e-05,
      "loss": 0.0677,
      "step": 1675
    },
    {
      "epoch": 10.22,
      "learning_rate": 4.897709923664122e-05,
      "loss": 0.0535,
      "step": 1676
    },
    {
      "epoch": 10.23,
      "learning_rate": 4.894656488549619e-05,
      "loss": 0.063,
      "step": 1677
    },
    {
      "epoch": 10.23,
      "learning_rate": 4.891603053435115e-05,
      "loss": 0.0641,
      "step": 1678
    },
    {
      "epoch": 10.24,
      "learning_rate": 4.888549618320611e-05,
      "loss": 0.0631,
      "step": 1679
    },
    {
      "epoch": 10.24,
      "learning_rate": 4.885496183206107e-05,
      "loss": 0.0558,
      "step": 1680
    },
    {
      "epoch": 10.25,
      "learning_rate": 4.8824427480916034e-05,
      "loss": 0.067,
      "step": 1681
    },
    {
      "epoch": 10.26,
      "learning_rate": 4.8793893129771e-05,
      "loss": 0.0586,
      "step": 1682
    },
    {
      "epoch": 10.26,
      "learning_rate": 4.8763358778625956e-05,
      "loss": 0.057,
      "step": 1683
    },
    {
      "epoch": 10.27,
      "learning_rate": 4.8732824427480914e-05,
      "loss": 0.0569,
      "step": 1684
    },
    {
      "epoch": 10.27,
      "learning_rate": 4.8702290076335885e-05,
      "loss": 0.0643,
      "step": 1685
    },
    {
      "epoch": 10.28,
      "learning_rate": 4.867175572519084e-05,
      "loss": 0.0537,
      "step": 1686
    },
    {
      "epoch": 10.29,
      "learning_rate": 4.86412213740458e-05,
      "loss": 0.0527,
      "step": 1687
    },
    {
      "epoch": 10.29,
      "learning_rate": 4.8610687022900765e-05,
      "loss": 0.0764,
      "step": 1688
    },
    {
      "epoch": 10.3,
      "learning_rate": 4.858015267175573e-05,
      "loss": 0.0615,
      "step": 1689
    },
    {
      "epoch": 10.3,
      "learning_rate": 4.854961832061069e-05,
      "loss": 0.062,
      "step": 1690
    },
    {
      "epoch": 10.31,
      "learning_rate": 4.851908396946565e-05,
      "loss": 0.0663,
      "step": 1691
    },
    {
      "epoch": 10.32,
      "learning_rate": 4.848854961832061e-05,
      "loss": 0.0672,
      "step": 1692
    },
    {
      "epoch": 10.32,
      "learning_rate": 4.8458015267175574e-05,
      "loss": 0.0627,
      "step": 1693
    },
    {
      "epoch": 10.33,
      "learning_rate": 4.842748091603054e-05,
      "loss": 0.0736,
      "step": 1694
    },
    {
      "epoch": 10.34,
      "learning_rate": 4.8396946564885497e-05,
      "loss": 0.1104,
      "step": 1695
    },
    {
      "epoch": 10.34,
      "learning_rate": 4.836641221374046e-05,
      "loss": 0.0537,
      "step": 1696
    },
    {
      "epoch": 10.35,
      "learning_rate": 4.8335877862595426e-05,
      "loss": 0.0526,
      "step": 1697
    },
    {
      "epoch": 10.35,
      "learning_rate": 4.830534351145038e-05,
      "loss": 0.0825,
      "step": 1698
    },
    {
      "epoch": 10.36,
      "learning_rate": 4.827480916030535e-05,
      "loss": 0.0543,
      "step": 1699
    },
    {
      "epoch": 10.37,
      "learning_rate": 4.8244274809160306e-05,
      "loss": 0.0445,
      "step": 1700
    },
    {
      "epoch": 10.37,
      "learning_rate": 4.821374045801527e-05,
      "loss": 0.0617,
      "step": 1701
    },
    {
      "epoch": 10.38,
      "learning_rate": 4.8183206106870235e-05,
      "loss": 0.0588,
      "step": 1702
    },
    {
      "epoch": 10.38,
      "learning_rate": 4.815267175572519e-05,
      "loss": 0.0616,
      "step": 1703
    },
    {
      "epoch": 10.39,
      "learning_rate": 4.812213740458015e-05,
      "loss": 0.0701,
      "step": 1704
    },
    {
      "epoch": 10.4,
      "learning_rate": 4.809160305343512e-05,
      "loss": 0.0708,
      "step": 1705
    },
    {
      "epoch": 10.4,
      "learning_rate": 4.806106870229008e-05,
      "loss": 0.0697,
      "step": 1706
    },
    {
      "epoch": 10.41,
      "learning_rate": 4.803053435114504e-05,
      "loss": 0.063,
      "step": 1707
    },
    {
      "epoch": 10.41,
      "learning_rate": 4.8e-05,
      "loss": 0.0638,
      "step": 1708
    },
    {
      "epoch": 10.42,
      "learning_rate": 4.7969465648854966e-05,
      "loss": 0.0702,
      "step": 1709
    },
    {
      "epoch": 10.43,
      "learning_rate": 4.793893129770993e-05,
      "loss": 0.0726,
      "step": 1710
    },
    {
      "epoch": 10.43,
      "learning_rate": 4.790839694656489e-05,
      "loss": 0.0834,
      "step": 1711
    },
    {
      "epoch": 10.44,
      "learning_rate": 4.7877862595419846e-05,
      "loss": 0.0692,
      "step": 1712
    },
    {
      "epoch": 10.45,
      "learning_rate": 4.784732824427482e-05,
      "loss": 0.0643,
      "step": 1713
    },
    {
      "epoch": 10.45,
      "learning_rate": 4.7816793893129775e-05,
      "loss": 0.0601,
      "step": 1714
    },
    {
      "epoch": 10.46,
      "learning_rate": 4.778625954198473e-05,
      "loss": 0.058,
      "step": 1715
    },
    {
      "epoch": 10.46,
      "learning_rate": 4.77557251908397e-05,
      "loss": 0.0806,
      "step": 1716
    },
    {
      "epoch": 10.47,
      "learning_rate": 4.772519083969466e-05,
      "loss": 0.0549,
      "step": 1717
    },
    {
      "epoch": 10.48,
      "learning_rate": 4.769465648854962e-05,
      "loss": 0.0609,
      "step": 1718
    },
    {
      "epoch": 10.48,
      "learning_rate": 4.7664122137404584e-05,
      "loss": 0.0599,
      "step": 1719
    },
    {
      "epoch": 10.49,
      "learning_rate": 4.763358778625954e-05,
      "loss": 0.0633,
      "step": 1720
    },
    {
      "epoch": 10.49,
      "learning_rate": 4.7603053435114506e-05,
      "loss": 0.0591,
      "step": 1721
    },
    {
      "epoch": 10.5,
      "learning_rate": 4.757251908396947e-05,
      "loss": 0.0538,
      "step": 1722
    },
    {
      "epoch": 10.51,
      "learning_rate": 4.754198473282443e-05,
      "loss": 0.0642,
      "step": 1723
    },
    {
      "epoch": 10.51,
      "learning_rate": 4.751145038167939e-05,
      "loss": 0.0693,
      "step": 1724
    },
    {
      "epoch": 10.52,
      "learning_rate": 4.748091603053436e-05,
      "loss": 0.0486,
      "step": 1725
    },
    {
      "epoch": 10.52,
      "learning_rate": 4.7450381679389315e-05,
      "loss": 0.0557,
      "step": 1726
    },
    {
      "epoch": 10.53,
      "learning_rate": 4.741984732824428e-05,
      "loss": 0.069,
      "step": 1727
    },
    {
      "epoch": 10.54,
      "learning_rate": 4.738931297709924e-05,
      "loss": 0.0638,
      "step": 1728
    },
    {
      "epoch": 10.54,
      "learning_rate": 4.73587786259542e-05,
      "loss": 0.0639,
      "step": 1729
    },
    {
      "epoch": 10.55,
      "learning_rate": 4.7328244274809166e-05,
      "loss": 0.0819,
      "step": 1730
    },
    {
      "epoch": 10.55,
      "learning_rate": 4.7297709923664124e-05,
      "loss": 0.0632,
      "step": 1731
    },
    {
      "epoch": 10.56,
      "learning_rate": 4.726717557251908e-05,
      "loss": 0.0663,
      "step": 1732
    },
    {
      "epoch": 10.57,
      "learning_rate": 4.723664122137405e-05,
      "loss": 0.0576,
      "step": 1733
    },
    {
      "epoch": 10.57,
      "learning_rate": 4.720610687022901e-05,
      "loss": 0.0613,
      "step": 1734
    },
    {
      "epoch": 10.58,
      "learning_rate": 4.717557251908397e-05,
      "loss": 0.0588,
      "step": 1735
    },
    {
      "epoch": 10.59,
      "learning_rate": 4.714503816793893e-05,
      "loss": 0.0533,
      "step": 1736
    },
    {
      "epoch": 10.59,
      "learning_rate": 4.71145038167939e-05,
      "loss": 0.0693,
      "step": 1737
    },
    {
      "epoch": 10.6,
      "learning_rate": 4.7083969465648855e-05,
      "loss": 0.0535,
      "step": 1738
    },
    {
      "epoch": 10.6,
      "learning_rate": 4.705343511450382e-05,
      "loss": 0.0551,
      "step": 1739
    },
    {
      "epoch": 10.61,
      "learning_rate": 4.702290076335878e-05,
      "loss": 0.0541,
      "step": 1740
    },
    {
      "epoch": 10.62,
      "learning_rate": 4.699236641221375e-05,
      "loss": 0.0652,
      "step": 1741
    },
    {
      "epoch": 10.62,
      "learning_rate": 4.6961832061068706e-05,
      "loss": 0.0606,
      "step": 1742
    },
    {
      "epoch": 10.63,
      "learning_rate": 4.6931297709923664e-05,
      "loss": 0.0542,
      "step": 1743
    },
    {
      "epoch": 10.63,
      "learning_rate": 4.690076335877863e-05,
      "loss": 0.0642,
      "step": 1744
    },
    {
      "epoch": 10.64,
      "learning_rate": 4.687022900763359e-05,
      "loss": 0.0541,
      "step": 1745
    },
    {
      "epoch": 10.65,
      "learning_rate": 4.683969465648855e-05,
      "loss": 0.061,
      "step": 1746
    },
    {
      "epoch": 10.65,
      "learning_rate": 4.6809160305343515e-05,
      "loss": 0.0555,
      "step": 1747
    },
    {
      "epoch": 10.66,
      "learning_rate": 4.677862595419847e-05,
      "loss": 0.0683,
      "step": 1748
    },
    {
      "epoch": 10.66,
      "learning_rate": 4.674809160305344e-05,
      "loss": 0.0788,
      "step": 1749
    },
    {
      "epoch": 10.67,
      "learning_rate": 4.67175572519084e-05,
      "loss": 0.0703,
      "step": 1750
    },
    {
      "epoch": 10.68,
      "learning_rate": 4.668702290076336e-05,
      "loss": 0.0556,
      "step": 1751
    },
    {
      "epoch": 10.68,
      "learning_rate": 4.6656488549618324e-05,
      "loss": 0.0721,
      "step": 1752
    },
    {
      "epoch": 10.69,
      "learning_rate": 4.662595419847329e-05,
      "loss": 0.0584,
      "step": 1753
    },
    {
      "epoch": 10.7,
      "learning_rate": 4.6595419847328246e-05,
      "loss": 0.0622,
      "step": 1754
    },
    {
      "epoch": 10.7,
      "learning_rate": 4.656488549618321e-05,
      "loss": 0.0706,
      "step": 1755
    },
    {
      "epoch": 10.71,
      "learning_rate": 4.653435114503817e-05,
      "loss": 0.0733,
      "step": 1756
    },
    {
      "epoch": 10.71,
      "learning_rate": 4.650381679389313e-05,
      "loss": 0.0626,
      "step": 1757
    },
    {
      "epoch": 10.72,
      "learning_rate": 4.64732824427481e-05,
      "loss": 0.0683,
      "step": 1758
    },
    {
      "epoch": 10.73,
      "learning_rate": 4.6442748091603055e-05,
      "loss": 0.0601,
      "step": 1759
    },
    {
      "epoch": 10.73,
      "learning_rate": 4.641221374045801e-05,
      "loss": 0.0623,
      "step": 1760
    },
    {
      "epoch": 10.74,
      "learning_rate": 4.6381679389312984e-05,
      "loss": 0.0512,
      "step": 1761
    },
    {
      "epoch": 10.74,
      "learning_rate": 4.635114503816794e-05,
      "loss": 0.0574,
      "step": 1762
    },
    {
      "epoch": 10.75,
      "learning_rate": 4.63206106870229e-05,
      "loss": 0.0574,
      "step": 1763
    },
    {
      "epoch": 10.76,
      "learning_rate": 4.6290076335877864e-05,
      "loss": 0.0656,
      "step": 1764
    },
    {
      "epoch": 10.76,
      "learning_rate": 4.625954198473283e-05,
      "loss": 0.0602,
      "step": 1765
    },
    {
      "epoch": 10.77,
      "learning_rate": 4.6229007633587787e-05,
      "loss": 0.0635,
      "step": 1766
    },
    {
      "epoch": 10.77,
      "learning_rate": 4.619847328244275e-05,
      "loss": 0.05,
      "step": 1767
    },
    {
      "epoch": 10.78,
      "learning_rate": 4.616793893129771e-05,
      "loss": 0.062,
      "step": 1768
    },
    {
      "epoch": 10.79,
      "learning_rate": 4.613740458015267e-05,
      "loss": 0.0675,
      "step": 1769
    },
    {
      "epoch": 10.79,
      "learning_rate": 4.610687022900764e-05,
      "loss": 0.0574,
      "step": 1770
    },
    {
      "epoch": 10.8,
      "learning_rate": 4.6076335877862596e-05,
      "loss": 0.0597,
      "step": 1771
    },
    {
      "epoch": 10.8,
      "learning_rate": 4.604580152671756e-05,
      "loss": 0.0691,
      "step": 1772
    },
    {
      "epoch": 10.81,
      "learning_rate": 4.6015267175572525e-05,
      "loss": 0.0589,
      "step": 1773
    },
    {
      "epoch": 10.82,
      "learning_rate": 4.598473282442748e-05,
      "loss": 0.0669,
      "step": 1774
    },
    {
      "epoch": 10.82,
      "learning_rate": 4.595419847328245e-05,
      "loss": 0.0623,
      "step": 1775
    },
    {
      "epoch": 10.83,
      "learning_rate": 4.5923664122137405e-05,
      "loss": 0.0536,
      "step": 1776
    },
    {
      "epoch": 10.84,
      "learning_rate": 4.589312977099237e-05,
      "loss": 0.0854,
      "step": 1777
    },
    {
      "epoch": 10.84,
      "learning_rate": 4.5862595419847334e-05,
      "loss": 0.0619,
      "step": 1778
    },
    {
      "epoch": 10.85,
      "learning_rate": 4.583206106870229e-05,
      "loss": 0.0624,
      "step": 1779
    },
    {
      "epoch": 10.85,
      "learning_rate": 4.5801526717557256e-05,
      "loss": 0.0669,
      "step": 1780
    },
    {
      "epoch": 10.86,
      "learning_rate": 4.577099236641222e-05,
      "loss": 0.0652,
      "step": 1781
    },
    {
      "epoch": 10.87,
      "learning_rate": 4.574045801526718e-05,
      "loss": 0.0682,
      "step": 1782
    },
    {
      "epoch": 10.87,
      "learning_rate": 4.570992366412214e-05,
      "loss": 0.0501,
      "step": 1783
    },
    {
      "epoch": 10.88,
      "learning_rate": 4.56793893129771e-05,
      "loss": 0.0701,
      "step": 1784
    },
    {
      "epoch": 10.88,
      "learning_rate": 4.5648854961832065e-05,
      "loss": 0.0565,
      "step": 1785
    },
    {
      "epoch": 10.89,
      "learning_rate": 4.561832061068703e-05,
      "loss": 0.0725,
      "step": 1786
    },
    {
      "epoch": 10.9,
      "learning_rate": 4.558778625954199e-05,
      "loss": 0.0729,
      "step": 1787
    },
    {
      "epoch": 10.9,
      "learning_rate": 4.5557251908396945e-05,
      "loss": 0.0859,
      "step": 1788
    },
    {
      "epoch": 10.91,
      "learning_rate": 4.5526717557251916e-05,
      "loss": 0.0768,
      "step": 1789
    },
    {
      "epoch": 10.91,
      "learning_rate": 4.5496183206106874e-05,
      "loss": 0.0479,
      "step": 1790
    },
    {
      "epoch": 10.92,
      "learning_rate": 4.546564885496183e-05,
      "loss": 0.0668,
      "step": 1791
    },
    {
      "epoch": 10.93,
      "learning_rate": 4.5435114503816796e-05,
      "loss": 0.0565,
      "step": 1792
    },
    {
      "epoch": 10.93,
      "learning_rate": 4.5404580152671754e-05,
      "loss": 0.0727,
      "step": 1793
    },
    {
      "epoch": 10.94,
      "learning_rate": 4.537404580152672e-05,
      "loss": 0.0781,
      "step": 1794
    },
    {
      "epoch": 10.95,
      "learning_rate": 4.534351145038168e-05,
      "loss": 0.0636,
      "step": 1795
    },
    {
      "epoch": 10.95,
      "learning_rate": 4.531297709923664e-05,
      "loss": 0.0684,
      "step": 1796
    },
    {
      "epoch": 10.96,
      "learning_rate": 4.5282442748091605e-05,
      "loss": 0.0787,
      "step": 1797
    },
    {
      "epoch": 10.96,
      "learning_rate": 4.525190839694657e-05,
      "loss": 0.0448,
      "step": 1798
    },
    {
      "epoch": 10.97,
      "learning_rate": 4.522137404580153e-05,
      "loss": 0.0748,
      "step": 1799
    },
    {
      "epoch": 10.98,
      "learning_rate": 4.519083969465649e-05,
      "loss": 0.0747,
      "step": 1800
    },
    {
      "epoch": 10.98,
      "learning_rate": 4.516030534351145e-05,
      "loss": 0.0618,
      "step": 1801
    },
    {
      "epoch": 10.99,
      "learning_rate": 4.5129770992366414e-05,
      "loss": 0.0731,
      "step": 1802
    },
    {
      "epoch": 10.99,
      "learning_rate": 4.509923664122138e-05,
      "loss": 0.0581,
      "step": 1803
    },
    {
      "epoch": 11.0,
      "learning_rate": 4.5068702290076336e-05,
      "loss": 0.0682,
      "step": 1804
    },
    {
      "epoch": 11.01,
      "learning_rate": 4.5038167938931294e-05,
      "loss": 0.0698,
      "step": 1805
    },
    {
      "epoch": 11.01,
      "learning_rate": 4.5007633587786265e-05,
      "loss": 0.0445,
      "step": 1806
    },
    {
      "epoch": 11.02,
      "learning_rate": 4.497709923664122e-05,
      "loss": 0.0484,
      "step": 1807
    },
    {
      "epoch": 11.02,
      "learning_rate": 4.494656488549619e-05,
      "loss": 0.0488,
      "step": 1808
    },
    {
      "epoch": 11.03,
      "learning_rate": 4.4916030534351145e-05,
      "loss": 0.0468,
      "step": 1809
    },
    {
      "epoch": 11.04,
      "learning_rate": 4.488549618320611e-05,
      "loss": 0.0473,
      "step": 1810
    },
    {
      "epoch": 11.04,
      "learning_rate": 4.4854961832061074e-05,
      "loss": 0.0563,
      "step": 1811
    },
    {
      "epoch": 11.05,
      "learning_rate": 4.482442748091603e-05,
      "loss": 0.0493,
      "step": 1812
    },
    {
      "epoch": 11.05,
      "learning_rate": 4.479389312977099e-05,
      "loss": 0.0548,
      "step": 1813
    },
    {
      "epoch": 11.06,
      "learning_rate": 4.476335877862596e-05,
      "loss": 0.0472,
      "step": 1814
    },
    {
      "epoch": 11.07,
      "learning_rate": 4.473282442748092e-05,
      "loss": 0.0632,
      "step": 1815
    },
    {
      "epoch": 11.07,
      "learning_rate": 4.4702290076335876e-05,
      "loss": 0.0537,
      "step": 1816
    },
    {
      "epoch": 11.08,
      "learning_rate": 4.467175572519084e-05,
      "loss": 0.0552,
      "step": 1817
    },
    {
      "epoch": 11.09,
      "learning_rate": 4.4641221374045805e-05,
      "loss": 0.054,
      "step": 1818
    },
    {
      "epoch": 11.09,
      "learning_rate": 4.461068702290076e-05,
      "loss": 0.0546,
      "step": 1819
    },
    {
      "epoch": 11.1,
      "learning_rate": 4.458015267175573e-05,
      "loss": 0.0659,
      "step": 1820
    },
    {
      "epoch": 11.1,
      "learning_rate": 4.4549618320610685e-05,
      "loss": 0.0644,
      "step": 1821
    },
    {
      "epoch": 11.11,
      "learning_rate": 4.451908396946565e-05,
      "loss": 0.0589,
      "step": 1822
    },
    {
      "epoch": 11.12,
      "learning_rate": 4.4488549618320614e-05,
      "loss": 0.0619,
      "step": 1823
    },
    {
      "epoch": 11.12,
      "learning_rate": 4.445801526717557e-05,
      "loss": 0.0595,
      "step": 1824
    },
    {
      "epoch": 11.13,
      "learning_rate": 4.4427480916030536e-05,
      "loss": 0.0424,
      "step": 1825
    },
    {
      "epoch": 11.13,
      "learning_rate": 4.43969465648855e-05,
      "loss": 0.0583,
      "step": 1826
    },
    {
      "epoch": 11.14,
      "learning_rate": 4.436641221374046e-05,
      "loss": 0.059,
      "step": 1827
    },
    {
      "epoch": 11.15,
      "learning_rate": 4.433587786259542e-05,
      "loss": 0.0604,
      "step": 1828
    },
    {
      "epoch": 11.15,
      "learning_rate": 4.430534351145038e-05,
      "loss": 0.0424,
      "step": 1829
    },
    {
      "epoch": 11.16,
      "learning_rate": 4.4274809160305345e-05,
      "loss": 0.0536,
      "step": 1830
    },
    {
      "epoch": 11.16,
      "learning_rate": 4.424427480916031e-05,
      "loss": 0.0531,
      "step": 1831
    },
    {
      "epoch": 11.17,
      "learning_rate": 4.421374045801527e-05,
      "loss": 0.0516,
      "step": 1832
    },
    {
      "epoch": 11.18,
      "learning_rate": 4.4183206106870225e-05,
      "loss": 0.069,
      "step": 1833
    },
    {
      "epoch": 11.18,
      "learning_rate": 4.41526717557252e-05,
      "loss": 0.0508,
      "step": 1834
    },
    {
      "epoch": 11.19,
      "learning_rate": 4.4122137404580154e-05,
      "loss": 0.0637,
      "step": 1835
    },
    {
      "epoch": 11.2,
      "learning_rate": 4.409160305343511e-05,
      "loss": 0.0885,
      "step": 1836
    },
    {
      "epoch": 11.2,
      "learning_rate": 4.406106870229008e-05,
      "loss": 0.0463,
      "step": 1837
    },
    {
      "epoch": 11.21,
      "learning_rate": 4.403053435114504e-05,
      "loss": 0.0454,
      "step": 1838
    },
    {
      "epoch": 11.21,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.0547,
      "step": 1839
    },
    {
      "epoch": 11.22,
      "learning_rate": 4.396946564885496e-05,
      "loss": 0.0587,
      "step": 1840
    },
    {
      "epoch": 11.23,
      "learning_rate": 4.393893129770992e-05,
      "loss": 0.0522,
      "step": 1841
    },
    {
      "epoch": 11.23,
      "learning_rate": 4.390839694656489e-05,
      "loss": 0.0655,
      "step": 1842
    },
    {
      "epoch": 11.24,
      "learning_rate": 4.387786259541985e-05,
      "loss": 0.0591,
      "step": 1843
    },
    {
      "epoch": 11.24,
      "learning_rate": 4.384732824427481e-05,
      "loss": 0.0641,
      "step": 1844
    },
    {
      "epoch": 11.25,
      "learning_rate": 4.381679389312977e-05,
      "loss": 0.0655,
      "step": 1845
    },
    {
      "epoch": 11.26,
      "learning_rate": 4.378625954198474e-05,
      "loss": 0.063,
      "step": 1846
    },
    {
      "epoch": 11.26,
      "learning_rate": 4.3755725190839695e-05,
      "loss": 0.0544,
      "step": 1847
    },
    {
      "epoch": 11.27,
      "learning_rate": 4.372519083969466e-05,
      "loss": 0.0439,
      "step": 1848
    },
    {
      "epoch": 11.27,
      "learning_rate": 4.369465648854962e-05,
      "loss": 0.0592,
      "step": 1849
    },
    {
      "epoch": 11.28,
      "learning_rate": 4.366412213740458e-05,
      "loss": 0.0691,
      "step": 1850
    },
    {
      "epoch": 11.29,
      "learning_rate": 4.3633587786259546e-05,
      "loss": 0.0531,
      "step": 1851
    },
    {
      "epoch": 11.29,
      "learning_rate": 4.3603053435114504e-05,
      "loss": 0.0586,
      "step": 1852
    },
    {
      "epoch": 11.3,
      "learning_rate": 4.357251908396947e-05,
      "loss": 0.0586,
      "step": 1853
    },
    {
      "epoch": 11.3,
      "learning_rate": 4.354198473282443e-05,
      "loss": 0.0587,
      "step": 1854
    },
    {
      "epoch": 11.31,
      "learning_rate": 4.351145038167939e-05,
      "loss": 0.0515,
      "step": 1855
    },
    {
      "epoch": 11.32,
      "learning_rate": 4.3480916030534355e-05,
      "loss": 0.0532,
      "step": 1856
    },
    {
      "epoch": 11.32,
      "learning_rate": 4.345038167938931e-05,
      "loss": 0.0701,
      "step": 1857
    },
    {
      "epoch": 11.33,
      "learning_rate": 4.341984732824428e-05,
      "loss": 0.0648,
      "step": 1858
    },
    {
      "epoch": 11.34,
      "learning_rate": 4.338931297709924e-05,
      "loss": 0.0544,
      "step": 1859
    },
    {
      "epoch": 11.34,
      "learning_rate": 4.33587786259542e-05,
      "loss": 0.0532,
      "step": 1860
    },
    {
      "epoch": 11.35,
      "learning_rate": 4.332824427480916e-05,
      "loss": 0.0635,
      "step": 1861
    },
    {
      "epoch": 11.35,
      "learning_rate": 4.329770992366413e-05,
      "loss": 0.0708,
      "step": 1862
    },
    {
      "epoch": 11.36,
      "learning_rate": 4.3267175572519086e-05,
      "loss": 0.0584,
      "step": 1863
    },
    {
      "epoch": 11.37,
      "learning_rate": 4.3236641221374044e-05,
      "loss": 0.068,
      "step": 1864
    },
    {
      "epoch": 11.37,
      "learning_rate": 4.320610687022901e-05,
      "loss": 0.0552,
      "step": 1865
    },
    {
      "epoch": 11.38,
      "learning_rate": 4.317557251908397e-05,
      "loss": 0.0642,
      "step": 1866
    },
    {
      "epoch": 11.38,
      "learning_rate": 4.314503816793893e-05,
      "loss": 0.0562,
      "step": 1867
    },
    {
      "epoch": 11.39,
      "learning_rate": 4.3114503816793895e-05,
      "loss": 0.0913,
      "step": 1868
    },
    {
      "epoch": 11.4,
      "learning_rate": 4.308396946564885e-05,
      "loss": 0.0656,
      "step": 1869
    },
    {
      "epoch": 11.4,
      "learning_rate": 4.3053435114503824e-05,
      "loss": 0.0547,
      "step": 1870
    },
    {
      "epoch": 11.41,
      "learning_rate": 4.302290076335878e-05,
      "loss": 0.067,
      "step": 1871
    },
    {
      "epoch": 11.41,
      "learning_rate": 4.299236641221374e-05,
      "loss": 0.0708,
      "step": 1872
    },
    {
      "epoch": 11.42,
      "learning_rate": 4.2961832061068704e-05,
      "loss": 0.0609,
      "step": 1873
    },
    {
      "epoch": 11.43,
      "learning_rate": 4.293129770992367e-05,
      "loss": 0.0648,
      "step": 1874
    },
    {
      "epoch": 11.43,
      "learning_rate": 4.2900763358778626e-05,
      "loss": 0.0626,
      "step": 1875
    },
    {
      "epoch": 11.44,
      "learning_rate": 4.287022900763359e-05,
      "loss": 0.0843,
      "step": 1876
    },
    {
      "epoch": 11.45,
      "learning_rate": 4.283969465648855e-05,
      "loss": 0.0872,
      "step": 1877
    },
    {
      "epoch": 11.45,
      "learning_rate": 4.280916030534351e-05,
      "loss": 0.0599,
      "step": 1878
    },
    {
      "epoch": 11.46,
      "learning_rate": 4.277862595419848e-05,
      "loss": 0.0605,
      "step": 1879
    },
    {
      "epoch": 11.46,
      "learning_rate": 4.2748091603053435e-05,
      "loss": 0.0623,
      "step": 1880
    },
    {
      "epoch": 11.47,
      "learning_rate": 4.27175572519084e-05,
      "loss": 0.0616,
      "step": 1881
    },
    {
      "epoch": 11.48,
      "learning_rate": 4.2687022900763364e-05,
      "loss": 0.0561,
      "step": 1882
    },
    {
      "epoch": 11.48,
      "learning_rate": 4.265648854961832e-05,
      "loss": 0.063,
      "step": 1883
    },
    {
      "epoch": 11.49,
      "learning_rate": 4.2625954198473286e-05,
      "loss": 0.0687,
      "step": 1884
    },
    {
      "epoch": 11.49,
      "learning_rate": 4.2595419847328244e-05,
      "loss": 0.0684,
      "step": 1885
    },
    {
      "epoch": 11.5,
      "learning_rate": 4.256488549618321e-05,
      "loss": 0.0659,
      "step": 1886
    },
    {
      "epoch": 11.51,
      "learning_rate": 4.253435114503817e-05,
      "loss": 0.0563,
      "step": 1887
    },
    {
      "epoch": 11.51,
      "learning_rate": 4.250381679389313e-05,
      "loss": 0.0502,
      "step": 1888
    },
    {
      "epoch": 11.52,
      "learning_rate": 4.247328244274809e-05,
      "loss": 0.0567,
      "step": 1889
    },
    {
      "epoch": 11.52,
      "learning_rate": 4.244274809160306e-05,
      "loss": 0.0736,
      "step": 1890
    },
    {
      "epoch": 11.53,
      "learning_rate": 4.241221374045802e-05,
      "loss": 0.0735,
      "step": 1891
    },
    {
      "epoch": 11.54,
      "learning_rate": 4.2381679389312975e-05,
      "loss": 0.0564,
      "step": 1892
    },
    {
      "epoch": 11.54,
      "learning_rate": 4.235114503816794e-05,
      "loss": 0.0722,
      "step": 1893
    },
    {
      "epoch": 11.55,
      "learning_rate": 4.2320610687022904e-05,
      "loss": 0.0538,
      "step": 1894
    },
    {
      "epoch": 11.55,
      "learning_rate": 4.229007633587786e-05,
      "loss": 0.0617,
      "step": 1895
    },
    {
      "epoch": 11.56,
      "learning_rate": 4.2259541984732826e-05,
      "loss": 0.0629,
      "step": 1896
    },
    {
      "epoch": 11.57,
      "learning_rate": 4.2229007633587784e-05,
      "loss": 0.0561,
      "step": 1897
    },
    {
      "epoch": 11.57,
      "learning_rate": 4.219847328244275e-05,
      "loss": 0.0522,
      "step": 1898
    },
    {
      "epoch": 11.58,
      "learning_rate": 4.216793893129771e-05,
      "loss": 0.068,
      "step": 1899
    },
    {
      "epoch": 11.59,
      "learning_rate": 4.213740458015267e-05,
      "loss": 0.0793,
      "step": 1900
    },
    {
      "epoch": 11.59,
      "learning_rate": 4.2106870229007635e-05,
      "loss": 0.062,
      "step": 1901
    },
    {
      "epoch": 11.6,
      "learning_rate": 4.20763358778626e-05,
      "loss": 0.0707,
      "step": 1902
    },
    {
      "epoch": 11.6,
      "learning_rate": 4.204580152671756e-05,
      "loss": 0.0554,
      "step": 1903
    },
    {
      "epoch": 11.61,
      "learning_rate": 4.201526717557252e-05,
      "loss": 0.0684,
      "step": 1904
    },
    {
      "epoch": 11.62,
      "learning_rate": 4.198473282442748e-05,
      "loss": 0.0612,
      "step": 1905
    },
    {
      "epoch": 11.62,
      "learning_rate": 4.1954198473282444e-05,
      "loss": 0.0832,
      "step": 1906
    },
    {
      "epoch": 11.63,
      "learning_rate": 4.192366412213741e-05,
      "loss": 0.059,
      "step": 1907
    },
    {
      "epoch": 11.63,
      "learning_rate": 4.189312977099237e-05,
      "loss": 0.0529,
      "step": 1908
    },
    {
      "epoch": 11.64,
      "learning_rate": 4.186259541984733e-05,
      "loss": 0.0683,
      "step": 1909
    },
    {
      "epoch": 11.65,
      "learning_rate": 4.1832061068702296e-05,
      "loss": 0.067,
      "step": 1910
    },
    {
      "epoch": 11.65,
      "learning_rate": 4.1801526717557253e-05,
      "loss": 0.0651,
      "step": 1911
    },
    {
      "epoch": 11.66,
      "learning_rate": 4.177099236641222e-05,
      "loss": 0.0706,
      "step": 1912
    },
    {
      "epoch": 11.66,
      "learning_rate": 4.1740458015267176e-05,
      "loss": 0.0518,
      "step": 1913
    },
    {
      "epoch": 11.67,
      "learning_rate": 4.170992366412214e-05,
      "loss": 0.0739,
      "step": 1914
    },
    {
      "epoch": 11.68,
      "learning_rate": 4.1679389312977105e-05,
      "loss": 0.0629,
      "step": 1915
    },
    {
      "epoch": 11.68,
      "learning_rate": 4.164885496183206e-05,
      "loss": 0.0589,
      "step": 1916
    },
    {
      "epoch": 11.69,
      "learning_rate": 4.161832061068702e-05,
      "loss": 0.0523,
      "step": 1917
    },
    {
      "epoch": 11.7,
      "learning_rate": 4.158778625954199e-05,
      "loss": 0.0605,
      "step": 1918
    },
    {
      "epoch": 11.7,
      "learning_rate": 4.155725190839695e-05,
      "loss": 0.0617,
      "step": 1919
    },
    {
      "epoch": 11.71,
      "learning_rate": 4.152671755725191e-05,
      "loss": 0.0663,
      "step": 1920
    },
    {
      "epoch": 11.71,
      "learning_rate": 4.149618320610687e-05,
      "loss": 0.0763,
      "step": 1921
    },
    {
      "epoch": 11.72,
      "learning_rate": 4.1465648854961836e-05,
      "loss": 0.0566,
      "step": 1922
    },
    {
      "epoch": 11.73,
      "learning_rate": 4.1435114503816794e-05,
      "loss": 0.0713,
      "step": 1923
    },
    {
      "epoch": 11.73,
      "learning_rate": 4.140458015267176e-05,
      "loss": 0.0594,
      "step": 1924
    },
    {
      "epoch": 11.74,
      "learning_rate": 4.1374045801526716e-05,
      "loss": 0.0565,
      "step": 1925
    },
    {
      "epoch": 11.74,
      "learning_rate": 4.134351145038168e-05,
      "loss": 0.0684,
      "step": 1926
    },
    {
      "epoch": 11.75,
      "learning_rate": 4.1312977099236645e-05,
      "loss": 0.067,
      "step": 1927
    },
    {
      "epoch": 11.76,
      "learning_rate": 4.12824427480916e-05,
      "loss": 0.0604,
      "step": 1928
    },
    {
      "epoch": 11.76,
      "learning_rate": 4.125190839694657e-05,
      "loss": 0.0765,
      "step": 1929
    },
    {
      "epoch": 11.77,
      "learning_rate": 4.122137404580153e-05,
      "loss": 0.0708,
      "step": 1930
    },
    {
      "epoch": 11.77,
      "learning_rate": 4.119083969465649e-05,
      "loss": 0.0828,
      "step": 1931
    },
    {
      "epoch": 11.78,
      "learning_rate": 4.1160305343511454e-05,
      "loss": 0.0606,
      "step": 1932
    },
    {
      "epoch": 11.79,
      "learning_rate": 4.112977099236641e-05,
      "loss": 0.0616,
      "step": 1933
    },
    {
      "epoch": 11.79,
      "learning_rate": 4.1099236641221376e-05,
      "loss": 0.0669,
      "step": 1934
    },
    {
      "epoch": 11.8,
      "learning_rate": 4.106870229007634e-05,
      "loss": 0.13,
      "step": 1935
    },
    {
      "epoch": 11.8,
      "learning_rate": 4.10381679389313e-05,
      "loss": 0.0586,
      "step": 1936
    },
    {
      "epoch": 11.81,
      "learning_rate": 4.100763358778626e-05,
      "loss": 0.0604,
      "step": 1937
    },
    {
      "epoch": 11.82,
      "learning_rate": 4.097709923664123e-05,
      "loss": 0.0712,
      "step": 1938
    },
    {
      "epoch": 11.82,
      "learning_rate": 4.0946564885496185e-05,
      "loss": 0.0653,
      "step": 1939
    },
    {
      "epoch": 11.83,
      "learning_rate": 4.091603053435115e-05,
      "loss": 0.0593,
      "step": 1940
    },
    {
      "epoch": 11.84,
      "learning_rate": 4.088549618320611e-05,
      "loss": 0.0653,
      "step": 1941
    },
    {
      "epoch": 11.84,
      "learning_rate": 4.085496183206107e-05,
      "loss": 0.064,
      "step": 1942
    },
    {
      "epoch": 11.85,
      "learning_rate": 4.0824427480916036e-05,
      "loss": 0.0576,
      "step": 1943
    },
    {
      "epoch": 11.85,
      "learning_rate": 4.0793893129770994e-05,
      "loss": 0.0677,
      "step": 1944
    },
    {
      "epoch": 11.86,
      "learning_rate": 4.076335877862595e-05,
      "loss": 0.0546,
      "step": 1945
    },
    {
      "epoch": 11.87,
      "learning_rate": 4.073282442748092e-05,
      "loss": 0.0771,
      "step": 1946
    },
    {
      "epoch": 11.87,
      "learning_rate": 4.070229007633588e-05,
      "loss": 0.0531,
      "step": 1947
    },
    {
      "epoch": 11.88,
      "learning_rate": 4.067175572519084e-05,
      "loss": 0.0611,
      "step": 1948
    },
    {
      "epoch": 11.88,
      "learning_rate": 4.06412213740458e-05,
      "loss": 0.0579,
      "step": 1949
    },
    {
      "epoch": 11.89,
      "learning_rate": 4.061068702290077e-05,
      "loss": 0.0659,
      "step": 1950
    },
    {
      "epoch": 11.9,
      "learning_rate": 4.0580152671755725e-05,
      "loss": 0.0696,
      "step": 1951
    },
    {
      "epoch": 11.9,
      "learning_rate": 4.054961832061069e-05,
      "loss": 0.0586,
      "step": 1952
    },
    {
      "epoch": 11.91,
      "learning_rate": 4.051908396946565e-05,
      "loss": 0.0618,
      "step": 1953
    },
    {
      "epoch": 11.91,
      "learning_rate": 4.048854961832061e-05,
      "loss": 0.0626,
      "step": 1954
    },
    {
      "epoch": 11.92,
      "learning_rate": 4.0458015267175576e-05,
      "loss": 0.0628,
      "step": 1955
    },
    {
      "epoch": 11.93,
      "learning_rate": 4.0427480916030534e-05,
      "loss": 0.0626,
      "step": 1956
    },
    {
      "epoch": 11.93,
      "learning_rate": 4.03969465648855e-05,
      "loss": 0.062,
      "step": 1957
    },
    {
      "epoch": 11.94,
      "learning_rate": 4.036641221374046e-05,
      "loss": 0.0658,
      "step": 1958
    },
    {
      "epoch": 11.95,
      "learning_rate": 4.033587786259542e-05,
      "loss": 0.0539,
      "step": 1959
    },
    {
      "epoch": 11.95,
      "learning_rate": 4.0305343511450385e-05,
      "loss": 0.0528,
      "step": 1960
    },
    {
      "epoch": 11.96,
      "learning_rate": 4.027480916030534e-05,
      "loss": 0.0809,
      "step": 1961
    },
    {
      "epoch": 11.96,
      "learning_rate": 4.024427480916031e-05,
      "loss": 0.0628,
      "step": 1962
    },
    {
      "epoch": 11.97,
      "learning_rate": 4.021374045801527e-05,
      "loss": 0.0651,
      "step": 1963
    },
    {
      "epoch": 11.98,
      "learning_rate": 4.018320610687023e-05,
      "loss": 0.077,
      "step": 1964
    },
    {
      "epoch": 11.98,
      "learning_rate": 4.015267175572519e-05,
      "loss": 0.0562,
      "step": 1965
    },
    {
      "epoch": 11.99,
      "learning_rate": 4.012213740458016e-05,
      "loss": 0.0681,
      "step": 1966
    },
    {
      "epoch": 11.99,
      "learning_rate": 4.0091603053435117e-05,
      "loss": 0.0698,
      "step": 1967
    },
    {
      "epoch": 12.0,
      "learning_rate": 4.006106870229008e-05,
      "loss": 0.0605,
      "step": 1968
    },
    {
      "epoch": 12.01,
      "learning_rate": 4.003053435114504e-05,
      "loss": 0.0545,
      "step": 1969
    },
    {
      "epoch": 12.01,
      "learning_rate": 4e-05,
      "loss": 0.0534,
      "step": 1970
    },
    {
      "epoch": 12.02,
      "learning_rate": 3.996946564885497e-05,
      "loss": 0.0697,
      "step": 1971
    },
    {
      "epoch": 12.02,
      "learning_rate": 3.9938931297709926e-05,
      "loss": 0.0561,
      "step": 1972
    },
    {
      "epoch": 12.03,
      "learning_rate": 3.990839694656488e-05,
      "loss": 0.0498,
      "step": 1973
    },
    {
      "epoch": 12.04,
      "learning_rate": 3.9877862595419855e-05,
      "loss": 0.0546,
      "step": 1974
    },
    {
      "epoch": 12.04,
      "learning_rate": 3.984732824427481e-05,
      "loss": 0.056,
      "step": 1975
    },
    {
      "epoch": 12.05,
      "learning_rate": 3.981679389312977e-05,
      "loss": 0.0714,
      "step": 1976
    },
    {
      "epoch": 12.05,
      "learning_rate": 3.9786259541984734e-05,
      "loss": 0.0622,
      "step": 1977
    },
    {
      "epoch": 12.06,
      "learning_rate": 3.97557251908397e-05,
      "loss": 0.0556,
      "step": 1978
    },
    {
      "epoch": 12.07,
      "learning_rate": 3.972519083969466e-05,
      "loss": 0.0568,
      "step": 1979
    },
    {
      "epoch": 12.07,
      "learning_rate": 3.969465648854962e-05,
      "loss": 0.0646,
      "step": 1980
    },
    {
      "epoch": 12.08,
      "learning_rate": 3.966412213740458e-05,
      "loss": 0.0514,
      "step": 1981
    },
    {
      "epoch": 12.09,
      "learning_rate": 3.9633587786259543e-05,
      "loss": 0.059,
      "step": 1982
    },
    {
      "epoch": 12.09,
      "learning_rate": 3.960305343511451e-05,
      "loss": 0.0644,
      "step": 1983
    },
    {
      "epoch": 12.1,
      "learning_rate": 3.9572519083969466e-05,
      "loss": 0.0388,
      "step": 1984
    },
    {
      "epoch": 12.1,
      "learning_rate": 3.954198473282443e-05,
      "loss": 0.0544,
      "step": 1985
    },
    {
      "epoch": 12.11,
      "learning_rate": 3.9511450381679395e-05,
      "loss": 0.05,
      "step": 1986
    },
    {
      "epoch": 12.12,
      "learning_rate": 3.948091603053435e-05,
      "loss": 0.0503,
      "step": 1987
    },
    {
      "epoch": 12.12,
      "learning_rate": 3.945038167938932e-05,
      "loss": 0.0455,
      "step": 1988
    },
    {
      "epoch": 12.13,
      "learning_rate": 3.9419847328244275e-05,
      "loss": 0.0654,
      "step": 1989
    },
    {
      "epoch": 12.13,
      "learning_rate": 3.938931297709924e-05,
      "loss": 0.0576,
      "step": 1990
    },
    {
      "epoch": 12.14,
      "learning_rate": 3.9358778625954204e-05,
      "loss": 0.0709,
      "step": 1991
    },
    {
      "epoch": 12.15,
      "learning_rate": 3.932824427480916e-05,
      "loss": 0.0658,
      "step": 1992
    },
    {
      "epoch": 12.15,
      "learning_rate": 3.929770992366412e-05,
      "loss": 0.0586,
      "step": 1993
    },
    {
      "epoch": 12.16,
      "learning_rate": 3.926717557251909e-05,
      "loss": 0.0523,
      "step": 1994
    },
    {
      "epoch": 12.16,
      "learning_rate": 3.923664122137405e-05,
      "loss": 0.063,
      "step": 1995
    },
    {
      "epoch": 12.17,
      "learning_rate": 3.9206106870229006e-05,
      "loss": 0.0452,
      "step": 1996
    },
    {
      "epoch": 12.18,
      "learning_rate": 3.917557251908397e-05,
      "loss": 0.0657,
      "step": 1997
    },
    {
      "epoch": 12.18,
      "learning_rate": 3.9145038167938935e-05,
      "loss": 0.0615,
      "step": 1998
    },
    {
      "epoch": 12.19,
      "learning_rate": 3.91145038167939e-05,
      "loss": 0.0563,
      "step": 1999
    },
    {
      "epoch": 12.2,
      "learning_rate": 3.908396946564886e-05,
      "loss": 0.0537,
      "step": 2000
    },
    {
      "epoch": 12.2,
      "learning_rate": 3.9053435114503815e-05,
      "loss": 0.0521,
      "step": 2001
    },
    {
      "epoch": 12.21,
      "learning_rate": 3.9022900763358786e-05,
      "loss": 0.063,
      "step": 2002
    },
    {
      "epoch": 12.21,
      "learning_rate": 3.8992366412213744e-05,
      "loss": 0.0602,
      "step": 2003
    },
    {
      "epoch": 12.22,
      "learning_rate": 3.89618320610687e-05,
      "loss": 0.0508,
      "step": 2004
    },
    {
      "epoch": 12.23,
      "learning_rate": 3.8931297709923666e-05,
      "loss": 0.0522,
      "step": 2005
    },
    {
      "epoch": 12.23,
      "learning_rate": 3.890076335877863e-05,
      "loss": 0.0574,
      "step": 2006
    },
    {
      "epoch": 12.24,
      "learning_rate": 3.887022900763359e-05,
      "loss": 0.0687,
      "step": 2007
    },
    {
      "epoch": 12.24,
      "learning_rate": 3.883969465648855e-05,
      "loss": 0.0538,
      "step": 2008
    },
    {
      "epoch": 12.25,
      "learning_rate": 3.880916030534351e-05,
      "loss": 0.0614,
      "step": 2009
    },
    {
      "epoch": 12.26,
      "learning_rate": 3.8778625954198475e-05,
      "loss": 0.0663,
      "step": 2010
    },
    {
      "epoch": 12.26,
      "learning_rate": 3.874809160305344e-05,
      "loss": 0.065,
      "step": 2011
    },
    {
      "epoch": 12.27,
      "learning_rate": 3.87175572519084e-05,
      "loss": 0.0641,
      "step": 2012
    },
    {
      "epoch": 12.27,
      "learning_rate": 3.868702290076336e-05,
      "loss": 0.0611,
      "step": 2013
    },
    {
      "epoch": 12.28,
      "learning_rate": 3.8656488549618326e-05,
      "loss": 0.0861,
      "step": 2014
    },
    {
      "epoch": 12.29,
      "learning_rate": 3.8625954198473284e-05,
      "loss": 0.0634,
      "step": 2015
    },
    {
      "epoch": 12.29,
      "learning_rate": 3.859541984732825e-05,
      "loss": 0.0552,
      "step": 2016
    },
    {
      "epoch": 12.3,
      "learning_rate": 3.8564885496183206e-05,
      "loss": 0.0584,
      "step": 2017
    },
    {
      "epoch": 12.3,
      "learning_rate": 3.853435114503817e-05,
      "loss": 0.045,
      "step": 2018
    },
    {
      "epoch": 12.31,
      "learning_rate": 3.8503816793893135e-05,
      "loss": 0.0552,
      "step": 2019
    },
    {
      "epoch": 12.32,
      "learning_rate": 3.847328244274809e-05,
      "loss": 0.044,
      "step": 2020
    },
    {
      "epoch": 12.32,
      "learning_rate": 3.844274809160305e-05,
      "loss": 0.0522,
      "step": 2021
    },
    {
      "epoch": 12.33,
      "learning_rate": 3.841221374045802e-05,
      "loss": 0.0515,
      "step": 2022
    },
    {
      "epoch": 12.34,
      "learning_rate": 3.838167938931298e-05,
      "loss": 0.0598,
      "step": 2023
    },
    {
      "epoch": 12.34,
      "learning_rate": 3.835114503816794e-05,
      "loss": 0.0695,
      "step": 2024
    },
    {
      "epoch": 12.35,
      "learning_rate": 3.83206106870229e-05,
      "loss": 0.0572,
      "step": 2025
    },
    {
      "epoch": 12.35,
      "learning_rate": 3.8290076335877866e-05,
      "loss": 0.0696,
      "step": 2026
    },
    {
      "epoch": 12.36,
      "learning_rate": 3.8259541984732824e-05,
      "loss": 0.0539,
      "step": 2027
    },
    {
      "epoch": 12.37,
      "learning_rate": 3.822900763358779e-05,
      "loss": 0.0584,
      "step": 2028
    },
    {
      "epoch": 12.37,
      "learning_rate": 3.8198473282442746e-05,
      "loss": 0.0726,
      "step": 2029
    },
    {
      "epoch": 12.38,
      "learning_rate": 3.816793893129771e-05,
      "loss": 0.0606,
      "step": 2030
    },
    {
      "epoch": 12.38,
      "learning_rate": 3.8137404580152675e-05,
      "loss": 0.0523,
      "step": 2031
    },
    {
      "epoch": 12.39,
      "learning_rate": 3.810687022900763e-05,
      "loss": 0.0594,
      "step": 2032
    },
    {
      "epoch": 12.4,
      "learning_rate": 3.80763358778626e-05,
      "loss": 0.0499,
      "step": 2033
    },
    {
      "epoch": 12.4,
      "learning_rate": 3.804580152671756e-05,
      "loss": 0.0615,
      "step": 2034
    },
    {
      "epoch": 12.41,
      "learning_rate": 3.801526717557252e-05,
      "loss": 0.0487,
      "step": 2035
    },
    {
      "epoch": 12.41,
      "learning_rate": 3.7984732824427484e-05,
      "loss": 0.0597,
      "step": 2036
    },
    {
      "epoch": 12.42,
      "learning_rate": 3.795419847328244e-05,
      "loss": 0.0656,
      "step": 2037
    },
    {
      "epoch": 12.43,
      "learning_rate": 3.7923664122137407e-05,
      "loss": 0.0574,
      "step": 2038
    },
    {
      "epoch": 12.43,
      "learning_rate": 3.789312977099237e-05,
      "loss": 0.0706,
      "step": 2039
    },
    {
      "epoch": 12.44,
      "learning_rate": 3.786259541984733e-05,
      "loss": 0.0707,
      "step": 2040
    },
    {
      "epoch": 12.45,
      "learning_rate": 3.783206106870229e-05,
      "loss": 0.0524,
      "step": 2041
    },
    {
      "epoch": 12.45,
      "learning_rate": 3.780152671755726e-05,
      "loss": 0.0363,
      "step": 2042
    },
    {
      "epoch": 12.46,
      "learning_rate": 3.7770992366412216e-05,
      "loss": 0.0675,
      "step": 2043
    },
    {
      "epoch": 12.46,
      "learning_rate": 3.774045801526718e-05,
      "loss": 0.0555,
      "step": 2044
    },
    {
      "epoch": 12.47,
      "learning_rate": 3.770992366412214e-05,
      "loss": 0.058,
      "step": 2045
    },
    {
      "epoch": 12.48,
      "learning_rate": 3.76793893129771e-05,
      "loss": 0.0533,
      "step": 2046
    },
    {
      "epoch": 12.48,
      "learning_rate": 3.764885496183207e-05,
      "loss": 0.0549,
      "step": 2047
    },
    {
      "epoch": 12.49,
      "learning_rate": 3.7618320610687025e-05,
      "loss": 0.0643,
      "step": 2048
    },
    {
      "epoch": 12.49,
      "learning_rate": 3.758778625954198e-05,
      "loss": 0.0579,
      "step": 2049
    },
    {
      "epoch": 12.5,
      "learning_rate": 3.7557251908396954e-05,
      "loss": 0.0673,
      "step": 2050
    },
    {
      "epoch": 12.51,
      "learning_rate": 3.752671755725191e-05,
      "loss": 0.0598,
      "step": 2051
    },
    {
      "epoch": 12.51,
      "learning_rate": 3.749618320610687e-05,
      "loss": 0.0684,
      "step": 2052
    },
    {
      "epoch": 12.52,
      "learning_rate": 3.7465648854961833e-05,
      "loss": 0.0684,
      "step": 2053
    },
    {
      "epoch": 12.52,
      "learning_rate": 3.74351145038168e-05,
      "loss": 0.052,
      "step": 2054
    },
    {
      "epoch": 12.53,
      "learning_rate": 3.7404580152671756e-05,
      "loss": 0.0685,
      "step": 2055
    },
    {
      "epoch": 12.54,
      "learning_rate": 3.737404580152672e-05,
      "loss": 0.0646,
      "step": 2056
    },
    {
      "epoch": 12.54,
      "learning_rate": 3.734351145038168e-05,
      "loss": 0.0597,
      "step": 2057
    },
    {
      "epoch": 12.55,
      "learning_rate": 3.731297709923664e-05,
      "loss": 0.0674,
      "step": 2058
    },
    {
      "epoch": 12.55,
      "learning_rate": 3.728244274809161e-05,
      "loss": 0.0522,
      "step": 2059
    },
    {
      "epoch": 12.56,
      "learning_rate": 3.7251908396946565e-05,
      "loss": 0.0708,
      "step": 2060
    },
    {
      "epoch": 12.57,
      "learning_rate": 3.722137404580153e-05,
      "loss": 0.0601,
      "step": 2061
    },
    {
      "epoch": 12.57,
      "learning_rate": 3.7190839694656494e-05,
      "loss": 0.061,
      "step": 2062
    },
    {
      "epoch": 12.58,
      "learning_rate": 3.716030534351145e-05,
      "loss": 0.0481,
      "step": 2063
    },
    {
      "epoch": 12.59,
      "learning_rate": 3.7129770992366416e-05,
      "loss": 0.0536,
      "step": 2064
    },
    {
      "epoch": 12.59,
      "learning_rate": 3.7099236641221374e-05,
      "loss": 0.055,
      "step": 2065
    },
    {
      "epoch": 12.6,
      "learning_rate": 3.706870229007634e-05,
      "loss": 0.0579,
      "step": 2066
    },
    {
      "epoch": 12.6,
      "learning_rate": 3.70381679389313e-05,
      "loss": 0.0614,
      "step": 2067
    },
    {
      "epoch": 12.61,
      "learning_rate": 3.700763358778626e-05,
      "loss": 0.0607,
      "step": 2068
    },
    {
      "epoch": 12.62,
      "learning_rate": 3.6977099236641225e-05,
      "loss": 0.0584,
      "step": 2069
    },
    {
      "epoch": 12.62,
      "learning_rate": 3.694656488549619e-05,
      "loss": 0.0544,
      "step": 2070
    },
    {
      "epoch": 12.63,
      "learning_rate": 3.691603053435115e-05,
      "loss": 0.0644,
      "step": 2071
    },
    {
      "epoch": 12.63,
      "learning_rate": 3.688549618320611e-05,
      "loss": 0.0607,
      "step": 2072
    },
    {
      "epoch": 12.64,
      "learning_rate": 3.685496183206107e-05,
      "loss": 0.0609,
      "step": 2073
    },
    {
      "epoch": 12.65,
      "learning_rate": 3.6824427480916034e-05,
      "loss": 0.0584,
      "step": 2074
    },
    {
      "epoch": 12.65,
      "learning_rate": 3.6793893129771e-05,
      "loss": 0.0561,
      "step": 2075
    },
    {
      "epoch": 12.66,
      "learning_rate": 3.6763358778625956e-05,
      "loss": 0.0618,
      "step": 2076
    },
    {
      "epoch": 12.66,
      "learning_rate": 3.6732824427480914e-05,
      "loss": 0.0605,
      "step": 2077
    },
    {
      "epoch": 12.67,
      "learning_rate": 3.6702290076335885e-05,
      "loss": 0.0738,
      "step": 2078
    },
    {
      "epoch": 12.68,
      "learning_rate": 3.667175572519084e-05,
      "loss": 0.0568,
      "step": 2079
    },
    {
      "epoch": 12.68,
      "learning_rate": 3.66412213740458e-05,
      "loss": 0.0571,
      "step": 2080
    },
    {
      "epoch": 12.69,
      "learning_rate": 3.6610687022900765e-05,
      "loss": 0.0529,
      "step": 2081
    },
    {
      "epoch": 12.7,
      "learning_rate": 3.658015267175573e-05,
      "loss": 0.0602,
      "step": 2082
    },
    {
      "epoch": 12.7,
      "learning_rate": 3.654961832061069e-05,
      "loss": 0.0654,
      "step": 2083
    },
    {
      "epoch": 12.71,
      "learning_rate": 3.651908396946565e-05,
      "loss": 0.0638,
      "step": 2084
    },
    {
      "epoch": 12.71,
      "learning_rate": 3.648854961832061e-05,
      "loss": 0.066,
      "step": 2085
    },
    {
      "epoch": 12.72,
      "learning_rate": 3.6458015267175574e-05,
      "loss": 0.0653,
      "step": 2086
    },
    {
      "epoch": 12.73,
      "learning_rate": 3.642748091603054e-05,
      "loss": 0.0564,
      "step": 2087
    },
    {
      "epoch": 12.73,
      "learning_rate": 3.6396946564885496e-05,
      "loss": 0.0688,
      "step": 2088
    },
    {
      "epoch": 12.74,
      "learning_rate": 3.636641221374046e-05,
      "loss": 0.0491,
      "step": 2089
    },
    {
      "epoch": 12.74,
      "learning_rate": 3.633587786259542e-05,
      "loss": 0.0627,
      "step": 2090
    },
    {
      "epoch": 12.75,
      "learning_rate": 3.630534351145038e-05,
      "loss": 0.0648,
      "step": 2091
    },
    {
      "epoch": 12.76,
      "learning_rate": 3.627480916030535e-05,
      "loss": 0.0687,
      "step": 2092
    },
    {
      "epoch": 12.76,
      "learning_rate": 3.6244274809160305e-05,
      "loss": 0.068,
      "step": 2093
    },
    {
      "epoch": 12.77,
      "learning_rate": 3.621374045801526e-05,
      "loss": 0.0668,
      "step": 2094
    },
    {
      "epoch": 12.77,
      "learning_rate": 3.6183206106870234e-05,
      "loss": 0.0718,
      "step": 2095
    },
    {
      "epoch": 12.78,
      "learning_rate": 3.615267175572519e-05,
      "loss": 0.0608,
      "step": 2096
    },
    {
      "epoch": 12.79,
      "learning_rate": 3.612213740458015e-05,
      "loss": 0.0618,
      "step": 2097
    },
    {
      "epoch": 12.79,
      "learning_rate": 3.6091603053435114e-05,
      "loss": 0.0665,
      "step": 2098
    },
    {
      "epoch": 12.8,
      "learning_rate": 3.606106870229008e-05,
      "loss": 0.0724,
      "step": 2099
    },
    {
      "epoch": 12.8,
      "learning_rate": 3.603053435114504e-05,
      "loss": 0.0989,
      "step": 2100
    },
    {
      "epoch": 12.81,
      "learning_rate": 3.6e-05,
      "loss": 0.0752,
      "step": 2101
    },
    {
      "epoch": 12.82,
      "learning_rate": 3.596946564885496e-05,
      "loss": 0.052,
      "step": 2102
    },
    {
      "epoch": 12.82,
      "learning_rate": 3.593893129770993e-05,
      "loss": 0.0538,
      "step": 2103
    },
    {
      "epoch": 12.83,
      "learning_rate": 3.590839694656489e-05,
      "loss": 0.065,
      "step": 2104
    },
    {
      "epoch": 12.84,
      "learning_rate": 3.5877862595419845e-05,
      "loss": 0.0553,
      "step": 2105
    },
    {
      "epoch": 12.84,
      "learning_rate": 3.584732824427481e-05,
      "loss": 0.0679,
      "step": 2106
    },
    {
      "epoch": 12.85,
      "learning_rate": 3.5816793893129774e-05,
      "loss": 0.0675,
      "step": 2107
    },
    {
      "epoch": 12.85,
      "learning_rate": 3.578625954198473e-05,
      "loss": 0.048,
      "step": 2108
    },
    {
      "epoch": 12.86,
      "learning_rate": 3.57557251908397e-05,
      "loss": 0.0599,
      "step": 2109
    },
    {
      "epoch": 12.87,
      "learning_rate": 3.5725190839694654e-05,
      "loss": 0.064,
      "step": 2110
    },
    {
      "epoch": 12.87,
      "learning_rate": 3.569465648854962e-05,
      "loss": 0.0596,
      "step": 2111
    },
    {
      "epoch": 12.88,
      "learning_rate": 3.566412213740458e-05,
      "loss": 0.0818,
      "step": 2112
    },
    {
      "epoch": 12.88,
      "learning_rate": 3.563358778625954e-05,
      "loss": 0.0568,
      "step": 2113
    },
    {
      "epoch": 12.89,
      "learning_rate": 3.5603053435114506e-05,
      "loss": 0.0493,
      "step": 2114
    },
    {
      "epoch": 12.9,
      "learning_rate": 3.557251908396947e-05,
      "loss": 0.0618,
      "step": 2115
    },
    {
      "epoch": 12.9,
      "learning_rate": 3.554198473282443e-05,
      "loss": 0.0714,
      "step": 2116
    },
    {
      "epoch": 12.91,
      "learning_rate": 3.551145038167939e-05,
      "loss": 0.0766,
      "step": 2117
    },
    {
      "epoch": 12.91,
      "learning_rate": 3.548091603053435e-05,
      "loss": 0.0581,
      "step": 2118
    },
    {
      "epoch": 12.92,
      "learning_rate": 3.5450381679389315e-05,
      "loss": 0.0568,
      "step": 2119
    },
    {
      "epoch": 12.93,
      "learning_rate": 3.541984732824428e-05,
      "loss": 0.0631,
      "step": 2120
    },
    {
      "epoch": 12.93,
      "learning_rate": 3.538931297709924e-05,
      "loss": 0.0711,
      "step": 2121
    },
    {
      "epoch": 12.94,
      "learning_rate": 3.5358778625954195e-05,
      "loss": 0.0607,
      "step": 2122
    },
    {
      "epoch": 12.95,
      "learning_rate": 3.5328244274809166e-05,
      "loss": 0.0576,
      "step": 2123
    },
    {
      "epoch": 12.95,
      "learning_rate": 3.5297709923664124e-05,
      "loss": 0.0573,
      "step": 2124
    },
    {
      "epoch": 12.96,
      "learning_rate": 3.526717557251908e-05,
      "loss": 0.0668,
      "step": 2125
    },
    {
      "epoch": 12.96,
      "learning_rate": 3.5236641221374046e-05,
      "loss": 0.0588,
      "step": 2126
    },
    {
      "epoch": 12.97,
      "learning_rate": 3.520610687022901e-05,
      "loss": 0.0521,
      "step": 2127
    },
    {
      "epoch": 12.98,
      "learning_rate": 3.517557251908397e-05,
      "loss": 0.0625,
      "step": 2128
    },
    {
      "epoch": 12.98,
      "learning_rate": 3.514503816793893e-05,
      "loss": 0.0593,
      "step": 2129
    },
    {
      "epoch": 12.99,
      "learning_rate": 3.511450381679389e-05,
      "loss": 0.0617,
      "step": 2130
    },
    {
      "epoch": 12.99,
      "learning_rate": 3.508396946564886e-05,
      "loss": 0.0905,
      "step": 2131
    },
    {
      "epoch": 13.0,
      "learning_rate": 3.505343511450382e-05,
      "loss": 0.0867,
      "step": 2132
    },
    {
      "epoch": 13.01,
      "learning_rate": 3.502290076335878e-05,
      "loss": 0.0559,
      "step": 2133
    },
    {
      "epoch": 13.01,
      "learning_rate": 3.499236641221374e-05,
      "loss": 0.0575,
      "step": 2134
    },
    {
      "epoch": 13.02,
      "learning_rate": 3.4961832061068706e-05,
      "loss": 0.0592,
      "step": 2135
    },
    {
      "epoch": 13.02,
      "learning_rate": 3.4931297709923664e-05,
      "loss": 0.0519,
      "step": 2136
    },
    {
      "epoch": 13.03,
      "learning_rate": 3.490076335877863e-05,
      "loss": 0.0512,
      "step": 2137
    },
    {
      "epoch": 13.04,
      "learning_rate": 3.4870229007633586e-05,
      "loss": 0.0525,
      "step": 2138
    },
    {
      "epoch": 13.04,
      "learning_rate": 3.483969465648855e-05,
      "loss": 0.049,
      "step": 2139
    },
    {
      "epoch": 13.05,
      "learning_rate": 3.4809160305343515e-05,
      "loss": 0.0662,
      "step": 2140
    },
    {
      "epoch": 13.05,
      "learning_rate": 3.477862595419847e-05,
      "loss": 0.0628,
      "step": 2141
    },
    {
      "epoch": 13.06,
      "learning_rate": 3.474809160305344e-05,
      "loss": 0.0555,
      "step": 2142
    },
    {
      "epoch": 13.07,
      "learning_rate": 3.47175572519084e-05,
      "loss": 0.0661,
      "step": 2143
    },
    {
      "epoch": 13.07,
      "learning_rate": 3.468702290076336e-05,
      "loss": 0.0577,
      "step": 2144
    },
    {
      "epoch": 13.08,
      "learning_rate": 3.4656488549618324e-05,
      "loss": 0.0638,
      "step": 2145
    },
    {
      "epoch": 13.09,
      "learning_rate": 3.462595419847328e-05,
      "loss": 0.0481,
      "step": 2146
    },
    {
      "epoch": 13.09,
      "learning_rate": 3.4595419847328246e-05,
      "loss": 0.0494,
      "step": 2147
    },
    {
      "epoch": 13.1,
      "learning_rate": 3.456488549618321e-05,
      "loss": 0.0574,
      "step": 2148
    },
    {
      "epoch": 13.1,
      "learning_rate": 3.453435114503817e-05,
      "loss": 0.0541,
      "step": 2149
    },
    {
      "epoch": 13.11,
      "learning_rate": 3.4503816793893126e-05,
      "loss": 0.0572,
      "step": 2150
    },
    {
      "epoch": 13.12,
      "learning_rate": 3.44732824427481e-05,
      "loss": 0.0433,
      "step": 2151
    },
    {
      "epoch": 13.12,
      "learning_rate": 3.4442748091603055e-05,
      "loss": 0.0568,
      "step": 2152
    },
    {
      "epoch": 13.13,
      "learning_rate": 3.441221374045801e-05,
      "loss": 0.0524,
      "step": 2153
    },
    {
      "epoch": 13.13,
      "learning_rate": 3.438167938931298e-05,
      "loss": 0.0585,
      "step": 2154
    },
    {
      "epoch": 13.14,
      "learning_rate": 3.435114503816794e-05,
      "loss": 0.0558,
      "step": 2155
    },
    {
      "epoch": 13.15,
      "learning_rate": 3.43206106870229e-05,
      "loss": 0.0624,
      "step": 2156
    },
    {
      "epoch": 13.15,
      "learning_rate": 3.4290076335877864e-05,
      "loss": 0.0562,
      "step": 2157
    },
    {
      "epoch": 13.16,
      "learning_rate": 3.425954198473282e-05,
      "loss": 0.0625,
      "step": 2158
    },
    {
      "epoch": 13.16,
      "learning_rate": 3.4229007633587786e-05,
      "loss": 0.0517,
      "step": 2159
    },
    {
      "epoch": 13.17,
      "learning_rate": 3.419847328244275e-05,
      "loss": 0.0649,
      "step": 2160
    },
    {
      "epoch": 13.18,
      "learning_rate": 3.416793893129771e-05,
      "loss": 0.0569,
      "step": 2161
    },
    {
      "epoch": 13.18,
      "learning_rate": 3.413740458015267e-05,
      "loss": 0.0546,
      "step": 2162
    },
    {
      "epoch": 13.19,
      "learning_rate": 3.410687022900764e-05,
      "loss": 0.0702,
      "step": 2163
    },
    {
      "epoch": 13.2,
      "learning_rate": 3.4076335877862595e-05,
      "loss": 0.0566,
      "step": 2164
    },
    {
      "epoch": 13.2,
      "learning_rate": 3.404580152671756e-05,
      "loss": 0.0553,
      "step": 2165
    },
    {
      "epoch": 13.21,
      "learning_rate": 3.401526717557252e-05,
      "loss": 0.062,
      "step": 2166
    },
    {
      "epoch": 13.21,
      "learning_rate": 3.398473282442748e-05,
      "loss": 0.0755,
      "step": 2167
    },
    {
      "epoch": 13.22,
      "learning_rate": 3.3954198473282446e-05,
      "loss": 0.0539,
      "step": 2168
    },
    {
      "epoch": 13.23,
      "learning_rate": 3.3923664122137404e-05,
      "loss": 0.0575,
      "step": 2169
    },
    {
      "epoch": 13.23,
      "learning_rate": 3.389312977099237e-05,
      "loss": 0.0668,
      "step": 2170
    },
    {
      "epoch": 13.24,
      "learning_rate": 3.386259541984733e-05,
      "loss": 0.0556,
      "step": 2171
    },
    {
      "epoch": 13.24,
      "learning_rate": 3.383206106870229e-05,
      "loss": 0.0504,
      "step": 2172
    },
    {
      "epoch": 13.25,
      "learning_rate": 3.3801526717557255e-05,
      "loss": 0.0686,
      "step": 2173
    },
    {
      "epoch": 13.26,
      "learning_rate": 3.377099236641221e-05,
      "loss": 0.0577,
      "step": 2174
    },
    {
      "epoch": 13.26,
      "learning_rate": 3.374045801526718e-05,
      "loss": 0.0462,
      "step": 2175
    },
    {
      "epoch": 13.27,
      "learning_rate": 3.370992366412214e-05,
      "loss": 0.0608,
      "step": 2176
    },
    {
      "epoch": 13.27,
      "learning_rate": 3.36793893129771e-05,
      "loss": 0.0654,
      "step": 2177
    },
    {
      "epoch": 13.28,
      "learning_rate": 3.364885496183206e-05,
      "loss": 0.0609,
      "step": 2178
    },
    {
      "epoch": 13.29,
      "learning_rate": 3.361832061068703e-05,
      "loss": 0.0605,
      "step": 2179
    },
    {
      "epoch": 13.29,
      "learning_rate": 3.358778625954199e-05,
      "loss": 0.0523,
      "step": 2180
    },
    {
      "epoch": 13.3,
      "learning_rate": 3.3557251908396944e-05,
      "loss": 0.045,
      "step": 2181
    },
    {
      "epoch": 13.3,
      "learning_rate": 3.352671755725191e-05,
      "loss": 0.0643,
      "step": 2182
    },
    {
      "epoch": 13.31,
      "learning_rate": 3.349618320610687e-05,
      "loss": 0.0574,
      "step": 2183
    },
    {
      "epoch": 13.32,
      "learning_rate": 3.346564885496183e-05,
      "loss": 0.0556,
      "step": 2184
    },
    {
      "epoch": 13.32,
      "learning_rate": 3.3435114503816796e-05,
      "loss": 0.0495,
      "step": 2185
    },
    {
      "epoch": 13.33,
      "learning_rate": 3.340458015267175e-05,
      "loss": 0.0547,
      "step": 2186
    },
    {
      "epoch": 13.34,
      "learning_rate": 3.337404580152672e-05,
      "loss": 0.0649,
      "step": 2187
    },
    {
      "epoch": 13.34,
      "learning_rate": 3.334351145038168e-05,
      "loss": 0.062,
      "step": 2188
    },
    {
      "epoch": 13.35,
      "learning_rate": 3.331297709923664e-05,
      "loss": 0.0572,
      "step": 2189
    },
    {
      "epoch": 13.35,
      "learning_rate": 3.3282442748091605e-05,
      "loss": 0.0596,
      "step": 2190
    },
    {
      "epoch": 13.36,
      "learning_rate": 3.325190839694657e-05,
      "loss": 0.0545,
      "step": 2191
    },
    {
      "epoch": 13.37,
      "learning_rate": 3.322137404580153e-05,
      "loss": 0.0577,
      "step": 2192
    },
    {
      "epoch": 13.37,
      "learning_rate": 3.319083969465649e-05,
      "loss": 0.0577,
      "step": 2193
    },
    {
      "epoch": 13.38,
      "learning_rate": 3.316030534351145e-05,
      "loss": 0.0458,
      "step": 2194
    },
    {
      "epoch": 13.38,
      "learning_rate": 3.3129770992366414e-05,
      "loss": 0.0586,
      "step": 2195
    },
    {
      "epoch": 13.39,
      "learning_rate": 3.309923664122138e-05,
      "loss": 0.0539,
      "step": 2196
    },
    {
      "epoch": 13.4,
      "learning_rate": 3.3068702290076336e-05,
      "loss": 0.0613,
      "step": 2197
    },
    {
      "epoch": 13.4,
      "learning_rate": 3.30381679389313e-05,
      "loss": 0.0619,
      "step": 2198
    },
    {
      "epoch": 13.41,
      "learning_rate": 3.3007633587786265e-05,
      "loss": 0.0561,
      "step": 2199
    },
    {
      "epoch": 13.41,
      "learning_rate": 3.297709923664122e-05,
      "loss": 0.0563,
      "step": 2200
    },
    {
      "epoch": 13.42,
      "learning_rate": 3.294656488549619e-05,
      "loss": 0.0524,
      "step": 2201
    },
    {
      "epoch": 13.43,
      "learning_rate": 3.2916030534351145e-05,
      "loss": 0.0696,
      "step": 2202
    },
    {
      "epoch": 13.43,
      "learning_rate": 3.288549618320611e-05,
      "loss": 0.0535,
      "step": 2203
    },
    {
      "epoch": 13.44,
      "learning_rate": 3.2854961832061074e-05,
      "loss": 0.0547,
      "step": 2204
    },
    {
      "epoch": 13.45,
      "learning_rate": 3.282442748091603e-05,
      "loss": 0.061,
      "step": 2205
    },
    {
      "epoch": 13.45,
      "learning_rate": 3.279389312977099e-05,
      "loss": 0.0517,
      "step": 2206
    },
    {
      "epoch": 13.46,
      "learning_rate": 3.276335877862596e-05,
      "loss": 0.0656,
      "step": 2207
    },
    {
      "epoch": 13.46,
      "learning_rate": 3.273282442748092e-05,
      "loss": 0.0568,
      "step": 2208
    },
    {
      "epoch": 13.47,
      "learning_rate": 3.2702290076335876e-05,
      "loss": 0.0668,
      "step": 2209
    },
    {
      "epoch": 13.48,
      "learning_rate": 3.267175572519084e-05,
      "loss": 0.063,
      "step": 2210
    },
    {
      "epoch": 13.48,
      "learning_rate": 3.2641221374045805e-05,
      "loss": 0.0598,
      "step": 2211
    },
    {
      "epoch": 13.49,
      "learning_rate": 3.261068702290076e-05,
      "loss": 0.0634,
      "step": 2212
    },
    {
      "epoch": 13.49,
      "learning_rate": 3.258015267175573e-05,
      "loss": 0.0507,
      "step": 2213
    },
    {
      "epoch": 13.5,
      "learning_rate": 3.2549618320610685e-05,
      "loss": 0.0576,
      "step": 2214
    },
    {
      "epoch": 13.51,
      "learning_rate": 3.251908396946565e-05,
      "loss": 0.0502,
      "step": 2215
    },
    {
      "epoch": 13.51,
      "learning_rate": 3.2488549618320614e-05,
      "loss": 0.0719,
      "step": 2216
    },
    {
      "epoch": 13.52,
      "learning_rate": 3.245801526717557e-05,
      "loss": 0.0564,
      "step": 2217
    },
    {
      "epoch": 13.52,
      "learning_rate": 3.2427480916030536e-05,
      "loss": 0.0554,
      "step": 2218
    },
    {
      "epoch": 13.53,
      "learning_rate": 3.23969465648855e-05,
      "loss": 0.0596,
      "step": 2219
    },
    {
      "epoch": 13.54,
      "learning_rate": 3.236641221374046e-05,
      "loss": 0.0493,
      "step": 2220
    },
    {
      "epoch": 13.54,
      "learning_rate": 3.233587786259542e-05,
      "loss": 0.0444,
      "step": 2221
    },
    {
      "epoch": 13.55,
      "learning_rate": 3.230534351145038e-05,
      "loss": 0.062,
      "step": 2222
    },
    {
      "epoch": 13.55,
      "learning_rate": 3.2274809160305345e-05,
      "loss": 0.0531,
      "step": 2223
    },
    {
      "epoch": 13.56,
      "learning_rate": 3.224427480916031e-05,
      "loss": 0.0877,
      "step": 2224
    },
    {
      "epoch": 13.57,
      "learning_rate": 3.221374045801527e-05,
      "loss": 0.0714,
      "step": 2225
    },
    {
      "epoch": 13.57,
      "learning_rate": 3.2183206106870225e-05,
      "loss": 0.0611,
      "step": 2226
    },
    {
      "epoch": 13.58,
      "learning_rate": 3.2152671755725196e-05,
      "loss": 0.051,
      "step": 2227
    },
    {
      "epoch": 13.59,
      "learning_rate": 3.2122137404580154e-05,
      "loss": 0.0533,
      "step": 2228
    },
    {
      "epoch": 13.59,
      "learning_rate": 3.209160305343512e-05,
      "loss": 0.0587,
      "step": 2229
    },
    {
      "epoch": 13.6,
      "learning_rate": 3.2061068702290076e-05,
      "loss": 0.0664,
      "step": 2230
    },
    {
      "epoch": 13.6,
      "learning_rate": 3.203053435114504e-05,
      "loss": 0.0642,
      "step": 2231
    },
    {
      "epoch": 13.61,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.0497,
      "step": 2232
    },
    {
      "epoch": 13.62,
      "learning_rate": 3.196946564885496e-05,
      "loss": 0.0675,
      "step": 2233
    },
    {
      "epoch": 13.62,
      "learning_rate": 3.193893129770992e-05,
      "loss": 0.0565,
      "step": 2234
    },
    {
      "epoch": 13.63,
      "learning_rate": 3.190839694656489e-05,
      "loss": 0.0745,
      "step": 2235
    },
    {
      "epoch": 13.63,
      "learning_rate": 3.187786259541985e-05,
      "loss": 0.0618,
      "step": 2236
    },
    {
      "epoch": 13.64,
      "learning_rate": 3.184732824427481e-05,
      "loss": 0.055,
      "step": 2237
    },
    {
      "epoch": 13.65,
      "learning_rate": 3.181679389312977e-05,
      "loss": 0.0641,
      "step": 2238
    },
    {
      "epoch": 13.65,
      "learning_rate": 3.1786259541984737e-05,
      "loss": 0.0504,
      "step": 2239
    },
    {
      "epoch": 13.66,
      "learning_rate": 3.1755725190839694e-05,
      "loss": 0.0588,
      "step": 2240
    },
    {
      "epoch": 13.66,
      "learning_rate": 3.172519083969466e-05,
      "loss": 0.0507,
      "step": 2241
    },
    {
      "epoch": 13.67,
      "learning_rate": 3.1694656488549616e-05,
      "loss": 0.0525,
      "step": 2242
    },
    {
      "epoch": 13.68,
      "learning_rate": 3.166412213740458e-05,
      "loss": 0.0492,
      "step": 2243
    },
    {
      "epoch": 13.68,
      "learning_rate": 3.1633587786259545e-05,
      "loss": 0.0629,
      "step": 2244
    },
    {
      "epoch": 13.69,
      "learning_rate": 3.16030534351145e-05,
      "loss": 0.0544,
      "step": 2245
    },
    {
      "epoch": 13.7,
      "learning_rate": 3.157251908396947e-05,
      "loss": 0.0512,
      "step": 2246
    },
    {
      "epoch": 13.7,
      "learning_rate": 3.154198473282443e-05,
      "loss": 0.0683,
      "step": 2247
    },
    {
      "epoch": 13.71,
      "learning_rate": 3.151145038167939e-05,
      "loss": 0.0697,
      "step": 2248
    },
    {
      "epoch": 13.71,
      "learning_rate": 3.1480916030534354e-05,
      "loss": 0.0637,
      "step": 2249
    },
    {
      "epoch": 13.72,
      "learning_rate": 3.145038167938931e-05,
      "loss": 0.0644,
      "step": 2250
    },
    {
      "epoch": 13.73,
      "learning_rate": 3.141984732824428e-05,
      "loss": 0.0562,
      "step": 2251
    },
    {
      "epoch": 13.73,
      "learning_rate": 3.138931297709924e-05,
      "loss": 0.059,
      "step": 2252
    },
    {
      "epoch": 13.74,
      "learning_rate": 3.13587786259542e-05,
      "loss": 0.0603,
      "step": 2253
    },
    {
      "epoch": 13.74,
      "learning_rate": 3.132824427480916e-05,
      "loss": 0.0687,
      "step": 2254
    },
    {
      "epoch": 13.75,
      "learning_rate": 3.129770992366413e-05,
      "loss": 0.044,
      "step": 2255
    },
    {
      "epoch": 13.76,
      "learning_rate": 3.1267175572519086e-05,
      "loss": 0.0639,
      "step": 2256
    },
    {
      "epoch": 13.76,
      "learning_rate": 3.123664122137404e-05,
      "loss": 0.0699,
      "step": 2257
    },
    {
      "epoch": 13.77,
      "learning_rate": 3.120610687022901e-05,
      "loss": 0.0635,
      "step": 2258
    },
    {
      "epoch": 13.77,
      "learning_rate": 3.117557251908397e-05,
      "loss": 0.0504,
      "step": 2259
    },
    {
      "epoch": 13.78,
      "learning_rate": 3.114503816793894e-05,
      "loss": 0.0622,
      "step": 2260
    },
    {
      "epoch": 13.79,
      "learning_rate": 3.1114503816793895e-05,
      "loss": 0.05,
      "step": 2261
    },
    {
      "epoch": 13.79,
      "learning_rate": 3.108396946564885e-05,
      "loss": 0.0542,
      "step": 2262
    },
    {
      "epoch": 13.8,
      "learning_rate": 3.1053435114503824e-05,
      "loss": 0.0629,
      "step": 2263
    },
    {
      "epoch": 13.8,
      "learning_rate": 3.102290076335878e-05,
      "loss": 0.0516,
      "step": 2264
    },
    {
      "epoch": 13.81,
      "learning_rate": 3.099236641221374e-05,
      "loss": 0.0646,
      "step": 2265
    },
    {
      "epoch": 13.82,
      "learning_rate": 3.0961832061068704e-05,
      "loss": 0.0648,
      "step": 2266
    },
    {
      "epoch": 13.82,
      "learning_rate": 3.093129770992367e-05,
      "loss": 0.0601,
      "step": 2267
    },
    {
      "epoch": 13.83,
      "learning_rate": 3.0900763358778626e-05,
      "loss": 0.0698,
      "step": 2268
    },
    {
      "epoch": 13.84,
      "learning_rate": 3.087022900763359e-05,
      "loss": 0.0588,
      "step": 2269
    },
    {
      "epoch": 13.84,
      "learning_rate": 3.083969465648855e-05,
      "loss": 0.067,
      "step": 2270
    },
    {
      "epoch": 13.85,
      "learning_rate": 3.080916030534351e-05,
      "loss": 0.0727,
      "step": 2271
    },
    {
      "epoch": 13.85,
      "learning_rate": 3.077862595419848e-05,
      "loss": 0.0717,
      "step": 2272
    },
    {
      "epoch": 13.86,
      "learning_rate": 3.0748091603053435e-05,
      "loss": 0.068,
      "step": 2273
    },
    {
      "epoch": 13.87,
      "learning_rate": 3.07175572519084e-05,
      "loss": 0.052,
      "step": 2274
    },
    {
      "epoch": 13.87,
      "learning_rate": 3.0687022900763364e-05,
      "loss": 0.0542,
      "step": 2275
    },
    {
      "epoch": 13.88,
      "learning_rate": 3.065648854961832e-05,
      "loss": 0.0678,
      "step": 2276
    },
    {
      "epoch": 13.88,
      "learning_rate": 3.0625954198473286e-05,
      "loss": 0.0644,
      "step": 2277
    },
    {
      "epoch": 13.89,
      "learning_rate": 3.0595419847328244e-05,
      "loss": 0.072,
      "step": 2278
    },
    {
      "epoch": 13.9,
      "learning_rate": 3.056488549618321e-05,
      "loss": 0.0598,
      "step": 2279
    },
    {
      "epoch": 13.9,
      "learning_rate": 3.053435114503817e-05,
      "loss": 0.0966,
      "step": 2280
    },
    {
      "epoch": 13.91,
      "learning_rate": 3.050381679389313e-05,
      "loss": 0.0594,
      "step": 2281
    },
    {
      "epoch": 13.91,
      "learning_rate": 3.047328244274809e-05,
      "loss": 0.0561,
      "step": 2282
    },
    {
      "epoch": 13.92,
      "learning_rate": 3.0442748091603056e-05,
      "loss": 0.0574,
      "step": 2283
    },
    {
      "epoch": 13.93,
      "learning_rate": 3.0412213740458017e-05,
      "loss": 0.0612,
      "step": 2284
    },
    {
      "epoch": 13.93,
      "learning_rate": 3.038167938931298e-05,
      "loss": 0.0605,
      "step": 2285
    },
    {
      "epoch": 13.94,
      "learning_rate": 3.035114503816794e-05,
      "loss": 0.0794,
      "step": 2286
    },
    {
      "epoch": 13.95,
      "learning_rate": 3.0320610687022904e-05,
      "loss": 0.0575,
      "step": 2287
    },
    {
      "epoch": 13.95,
      "learning_rate": 3.0290076335877865e-05,
      "loss": 0.0622,
      "step": 2288
    },
    {
      "epoch": 13.96,
      "learning_rate": 3.0259541984732826e-05,
      "loss": 0.1082,
      "step": 2289
    },
    {
      "epoch": 13.96,
      "learning_rate": 3.0229007633587787e-05,
      "loss": 0.0783,
      "step": 2290
    },
    {
      "epoch": 13.97,
      "learning_rate": 3.0198473282442752e-05,
      "loss": 0.0561,
      "step": 2291
    },
    {
      "epoch": 13.98,
      "learning_rate": 3.0167938931297713e-05,
      "loss": 0.058,
      "step": 2292
    },
    {
      "epoch": 13.98,
      "learning_rate": 3.0137404580152674e-05,
      "loss": 0.0549,
      "step": 2293
    },
    {
      "epoch": 13.99,
      "learning_rate": 3.0106870229007632e-05,
      "loss": 0.0713,
      "step": 2294
    },
    {
      "epoch": 13.99,
      "learning_rate": 3.00763358778626e-05,
      "loss": 0.0594,
      "step": 2295
    },
    {
      "epoch": 14.0,
      "learning_rate": 3.004580152671756e-05,
      "loss": 0.0476,
      "step": 2296
    },
    {
      "epoch": 14.01,
      "learning_rate": 3.001526717557252e-05,
      "loss": 0.0481,
      "step": 2297
    },
    {
      "epoch": 14.01,
      "learning_rate": 2.998473282442748e-05,
      "loss": 0.0508,
      "step": 2298
    },
    {
      "epoch": 14.02,
      "learning_rate": 2.9954198473282448e-05,
      "loss": 0.0622,
      "step": 2299
    },
    {
      "epoch": 14.02,
      "learning_rate": 2.992366412213741e-05,
      "loss": 0.0534,
      "step": 2300
    },
    {
      "epoch": 14.03,
      "learning_rate": 2.9893129770992366e-05,
      "loss": 0.0517,
      "step": 2301
    },
    {
      "epoch": 14.04,
      "learning_rate": 2.9862595419847327e-05,
      "loss": 0.0606,
      "step": 2302
    },
    {
      "epoch": 14.04,
      "learning_rate": 2.9832061068702295e-05,
      "loss": 0.0568,
      "step": 2303
    },
    {
      "epoch": 14.05,
      "learning_rate": 2.9801526717557253e-05,
      "loss": 0.0629,
      "step": 2304
    },
    {
      "epoch": 14.05,
      "learning_rate": 2.9770992366412214e-05,
      "loss": 0.0576,
      "step": 2305
    },
    {
      "epoch": 14.06,
      "learning_rate": 2.9740458015267175e-05,
      "loss": 0.0499,
      "step": 2306
    },
    {
      "epoch": 14.07,
      "learning_rate": 2.970992366412214e-05,
      "loss": 0.0572,
      "step": 2307
    },
    {
      "epoch": 14.07,
      "learning_rate": 2.96793893129771e-05,
      "loss": 0.0551,
      "step": 2308
    },
    {
      "epoch": 14.08,
      "learning_rate": 2.9648854961832062e-05,
      "loss": 0.0561,
      "step": 2309
    },
    {
      "epoch": 14.09,
      "learning_rate": 2.9618320610687023e-05,
      "loss": 0.0549,
      "step": 2310
    },
    {
      "epoch": 14.09,
      "learning_rate": 2.9587786259541988e-05,
      "loss": 0.0498,
      "step": 2311
    },
    {
      "epoch": 14.1,
      "learning_rate": 2.955725190839695e-05,
      "loss": 0.046,
      "step": 2312
    },
    {
      "epoch": 14.1,
      "learning_rate": 2.952671755725191e-05,
      "loss": 0.0664,
      "step": 2313
    },
    {
      "epoch": 14.11,
      "learning_rate": 2.949618320610687e-05,
      "loss": 0.0619,
      "step": 2314
    },
    {
      "epoch": 14.12,
      "learning_rate": 2.9465648854961836e-05,
      "loss": 0.061,
      "step": 2315
    },
    {
      "epoch": 14.12,
      "learning_rate": 2.9435114503816797e-05,
      "loss": 0.043,
      "step": 2316
    },
    {
      "epoch": 14.13,
      "learning_rate": 2.9404580152671758e-05,
      "loss": 0.0573,
      "step": 2317
    },
    {
      "epoch": 14.13,
      "learning_rate": 2.937404580152672e-05,
      "loss": 0.0523,
      "step": 2318
    },
    {
      "epoch": 14.14,
      "learning_rate": 2.9343511450381683e-05,
      "loss": 0.0536,
      "step": 2319
    },
    {
      "epoch": 14.15,
      "learning_rate": 2.9312977099236644e-05,
      "loss": 0.043,
      "step": 2320
    },
    {
      "epoch": 14.15,
      "learning_rate": 2.9282442748091606e-05,
      "loss": 0.058,
      "step": 2321
    },
    {
      "epoch": 14.16,
      "learning_rate": 2.9251908396946563e-05,
      "loss": 0.0523,
      "step": 2322
    },
    {
      "epoch": 14.16,
      "learning_rate": 2.922137404580153e-05,
      "loss": 0.0498,
      "step": 2323
    },
    {
      "epoch": 14.17,
      "learning_rate": 2.9190839694656492e-05,
      "loss": 0.0573,
      "step": 2324
    },
    {
      "epoch": 14.18,
      "learning_rate": 2.916030534351145e-05,
      "loss": 0.0595,
      "step": 2325
    },
    {
      "epoch": 14.18,
      "learning_rate": 2.912977099236641e-05,
      "loss": 0.0466,
      "step": 2326
    },
    {
      "epoch": 14.19,
      "learning_rate": 2.909923664122138e-05,
      "loss": 0.0675,
      "step": 2327
    },
    {
      "epoch": 14.2,
      "learning_rate": 2.9068702290076337e-05,
      "loss": 0.0525,
      "step": 2328
    },
    {
      "epoch": 14.2,
      "learning_rate": 2.9038167938931298e-05,
      "loss": 0.0602,
      "step": 2329
    },
    {
      "epoch": 14.21,
      "learning_rate": 2.900763358778626e-05,
      "loss": 0.0535,
      "step": 2330
    },
    {
      "epoch": 14.21,
      "learning_rate": 2.8977099236641227e-05,
      "loss": 0.0575,
      "step": 2331
    },
    {
      "epoch": 14.22,
      "learning_rate": 2.8946564885496185e-05,
      "loss": 0.051,
      "step": 2332
    },
    {
      "epoch": 14.23,
      "learning_rate": 2.8916030534351146e-05,
      "loss": 0.0548,
      "step": 2333
    },
    {
      "epoch": 14.23,
      "learning_rate": 2.8885496183206107e-05,
      "loss": 0.0732,
      "step": 2334
    },
    {
      "epoch": 14.24,
      "learning_rate": 2.885496183206107e-05,
      "loss": 0.0555,
      "step": 2335
    },
    {
      "epoch": 14.24,
      "learning_rate": 2.8824427480916033e-05,
      "loss": 0.0478,
      "step": 2336
    },
    {
      "epoch": 14.25,
      "learning_rate": 2.8793893129770994e-05,
      "loss": 0.0631,
      "step": 2337
    },
    {
      "epoch": 14.26,
      "learning_rate": 2.8763358778625955e-05,
      "loss": 0.0726,
      "step": 2338
    },
    {
      "epoch": 14.26,
      "learning_rate": 2.873282442748092e-05,
      "loss": 0.0509,
      "step": 2339
    },
    {
      "epoch": 14.27,
      "learning_rate": 2.870229007633588e-05,
      "loss": 0.0448,
      "step": 2340
    },
    {
      "epoch": 14.27,
      "learning_rate": 2.867175572519084e-05,
      "loss": 0.0505,
      "step": 2341
    },
    {
      "epoch": 14.28,
      "learning_rate": 2.8641221374045803e-05,
      "loss": 0.0534,
      "step": 2342
    },
    {
      "epoch": 14.29,
      "learning_rate": 2.8610687022900767e-05,
      "loss": 0.0572,
      "step": 2343
    },
    {
      "epoch": 14.29,
      "learning_rate": 2.8580152671755728e-05,
      "loss": 0.0535,
      "step": 2344
    },
    {
      "epoch": 14.3,
      "learning_rate": 2.854961832061069e-05,
      "loss": 0.0558,
      "step": 2345
    },
    {
      "epoch": 14.3,
      "learning_rate": 2.8519083969465647e-05,
      "loss": 0.044,
      "step": 2346
    },
    {
      "epoch": 14.31,
      "learning_rate": 2.8488549618320615e-05,
      "loss": 0.0666,
      "step": 2347
    },
    {
      "epoch": 14.32,
      "learning_rate": 2.8458015267175576e-05,
      "loss": 0.059,
      "step": 2348
    },
    {
      "epoch": 14.32,
      "learning_rate": 2.8427480916030537e-05,
      "loss": 0.0695,
      "step": 2349
    },
    {
      "epoch": 14.33,
      "learning_rate": 2.8396946564885495e-05,
      "loss": 0.0635,
      "step": 2350
    },
    {
      "epoch": 14.34,
      "learning_rate": 2.8366412213740463e-05,
      "loss": 0.0586,
      "step": 2351
    },
    {
      "epoch": 14.34,
      "learning_rate": 2.8335877862595424e-05,
      "loss": 0.0622,
      "step": 2352
    },
    {
      "epoch": 14.35,
      "learning_rate": 2.830534351145038e-05,
      "loss": 0.0609,
      "step": 2353
    },
    {
      "epoch": 14.35,
      "learning_rate": 2.8274809160305343e-05,
      "loss": 0.0469,
      "step": 2354
    },
    {
      "epoch": 14.36,
      "learning_rate": 2.824427480916031e-05,
      "loss": 0.0535,
      "step": 2355
    },
    {
      "epoch": 14.37,
      "learning_rate": 2.821374045801527e-05,
      "loss": 0.0687,
      "step": 2356
    },
    {
      "epoch": 14.37,
      "learning_rate": 2.818320610687023e-05,
      "loss": 0.0546,
      "step": 2357
    },
    {
      "epoch": 14.38,
      "learning_rate": 2.815267175572519e-05,
      "loss": 0.0509,
      "step": 2358
    },
    {
      "epoch": 14.38,
      "learning_rate": 2.8122137404580155e-05,
      "loss": 0.0891,
      "step": 2359
    },
    {
      "epoch": 14.39,
      "learning_rate": 2.8091603053435116e-05,
      "loss": 0.0546,
      "step": 2360
    },
    {
      "epoch": 14.4,
      "learning_rate": 2.8061068702290077e-05,
      "loss": 0.0564,
      "step": 2361
    },
    {
      "epoch": 14.4,
      "learning_rate": 2.803053435114504e-05,
      "loss": 0.0488,
      "step": 2362
    },
    {
      "epoch": 14.41,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.0579,
      "step": 2363
    },
    {
      "epoch": 14.41,
      "learning_rate": 2.7969465648854964e-05,
      "loss": 0.0559,
      "step": 2364
    },
    {
      "epoch": 14.42,
      "learning_rate": 2.7938931297709925e-05,
      "loss": 0.0602,
      "step": 2365
    },
    {
      "epoch": 14.43,
      "learning_rate": 2.7908396946564886e-05,
      "loss": 0.0597,
      "step": 2366
    },
    {
      "epoch": 14.43,
      "learning_rate": 2.787786259541985e-05,
      "loss": 0.0758,
      "step": 2367
    },
    {
      "epoch": 14.44,
      "learning_rate": 2.7847328244274812e-05,
      "loss": 0.0613,
      "step": 2368
    },
    {
      "epoch": 14.45,
      "learning_rate": 2.7816793893129773e-05,
      "loss": 0.0717,
      "step": 2369
    },
    {
      "epoch": 14.45,
      "learning_rate": 2.7786259541984734e-05,
      "loss": 0.0575,
      "step": 2370
    },
    {
      "epoch": 14.46,
      "learning_rate": 2.77557251908397e-05,
      "loss": 0.0511,
      "step": 2371
    },
    {
      "epoch": 14.46,
      "learning_rate": 2.772519083969466e-05,
      "loss": 0.0509,
      "step": 2372
    },
    {
      "epoch": 14.47,
      "learning_rate": 2.769465648854962e-05,
      "loss": 0.0497,
      "step": 2373
    },
    {
      "epoch": 14.48,
      "learning_rate": 2.766412213740458e-05,
      "loss": 0.0601,
      "step": 2374
    },
    {
      "epoch": 14.48,
      "learning_rate": 2.7633587786259547e-05,
      "loss": 0.0593,
      "step": 2375
    },
    {
      "epoch": 14.49,
      "learning_rate": 2.7603053435114508e-05,
      "loss": 0.0674,
      "step": 2376
    },
    {
      "epoch": 14.49,
      "learning_rate": 2.7572519083969465e-05,
      "loss": 0.0471,
      "step": 2377
    },
    {
      "epoch": 14.5,
      "learning_rate": 2.7541984732824426e-05,
      "loss": 0.0592,
      "step": 2378
    },
    {
      "epoch": 14.51,
      "learning_rate": 2.7511450381679394e-05,
      "loss": 0.0615,
      "step": 2379
    },
    {
      "epoch": 14.51,
      "learning_rate": 2.7480916030534355e-05,
      "loss": 0.0557,
      "step": 2380
    },
    {
      "epoch": 14.52,
      "learning_rate": 2.7450381679389313e-05,
      "loss": 0.0593,
      "step": 2381
    },
    {
      "epoch": 14.52,
      "learning_rate": 2.7419847328244274e-05,
      "loss": 0.0498,
      "step": 2382
    },
    {
      "epoch": 14.53,
      "learning_rate": 2.7389312977099242e-05,
      "loss": 0.0596,
      "step": 2383
    },
    {
      "epoch": 14.54,
      "learning_rate": 2.73587786259542e-05,
      "loss": 0.07,
      "step": 2384
    },
    {
      "epoch": 14.54,
      "learning_rate": 2.732824427480916e-05,
      "loss": 0.0635,
      "step": 2385
    },
    {
      "epoch": 14.55,
      "learning_rate": 2.7297709923664122e-05,
      "loss": 0.0569,
      "step": 2386
    },
    {
      "epoch": 14.55,
      "learning_rate": 2.7267175572519083e-05,
      "loss": 0.0576,
      "step": 2387
    },
    {
      "epoch": 14.56,
      "learning_rate": 2.7236641221374048e-05,
      "loss": 0.0529,
      "step": 2388
    },
    {
      "epoch": 14.57,
      "learning_rate": 2.720610687022901e-05,
      "loss": 0.0528,
      "step": 2389
    },
    {
      "epoch": 14.57,
      "learning_rate": 2.717557251908397e-05,
      "loss": 0.0601,
      "step": 2390
    },
    {
      "epoch": 14.58,
      "learning_rate": 2.714503816793893e-05,
      "loss": 0.0538,
      "step": 2391
    },
    {
      "epoch": 14.59,
      "learning_rate": 2.7114503816793896e-05,
      "loss": 0.0597,
      "step": 2392
    },
    {
      "epoch": 14.59,
      "learning_rate": 2.7083969465648857e-05,
      "loss": 0.0704,
      "step": 2393
    },
    {
      "epoch": 14.6,
      "learning_rate": 2.7053435114503818e-05,
      "loss": 0.0644,
      "step": 2394
    },
    {
      "epoch": 14.6,
      "learning_rate": 2.7022900763358776e-05,
      "loss": 0.0646,
      "step": 2395
    },
    {
      "epoch": 14.61,
      "learning_rate": 2.6992366412213744e-05,
      "loss": 0.0556,
      "step": 2396
    },
    {
      "epoch": 14.62,
      "learning_rate": 2.6961832061068705e-05,
      "loss": 0.0518,
      "step": 2397
    },
    {
      "epoch": 14.62,
      "learning_rate": 2.6931297709923666e-05,
      "loss": 0.0625,
      "step": 2398
    },
    {
      "epoch": 14.63,
      "learning_rate": 2.6900763358778623e-05,
      "loss": 0.0531,
      "step": 2399
    },
    {
      "epoch": 14.63,
      "learning_rate": 2.687022900763359e-05,
      "loss": 0.0515,
      "step": 2400
    },
    {
      "epoch": 14.64,
      "learning_rate": 2.6839694656488552e-05,
      "loss": 0.0588,
      "step": 2401
    },
    {
      "epoch": 14.65,
      "learning_rate": 2.680916030534351e-05,
      "loss": 0.0544,
      "step": 2402
    },
    {
      "epoch": 14.65,
      "learning_rate": 2.677862595419847e-05,
      "loss": 0.049,
      "step": 2403
    },
    {
      "epoch": 14.66,
      "learning_rate": 2.674809160305344e-05,
      "loss": 0.0765,
      "step": 2404
    },
    {
      "epoch": 14.66,
      "learning_rate": 2.6717557251908397e-05,
      "loss": 0.0568,
      "step": 2405
    },
    {
      "epoch": 14.67,
      "learning_rate": 2.6687022900763358e-05,
      "loss": 0.0645,
      "step": 2406
    },
    {
      "epoch": 14.68,
      "learning_rate": 2.665648854961832e-05,
      "loss": 0.0607,
      "step": 2407
    },
    {
      "epoch": 14.68,
      "learning_rate": 2.6625954198473284e-05,
      "loss": 0.0689,
      "step": 2408
    },
    {
      "epoch": 14.69,
      "learning_rate": 2.6595419847328245e-05,
      "loss": 0.0607,
      "step": 2409
    },
    {
      "epoch": 14.7,
      "learning_rate": 2.6564885496183206e-05,
      "loss": 0.0603,
      "step": 2410
    },
    {
      "epoch": 14.7,
      "learning_rate": 2.6534351145038167e-05,
      "loss": 0.0487,
      "step": 2411
    },
    {
      "epoch": 14.71,
      "learning_rate": 2.650381679389313e-05,
      "loss": 0.0692,
      "step": 2412
    },
    {
      "epoch": 14.71,
      "learning_rate": 2.6473282442748093e-05,
      "loss": 0.0546,
      "step": 2413
    },
    {
      "epoch": 14.72,
      "learning_rate": 2.6442748091603054e-05,
      "loss": 0.0729,
      "step": 2414
    },
    {
      "epoch": 14.73,
      "learning_rate": 2.6412213740458015e-05,
      "loss": 0.0522,
      "step": 2415
    },
    {
      "epoch": 14.73,
      "learning_rate": 2.638167938931298e-05,
      "loss": 0.0457,
      "step": 2416
    },
    {
      "epoch": 14.74,
      "learning_rate": 2.635114503816794e-05,
      "loss": 0.0525,
      "step": 2417
    },
    {
      "epoch": 14.74,
      "learning_rate": 2.63206106870229e-05,
      "loss": 0.0605,
      "step": 2418
    },
    {
      "epoch": 14.75,
      "learning_rate": 2.6290076335877863e-05,
      "loss": 0.0697,
      "step": 2419
    },
    {
      "epoch": 14.76,
      "learning_rate": 2.6259541984732827e-05,
      "loss": 0.0482,
      "step": 2420
    },
    {
      "epoch": 14.76,
      "learning_rate": 2.622900763358779e-05,
      "loss": 0.0406,
      "step": 2421
    },
    {
      "epoch": 14.77,
      "learning_rate": 2.619847328244275e-05,
      "loss": 0.0679,
      "step": 2422
    },
    {
      "epoch": 14.77,
      "learning_rate": 2.6167938931297707e-05,
      "loss": 0.0787,
      "step": 2423
    },
    {
      "epoch": 14.78,
      "learning_rate": 2.6137404580152675e-05,
      "loss": 0.0731,
      "step": 2424
    },
    {
      "epoch": 14.79,
      "learning_rate": 2.6106870229007636e-05,
      "loss": 0.0701,
      "step": 2425
    },
    {
      "epoch": 14.79,
      "learning_rate": 2.6076335877862594e-05,
      "loss": 0.0674,
      "step": 2426
    },
    {
      "epoch": 14.8,
      "learning_rate": 2.6045801526717555e-05,
      "loss": 0.0537,
      "step": 2427
    },
    {
      "epoch": 14.8,
      "learning_rate": 2.6015267175572523e-05,
      "loss": 0.0602,
      "step": 2428
    },
    {
      "epoch": 14.81,
      "learning_rate": 2.5984732824427484e-05,
      "loss": 0.0687,
      "step": 2429
    },
    {
      "epoch": 14.82,
      "learning_rate": 2.5954198473282442e-05,
      "loss": 0.054,
      "step": 2430
    },
    {
      "epoch": 14.82,
      "learning_rate": 2.5923664122137403e-05,
      "loss": 0.0521,
      "step": 2431
    },
    {
      "epoch": 14.83,
      "learning_rate": 2.589312977099237e-05,
      "loss": 0.0802,
      "step": 2432
    },
    {
      "epoch": 14.84,
      "learning_rate": 2.586259541984733e-05,
      "loss": 0.0444,
      "step": 2433
    },
    {
      "epoch": 14.84,
      "learning_rate": 2.583206106870229e-05,
      "loss": 0.0571,
      "step": 2434
    },
    {
      "epoch": 14.85,
      "learning_rate": 2.580152671755725e-05,
      "loss": 0.0762,
      "step": 2435
    },
    {
      "epoch": 14.85,
      "learning_rate": 2.5770992366412215e-05,
      "loss": 0.0584,
      "step": 2436
    },
    {
      "epoch": 14.86,
      "learning_rate": 2.5740458015267176e-05,
      "loss": 0.0557,
      "step": 2437
    },
    {
      "epoch": 14.87,
      "learning_rate": 2.5709923664122137e-05,
      "loss": 0.0597,
      "step": 2438
    },
    {
      "epoch": 14.87,
      "learning_rate": 2.56793893129771e-05,
      "loss": 0.0739,
      "step": 2439
    },
    {
      "epoch": 14.88,
      "learning_rate": 2.5648854961832063e-05,
      "loss": 0.0573,
      "step": 2440
    },
    {
      "epoch": 14.88,
      "learning_rate": 2.5618320610687024e-05,
      "loss": 0.0704,
      "step": 2441
    },
    {
      "epoch": 14.89,
      "learning_rate": 2.5587786259541985e-05,
      "loss": 0.0608,
      "step": 2442
    },
    {
      "epoch": 14.9,
      "learning_rate": 2.5557251908396946e-05,
      "loss": 0.0662,
      "step": 2443
    },
    {
      "epoch": 14.9,
      "learning_rate": 2.552671755725191e-05,
      "loss": 0.0542,
      "step": 2444
    },
    {
      "epoch": 14.91,
      "learning_rate": 2.5496183206106872e-05,
      "loss": 0.0642,
      "step": 2445
    },
    {
      "epoch": 14.91,
      "learning_rate": 2.5465648854961833e-05,
      "loss": 0.0663,
      "step": 2446
    },
    {
      "epoch": 14.92,
      "learning_rate": 2.5435114503816794e-05,
      "loss": 0.0586,
      "step": 2447
    },
    {
      "epoch": 14.93,
      "learning_rate": 2.540458015267176e-05,
      "loss": 0.0669,
      "step": 2448
    },
    {
      "epoch": 14.93,
      "learning_rate": 2.537404580152672e-05,
      "loss": 0.0611,
      "step": 2449
    },
    {
      "epoch": 14.94,
      "learning_rate": 2.534351145038168e-05,
      "loss": 0.0581,
      "step": 2450
    },
    {
      "epoch": 14.95,
      "learning_rate": 2.531297709923664e-05,
      "loss": 0.053,
      "step": 2451
    },
    {
      "epoch": 14.95,
      "learning_rate": 2.5282442748091607e-05,
      "loss": 0.0664,
      "step": 2452
    },
    {
      "epoch": 14.96,
      "learning_rate": 2.5251908396946568e-05,
      "loss": 0.0637,
      "step": 2453
    },
    {
      "epoch": 14.96,
      "learning_rate": 2.5221374045801525e-05,
      "loss": 0.0526,
      "step": 2454
    },
    {
      "epoch": 14.97,
      "learning_rate": 2.5190839694656487e-05,
      "loss": 0.0504,
      "step": 2455
    },
    {
      "epoch": 14.98,
      "learning_rate": 2.5160305343511454e-05,
      "loss": 0.0622,
      "step": 2456
    },
    {
      "epoch": 14.98,
      "learning_rate": 2.5129770992366412e-05,
      "loss": 0.0525,
      "step": 2457
    },
    {
      "epoch": 14.99,
      "learning_rate": 2.5099236641221373e-05,
      "loss": 0.0643,
      "step": 2458
    },
    {
      "epoch": 14.99,
      "learning_rate": 2.5068702290076334e-05,
      "loss": 0.0669,
      "step": 2459
    },
    {
      "epoch": 15.0,
      "learning_rate": 2.50381679389313e-05,
      "loss": 0.0572,
      "step": 2460
    },
    {
      "epoch": 15.01,
      "learning_rate": 2.500763358778626e-05,
      "loss": 0.0587,
      "step": 2461
    },
    {
      "epoch": 15.01,
      "learning_rate": 2.497709923664122e-05,
      "loss": 0.0509,
      "step": 2462
    },
    {
      "epoch": 15.02,
      "learning_rate": 2.4946564885496186e-05,
      "loss": 0.0496,
      "step": 2463
    },
    {
      "epoch": 15.02,
      "learning_rate": 2.4916030534351147e-05,
      "loss": 0.0437,
      "step": 2464
    },
    {
      "epoch": 15.03,
      "learning_rate": 2.4885496183206108e-05,
      "loss": 0.0588,
      "step": 2465
    },
    {
      "epoch": 15.04,
      "learning_rate": 2.485496183206107e-05,
      "loss": 0.0486,
      "step": 2466
    },
    {
      "epoch": 15.04,
      "learning_rate": 2.4824427480916034e-05,
      "loss": 0.0619,
      "step": 2467
    },
    {
      "epoch": 15.05,
      "learning_rate": 2.479389312977099e-05,
      "loss": 0.0413,
      "step": 2468
    },
    {
      "epoch": 15.05,
      "learning_rate": 2.4763358778625956e-05,
      "loss": 0.0572,
      "step": 2469
    },
    {
      "epoch": 15.06,
      "learning_rate": 2.4732824427480917e-05,
      "loss": 0.0507,
      "step": 2470
    },
    {
      "epoch": 15.07,
      "learning_rate": 2.4702290076335878e-05,
      "loss": 0.0536,
      "step": 2471
    },
    {
      "epoch": 15.07,
      "learning_rate": 2.467175572519084e-05,
      "loss": 0.0532,
      "step": 2472
    },
    {
      "epoch": 15.08,
      "learning_rate": 2.4641221374045804e-05,
      "loss": 0.0574,
      "step": 2473
    },
    {
      "epoch": 15.09,
      "learning_rate": 2.4610687022900765e-05,
      "loss": 0.0607,
      "step": 2474
    },
    {
      "epoch": 15.09,
      "learning_rate": 2.4580152671755726e-05,
      "loss": 0.0674,
      "step": 2475
    },
    {
      "epoch": 15.1,
      "learning_rate": 2.4549618320610687e-05,
      "loss": 0.0482,
      "step": 2476
    },
    {
      "epoch": 15.1,
      "learning_rate": 2.451908396946565e-05,
      "loss": 0.0593,
      "step": 2477
    },
    {
      "epoch": 15.11,
      "learning_rate": 2.448854961832061e-05,
      "loss": 0.0653,
      "step": 2478
    },
    {
      "epoch": 15.12,
      "learning_rate": 2.4458015267175574e-05,
      "loss": 0.053,
      "step": 2479
    },
    {
      "epoch": 15.12,
      "learning_rate": 2.4427480916030535e-05,
      "loss": 0.064,
      "step": 2480
    },
    {
      "epoch": 15.13,
      "learning_rate": 2.43969465648855e-05,
      "loss": 0.0642,
      "step": 2481
    },
    {
      "epoch": 15.13,
      "learning_rate": 2.4366412213740457e-05,
      "loss": 0.0543,
      "step": 2482
    },
    {
      "epoch": 15.14,
      "learning_rate": 2.433587786259542e-05,
      "loss": 0.0522,
      "step": 2483
    },
    {
      "epoch": 15.15,
      "learning_rate": 2.4305343511450383e-05,
      "loss": 0.0614,
      "step": 2484
    },
    {
      "epoch": 15.15,
      "learning_rate": 2.4274809160305344e-05,
      "loss": 0.043,
      "step": 2485
    },
    {
      "epoch": 15.16,
      "learning_rate": 2.4244274809160305e-05,
      "loss": 0.0629,
      "step": 2486
    },
    {
      "epoch": 15.16,
      "learning_rate": 2.421374045801527e-05,
      "loss": 0.0652,
      "step": 2487
    },
    {
      "epoch": 15.17,
      "learning_rate": 2.418320610687023e-05,
      "loss": 0.0588,
      "step": 2488
    },
    {
      "epoch": 15.18,
      "learning_rate": 2.415267175572519e-05,
      "loss": 0.0656,
      "step": 2489
    },
    {
      "epoch": 15.18,
      "learning_rate": 2.4122137404580153e-05,
      "loss": 0.0628,
      "step": 2490
    },
    {
      "epoch": 15.19,
      "learning_rate": 2.4091603053435117e-05,
      "loss": 0.0633,
      "step": 2491
    },
    {
      "epoch": 15.2,
      "learning_rate": 2.4061068702290075e-05,
      "loss": 0.0581,
      "step": 2492
    },
    {
      "epoch": 15.2,
      "learning_rate": 2.403053435114504e-05,
      "loss": 0.0539,
      "step": 2493
    },
    {
      "epoch": 15.21,
      "learning_rate": 2.4e-05,
      "loss": 0.0526,
      "step": 2494
    },
    {
      "epoch": 15.21,
      "learning_rate": 2.3969465648854965e-05,
      "loss": 0.0602,
      "step": 2495
    },
    {
      "epoch": 15.22,
      "learning_rate": 2.3938931297709923e-05,
      "loss": 0.0529,
      "step": 2496
    },
    {
      "epoch": 15.23,
      "learning_rate": 2.3908396946564887e-05,
      "loss": 0.0542,
      "step": 2497
    },
    {
      "epoch": 15.23,
      "learning_rate": 2.387786259541985e-05,
      "loss": 0.0607,
      "step": 2498
    },
    {
      "epoch": 15.24,
      "learning_rate": 2.384732824427481e-05,
      "loss": 0.0521,
      "step": 2499
    },
    {
      "epoch": 15.24,
      "learning_rate": 2.381679389312977e-05,
      "loss": 0.0575,
      "step": 2500
    },
    {
      "epoch": 15.25,
      "learning_rate": 2.3786259541984735e-05,
      "loss": 0.0525,
      "step": 2501
    },
    {
      "epoch": 15.26,
      "learning_rate": 2.3755725190839696e-05,
      "loss": 0.0631,
      "step": 2502
    },
    {
      "epoch": 15.26,
      "learning_rate": 2.3725190839694657e-05,
      "loss": 0.0562,
      "step": 2503
    },
    {
      "epoch": 15.27,
      "learning_rate": 2.369465648854962e-05,
      "loss": 0.056,
      "step": 2504
    },
    {
      "epoch": 15.27,
      "learning_rate": 2.3664122137404583e-05,
      "loss": 0.0558,
      "step": 2505
    },
    {
      "epoch": 15.28,
      "learning_rate": 2.363358778625954e-05,
      "loss": 0.0494,
      "step": 2506
    },
    {
      "epoch": 15.29,
      "learning_rate": 2.3603053435114505e-05,
      "loss": 0.0523,
      "step": 2507
    },
    {
      "epoch": 15.29,
      "learning_rate": 2.3572519083969466e-05,
      "loss": 0.0645,
      "step": 2508
    },
    {
      "epoch": 15.3,
      "learning_rate": 2.3541984732824428e-05,
      "loss": 0.0505,
      "step": 2509
    },
    {
      "epoch": 15.3,
      "learning_rate": 2.351145038167939e-05,
      "loss": 0.0544,
      "step": 2510
    },
    {
      "epoch": 15.31,
      "learning_rate": 2.3480916030534353e-05,
      "loss": 0.0604,
      "step": 2511
    },
    {
      "epoch": 15.32,
      "learning_rate": 2.3450381679389314e-05,
      "loss": 0.0624,
      "step": 2512
    },
    {
      "epoch": 15.32,
      "learning_rate": 2.3419847328244275e-05,
      "loss": 0.0596,
      "step": 2513
    },
    {
      "epoch": 15.33,
      "learning_rate": 2.3389312977099236e-05,
      "loss": 0.0692,
      "step": 2514
    },
    {
      "epoch": 15.34,
      "learning_rate": 2.33587786259542e-05,
      "loss": 0.0506,
      "step": 2515
    },
    {
      "epoch": 15.34,
      "learning_rate": 2.3328244274809162e-05,
      "loss": 0.0509,
      "step": 2516
    },
    {
      "epoch": 15.35,
      "learning_rate": 2.3297709923664123e-05,
      "loss": 0.0471,
      "step": 2517
    },
    {
      "epoch": 15.35,
      "learning_rate": 2.3267175572519084e-05,
      "loss": 0.06,
      "step": 2518
    },
    {
      "epoch": 15.36,
      "learning_rate": 2.323664122137405e-05,
      "loss": 0.0456,
      "step": 2519
    },
    {
      "epoch": 15.37,
      "learning_rate": 2.3206106870229007e-05,
      "loss": 0.0465,
      "step": 2520
    },
    {
      "epoch": 15.37,
      "learning_rate": 2.317557251908397e-05,
      "loss": 0.062,
      "step": 2521
    },
    {
      "epoch": 15.38,
      "learning_rate": 2.3145038167938932e-05,
      "loss": 0.0541,
      "step": 2522
    },
    {
      "epoch": 15.38,
      "learning_rate": 2.3114503816793893e-05,
      "loss": 0.0494,
      "step": 2523
    },
    {
      "epoch": 15.39,
      "learning_rate": 2.3083969465648854e-05,
      "loss": 0.054,
      "step": 2524
    },
    {
      "epoch": 15.4,
      "learning_rate": 2.305343511450382e-05,
      "loss": 0.0622,
      "step": 2525
    },
    {
      "epoch": 15.4,
      "learning_rate": 2.302290076335878e-05,
      "loss": 0.0477,
      "step": 2526
    },
    {
      "epoch": 15.41,
      "learning_rate": 2.299236641221374e-05,
      "loss": 0.0616,
      "step": 2527
    },
    {
      "epoch": 15.41,
      "learning_rate": 2.2961832061068702e-05,
      "loss": 0.0659,
      "step": 2528
    },
    {
      "epoch": 15.42,
      "learning_rate": 2.2931297709923667e-05,
      "loss": 0.0583,
      "step": 2529
    },
    {
      "epoch": 15.43,
      "learning_rate": 2.2900763358778628e-05,
      "loss": 0.0698,
      "step": 2530
    },
    {
      "epoch": 15.43,
      "learning_rate": 2.287022900763359e-05,
      "loss": 0.0471,
      "step": 2531
    },
    {
      "epoch": 15.44,
      "learning_rate": 2.283969465648855e-05,
      "loss": 0.0553,
      "step": 2532
    },
    {
      "epoch": 15.45,
      "learning_rate": 2.2809160305343515e-05,
      "loss": 0.0685,
      "step": 2533
    },
    {
      "epoch": 15.45,
      "learning_rate": 2.2778625954198472e-05,
      "loss": 0.0495,
      "step": 2534
    },
    {
      "epoch": 15.46,
      "learning_rate": 2.2748091603053437e-05,
      "loss": 0.0555,
      "step": 2535
    },
    {
      "epoch": 15.46,
      "learning_rate": 2.2717557251908398e-05,
      "loss": 0.0569,
      "step": 2536
    },
    {
      "epoch": 15.47,
      "learning_rate": 2.268702290076336e-05,
      "loss": 0.0573,
      "step": 2537
    },
    {
      "epoch": 15.48,
      "learning_rate": 2.265648854961832e-05,
      "loss": 0.0448,
      "step": 2538
    },
    {
      "epoch": 15.48,
      "learning_rate": 2.2625954198473285e-05,
      "loss": 0.0632,
      "step": 2539
    },
    {
      "epoch": 15.49,
      "learning_rate": 2.2595419847328246e-05,
      "loss": 0.062,
      "step": 2540
    },
    {
      "epoch": 15.49,
      "learning_rate": 2.2564885496183207e-05,
      "loss": 0.0613,
      "step": 2541
    },
    {
      "epoch": 15.5,
      "learning_rate": 2.2534351145038168e-05,
      "loss": 0.0693,
      "step": 2542
    },
    {
      "epoch": 15.51,
      "learning_rate": 2.2503816793893133e-05,
      "loss": 0.0654,
      "step": 2543
    },
    {
      "epoch": 15.51,
      "learning_rate": 2.2473282442748094e-05,
      "loss": 0.0482,
      "step": 2544
    },
    {
      "epoch": 15.52,
      "learning_rate": 2.2442748091603055e-05,
      "loss": 0.0594,
      "step": 2545
    },
    {
      "epoch": 15.52,
      "learning_rate": 2.2412213740458016e-05,
      "loss": 0.0684,
      "step": 2546
    },
    {
      "epoch": 15.53,
      "learning_rate": 2.238167938931298e-05,
      "loss": 0.0608,
      "step": 2547
    },
    {
      "epoch": 15.54,
      "learning_rate": 2.2351145038167938e-05,
      "loss": 0.0551,
      "step": 2548
    },
    {
      "epoch": 15.54,
      "learning_rate": 2.2320610687022903e-05,
      "loss": 0.0658,
      "step": 2549
    },
    {
      "epoch": 15.55,
      "learning_rate": 2.2290076335877864e-05,
      "loss": 0.0534,
      "step": 2550
    },
    {
      "epoch": 15.55,
      "learning_rate": 2.2259541984732825e-05,
      "loss": 0.0664,
      "step": 2551
    },
    {
      "epoch": 15.56,
      "learning_rate": 2.2229007633587786e-05,
      "loss": 0.0635,
      "step": 2552
    },
    {
      "epoch": 15.57,
      "learning_rate": 2.219847328244275e-05,
      "loss": 0.0693,
      "step": 2553
    },
    {
      "epoch": 15.57,
      "learning_rate": 2.216793893129771e-05,
      "loss": 0.0638,
      "step": 2554
    },
    {
      "epoch": 15.58,
      "learning_rate": 2.2137404580152673e-05,
      "loss": 0.0622,
      "step": 2555
    },
    {
      "epoch": 15.59,
      "learning_rate": 2.2106870229007634e-05,
      "loss": 0.0619,
      "step": 2556
    },
    {
      "epoch": 15.59,
      "learning_rate": 2.20763358778626e-05,
      "loss": 0.0638,
      "step": 2557
    },
    {
      "epoch": 15.6,
      "learning_rate": 2.2045801526717556e-05,
      "loss": 0.0759,
      "step": 2558
    },
    {
      "epoch": 15.6,
      "learning_rate": 2.201526717557252e-05,
      "loss": 0.0657,
      "step": 2559
    },
    {
      "epoch": 15.61,
      "learning_rate": 2.198473282442748e-05,
      "loss": 0.0606,
      "step": 2560
    },
    {
      "epoch": 15.62,
      "learning_rate": 2.1954198473282446e-05,
      "loss": 0.0588,
      "step": 2561
    },
    {
      "epoch": 15.62,
      "learning_rate": 2.1923664122137404e-05,
      "loss": 0.061,
      "step": 2562
    },
    {
      "epoch": 15.63,
      "learning_rate": 2.189312977099237e-05,
      "loss": 0.0512,
      "step": 2563
    },
    {
      "epoch": 15.63,
      "learning_rate": 2.186259541984733e-05,
      "loss": 0.0653,
      "step": 2564
    },
    {
      "epoch": 15.64,
      "learning_rate": 2.183206106870229e-05,
      "loss": 0.055,
      "step": 2565
    },
    {
      "epoch": 15.65,
      "learning_rate": 2.1801526717557252e-05,
      "loss": 0.0566,
      "step": 2566
    },
    {
      "epoch": 15.65,
      "learning_rate": 2.1770992366412216e-05,
      "loss": 0.0729,
      "step": 2567
    },
    {
      "epoch": 15.66,
      "learning_rate": 2.1740458015267177e-05,
      "loss": 0.0501,
      "step": 2568
    },
    {
      "epoch": 15.66,
      "learning_rate": 2.170992366412214e-05,
      "loss": 0.0574,
      "step": 2569
    },
    {
      "epoch": 15.67,
      "learning_rate": 2.16793893129771e-05,
      "loss": 0.0693,
      "step": 2570
    },
    {
      "epoch": 15.68,
      "learning_rate": 2.1648854961832064e-05,
      "loss": 0.0479,
      "step": 2571
    },
    {
      "epoch": 15.68,
      "learning_rate": 2.1618320610687022e-05,
      "loss": 0.0449,
      "step": 2572
    },
    {
      "epoch": 15.69,
      "learning_rate": 2.1587786259541986e-05,
      "loss": 0.0462,
      "step": 2573
    },
    {
      "epoch": 15.7,
      "learning_rate": 2.1557251908396947e-05,
      "loss": 0.057,
      "step": 2574
    },
    {
      "epoch": 15.7,
      "learning_rate": 2.1526717557251912e-05,
      "loss": 0.0523,
      "step": 2575
    },
    {
      "epoch": 15.71,
      "learning_rate": 2.149618320610687e-05,
      "loss": 0.0606,
      "step": 2576
    },
    {
      "epoch": 15.71,
      "learning_rate": 2.1465648854961834e-05,
      "loss": 0.0494,
      "step": 2577
    },
    {
      "epoch": 15.72,
      "learning_rate": 2.1435114503816795e-05,
      "loss": 0.0473,
      "step": 2578
    },
    {
      "epoch": 15.73,
      "learning_rate": 2.1404580152671756e-05,
      "loss": 0.0571,
      "step": 2579
    },
    {
      "epoch": 15.73,
      "learning_rate": 2.1374045801526718e-05,
      "loss": 0.0375,
      "step": 2580
    },
    {
      "epoch": 15.74,
      "learning_rate": 2.1343511450381682e-05,
      "loss": 0.0562,
      "step": 2581
    },
    {
      "epoch": 15.74,
      "learning_rate": 2.1312977099236643e-05,
      "loss": 0.0561,
      "step": 2582
    },
    {
      "epoch": 15.75,
      "learning_rate": 2.1282442748091604e-05,
      "loss": 0.104,
      "step": 2583
    },
    {
      "epoch": 15.76,
      "learning_rate": 2.1251908396946565e-05,
      "loss": 0.075,
      "step": 2584
    },
    {
      "epoch": 15.76,
      "learning_rate": 2.122137404580153e-05,
      "loss": 0.0606,
      "step": 2585
    },
    {
      "epoch": 15.77,
      "learning_rate": 2.1190839694656488e-05,
      "loss": 0.0565,
      "step": 2586
    },
    {
      "epoch": 15.77,
      "learning_rate": 2.1160305343511452e-05,
      "loss": 0.0572,
      "step": 2587
    },
    {
      "epoch": 15.78,
      "learning_rate": 2.1129770992366413e-05,
      "loss": 0.0529,
      "step": 2588
    },
    {
      "epoch": 15.79,
      "learning_rate": 2.1099236641221374e-05,
      "loss": 0.0561,
      "step": 2589
    },
    {
      "epoch": 15.79,
      "learning_rate": 2.1068702290076335e-05,
      "loss": 0.0539,
      "step": 2590
    },
    {
      "epoch": 15.8,
      "learning_rate": 2.10381679389313e-05,
      "loss": 0.0686,
      "step": 2591
    },
    {
      "epoch": 15.8,
      "learning_rate": 2.100763358778626e-05,
      "loss": 0.051,
      "step": 2592
    },
    {
      "epoch": 15.81,
      "learning_rate": 2.0977099236641222e-05,
      "loss": 0.0663,
      "step": 2593
    },
    {
      "epoch": 15.82,
      "learning_rate": 2.0946564885496183e-05,
      "loss": 0.0687,
      "step": 2594
    },
    {
      "epoch": 15.82,
      "learning_rate": 2.0916030534351148e-05,
      "loss": 0.0487,
      "step": 2595
    },
    {
      "epoch": 15.83,
      "learning_rate": 2.088549618320611e-05,
      "loss": 0.0581,
      "step": 2596
    },
    {
      "epoch": 15.84,
      "learning_rate": 2.085496183206107e-05,
      "loss": 0.0635,
      "step": 2597
    },
    {
      "epoch": 15.84,
      "learning_rate": 2.082442748091603e-05,
      "loss": 0.0555,
      "step": 2598
    },
    {
      "epoch": 15.85,
      "learning_rate": 2.0793893129770996e-05,
      "loss": 0.0645,
      "step": 2599
    },
    {
      "epoch": 15.85,
      "learning_rate": 2.0763358778625953e-05,
      "loss": 0.0586,
      "step": 2600
    },
    {
      "epoch": 15.86,
      "learning_rate": 2.0732824427480918e-05,
      "loss": 0.0491,
      "step": 2601
    },
    {
      "epoch": 15.87,
      "learning_rate": 2.070229007633588e-05,
      "loss": 0.0605,
      "step": 2602
    },
    {
      "epoch": 15.87,
      "learning_rate": 2.067175572519084e-05,
      "loss": 0.053,
      "step": 2603
    },
    {
      "epoch": 15.88,
      "learning_rate": 2.06412213740458e-05,
      "loss": 0.0556,
      "step": 2604
    },
    {
      "epoch": 15.88,
      "learning_rate": 2.0610687022900766e-05,
      "loss": 0.049,
      "step": 2605
    },
    {
      "epoch": 15.89,
      "learning_rate": 2.0580152671755727e-05,
      "loss": 0.0543,
      "step": 2606
    },
    {
      "epoch": 15.9,
      "learning_rate": 2.0549618320610688e-05,
      "loss": 0.0626,
      "step": 2607
    },
    {
      "epoch": 15.9,
      "learning_rate": 2.051908396946565e-05,
      "loss": 0.0546,
      "step": 2608
    },
    {
      "epoch": 15.91,
      "learning_rate": 2.0488549618320614e-05,
      "loss": 0.0601,
      "step": 2609
    },
    {
      "epoch": 15.91,
      "learning_rate": 2.0458015267175575e-05,
      "loss": 0.0624,
      "step": 2610
    },
    {
      "epoch": 15.92,
      "learning_rate": 2.0427480916030536e-05,
      "loss": 0.0477,
      "step": 2611
    },
    {
      "epoch": 15.93,
      "learning_rate": 2.0396946564885497e-05,
      "loss": 0.0658,
      "step": 2612
    },
    {
      "epoch": 15.93,
      "learning_rate": 2.036641221374046e-05,
      "loss": 0.0459,
      "step": 2613
    },
    {
      "epoch": 15.94,
      "learning_rate": 2.033587786259542e-05,
      "loss": 0.0522,
      "step": 2614
    },
    {
      "epoch": 15.95,
      "learning_rate": 2.0305343511450384e-05,
      "loss": 0.0547,
      "step": 2615
    },
    {
      "epoch": 15.95,
      "learning_rate": 2.0274809160305345e-05,
      "loss": 0.0622,
      "step": 2616
    },
    {
      "epoch": 15.96,
      "learning_rate": 2.0244274809160306e-05,
      "loss": 0.0736,
      "step": 2617
    },
    {
      "epoch": 15.96,
      "learning_rate": 2.0213740458015267e-05,
      "loss": 0.0648,
      "step": 2618
    },
    {
      "epoch": 15.97,
      "learning_rate": 2.018320610687023e-05,
      "loss": 0.0614,
      "step": 2619
    },
    {
      "epoch": 15.98,
      "learning_rate": 2.0152671755725193e-05,
      "loss": 0.0668,
      "step": 2620
    },
    {
      "epoch": 15.98,
      "learning_rate": 2.0122137404580154e-05,
      "loss": 0.0583,
      "step": 2621
    },
    {
      "epoch": 15.99,
      "learning_rate": 2.0091603053435115e-05,
      "loss": 0.0475,
      "step": 2622
    },
    {
      "epoch": 15.99,
      "learning_rate": 2.006106870229008e-05,
      "loss": 0.0622,
      "step": 2623
    },
    {
      "epoch": 16.0,
      "learning_rate": 2.003053435114504e-05,
      "loss": 0.0426,
      "step": 2624
    },
    {
      "epoch": 16.01,
      "learning_rate": 2e-05,
      "loss": 0.0511,
      "step": 2625
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.9969465648854963e-05,
      "loss": 0.0572,
      "step": 2626
    },
    {
      "epoch": 16.02,
      "learning_rate": 1.9938931297709927e-05,
      "loss": 0.0645,
      "step": 2627
    },
    {
      "epoch": 16.02,
      "learning_rate": 1.9908396946564885e-05,
      "loss": 0.0537,
      "step": 2628
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.987786259541985e-05,
      "loss": 0.0507,
      "step": 2629
    },
    {
      "epoch": 16.04,
      "learning_rate": 1.984732824427481e-05,
      "loss": 0.0468,
      "step": 2630
    },
    {
      "epoch": 16.04,
      "learning_rate": 1.9816793893129772e-05,
      "loss": 0.0517,
      "step": 2631
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.9786259541984733e-05,
      "loss": 0.0471,
      "step": 2632
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.9755725190839697e-05,
      "loss": 0.0524,
      "step": 2633
    },
    {
      "epoch": 16.06,
      "learning_rate": 1.972519083969466e-05,
      "loss": 0.0498,
      "step": 2634
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.969465648854962e-05,
      "loss": 0.0691,
      "step": 2635
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.966412213740458e-05,
      "loss": 0.0459,
      "step": 2636
    },
    {
      "epoch": 16.08,
      "learning_rate": 1.9633587786259545e-05,
      "loss": 0.0574,
      "step": 2637
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.9603053435114503e-05,
      "loss": 0.0537,
      "step": 2638
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.9572519083969467e-05,
      "loss": 0.0368,
      "step": 2639
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.954198473282443e-05,
      "loss": 0.057,
      "step": 2640
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.9511450381679393e-05,
      "loss": 0.0539,
      "step": 2641
    },
    {
      "epoch": 16.11,
      "learning_rate": 1.948091603053435e-05,
      "loss": 0.0513,
      "step": 2642
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.9450381679389315e-05,
      "loss": 0.0578,
      "step": 2643
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.9419847328244276e-05,
      "loss": 0.0475,
      "step": 2644
    },
    {
      "epoch": 16.13,
      "learning_rate": 1.9389312977099238e-05,
      "loss": 0.0685,
      "step": 2645
    },
    {
      "epoch": 16.13,
      "learning_rate": 1.93587786259542e-05,
      "loss": 0.0585,
      "step": 2646
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.9328244274809163e-05,
      "loss": 0.0381,
      "step": 2647
    },
    {
      "epoch": 16.15,
      "learning_rate": 1.9297709923664124e-05,
      "loss": 0.0607,
      "step": 2648
    },
    {
      "epoch": 16.15,
      "learning_rate": 1.9267175572519085e-05,
      "loss": 0.0512,
      "step": 2649
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.9236641221374046e-05,
      "loss": 0.056,
      "step": 2650
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.920610687022901e-05,
      "loss": 0.058,
      "step": 2651
    },
    {
      "epoch": 16.17,
      "learning_rate": 1.917557251908397e-05,
      "loss": 0.0534,
      "step": 2652
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.9145038167938933e-05,
      "loss": 0.0542,
      "step": 2653
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.9114503816793894e-05,
      "loss": 0.0658,
      "step": 2654
    },
    {
      "epoch": 16.19,
      "learning_rate": 1.9083969465648855e-05,
      "loss": 0.0539,
      "step": 2655
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.9053435114503817e-05,
      "loss": 0.0469,
      "step": 2656
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.902290076335878e-05,
      "loss": 0.0658,
      "step": 2657
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.8992366412213742e-05,
      "loss": 0.0585,
      "step": 2658
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.8961832061068703e-05,
      "loss": 0.0503,
      "step": 2659
    },
    {
      "epoch": 16.22,
      "learning_rate": 1.8931297709923664e-05,
      "loss": 0.0487,
      "step": 2660
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.890076335877863e-05,
      "loss": 0.0503,
      "step": 2661
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.887022900763359e-05,
      "loss": 0.0588,
      "step": 2662
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.883969465648855e-05,
      "loss": 0.0529,
      "step": 2663
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.8809160305343512e-05,
      "loss": 0.0691,
      "step": 2664
    },
    {
      "epoch": 16.25,
      "learning_rate": 1.8778625954198477e-05,
      "loss": 0.0508,
      "step": 2665
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.8748091603053434e-05,
      "loss": 0.0594,
      "step": 2666
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.87175572519084e-05,
      "loss": 0.0563,
      "step": 2667
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.868702290076336e-05,
      "loss": 0.05,
      "step": 2668
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.865648854961832e-05,
      "loss": 0.0475,
      "step": 2669
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.8625954198473282e-05,
      "loss": 0.0658,
      "step": 2670
    },
    {
      "epoch": 16.29,
      "learning_rate": 1.8595419847328247e-05,
      "loss": 0.054,
      "step": 2671
    },
    {
      "epoch": 16.29,
      "learning_rate": 1.8564885496183208e-05,
      "loss": 0.0511,
      "step": 2672
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.853435114503817e-05,
      "loss": 0.0645,
      "step": 2673
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.850381679389313e-05,
      "loss": 0.0524,
      "step": 2674
    },
    {
      "epoch": 16.31,
      "learning_rate": 1.8473282442748095e-05,
      "loss": 0.0548,
      "step": 2675
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.8442748091603056e-05,
      "loss": 0.0538,
      "step": 2676
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.8412213740458017e-05,
      "loss": 0.056,
      "step": 2677
    },
    {
      "epoch": 16.33,
      "learning_rate": 1.8381679389312978e-05,
      "loss": 0.057,
      "step": 2678
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.8351145038167943e-05,
      "loss": 0.0584,
      "step": 2679
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.83206106870229e-05,
      "loss": 0.0574,
      "step": 2680
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.8290076335877865e-05,
      "loss": 0.0696,
      "step": 2681
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.8259541984732826e-05,
      "loss": 0.047,
      "step": 2682
    },
    {
      "epoch": 16.36,
      "learning_rate": 1.8229007633587787e-05,
      "loss": 0.06,
      "step": 2683
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.8198473282442748e-05,
      "loss": 0.052,
      "step": 2684
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.816793893129771e-05,
      "loss": 0.05,
      "step": 2685
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.8137404580152674e-05,
      "loss": 0.0465,
      "step": 2686
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.810687022900763e-05,
      "loss": 0.0537,
      "step": 2687
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.8076335877862596e-05,
      "loss": 0.0635,
      "step": 2688
    },
    {
      "epoch": 16.4,
      "learning_rate": 1.8045801526717557e-05,
      "loss": 0.053,
      "step": 2689
    },
    {
      "epoch": 16.4,
      "learning_rate": 1.801526717557252e-05,
      "loss": 0.0595,
      "step": 2690
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.798473282442748e-05,
      "loss": 0.0551,
      "step": 2691
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.7954198473282444e-05,
      "loss": 0.0518,
      "step": 2692
    },
    {
      "epoch": 16.42,
      "learning_rate": 1.7923664122137405e-05,
      "loss": 0.0548,
      "step": 2693
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.7893129770992366e-05,
      "loss": 0.0512,
      "step": 2694
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.7862595419847327e-05,
      "loss": 0.0477,
      "step": 2695
    },
    {
      "epoch": 16.44,
      "learning_rate": 1.783206106870229e-05,
      "loss": 0.0516,
      "step": 2696
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.7801526717557253e-05,
      "loss": 0.0551,
      "step": 2697
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.7770992366412214e-05,
      "loss": 0.0644,
      "step": 2698
    },
    {
      "epoch": 16.46,
      "learning_rate": 1.7740458015267175e-05,
      "loss": 0.06,
      "step": 2699
    },
    {
      "epoch": 16.46,
      "learning_rate": 1.770992366412214e-05,
      "loss": 0.0533,
      "step": 2700
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.7679389312977097e-05,
      "loss": 0.0587,
      "step": 2701
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.7648854961832062e-05,
      "loss": 0.067,
      "step": 2702
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.7618320610687023e-05,
      "loss": 0.0661,
      "step": 2703
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.7587786259541984e-05,
      "loss": 0.0558,
      "step": 2704
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.7557251908396945e-05,
      "loss": 0.0657,
      "step": 2705
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.752671755725191e-05,
      "loss": 0.0507,
      "step": 2706
    },
    {
      "epoch": 16.51,
      "learning_rate": 1.749618320610687e-05,
      "loss": 0.0525,
      "step": 2707
    },
    {
      "epoch": 16.51,
      "learning_rate": 1.7465648854961832e-05,
      "loss": 0.0537,
      "step": 2708
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.7435114503816793e-05,
      "loss": 0.0694,
      "step": 2709
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.7404580152671757e-05,
      "loss": 0.0533,
      "step": 2710
    },
    {
      "epoch": 16.53,
      "learning_rate": 1.737404580152672e-05,
      "loss": 0.0604,
      "step": 2711
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.734351145038168e-05,
      "loss": 0.0444,
      "step": 2712
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.731297709923664e-05,
      "loss": 0.057,
      "step": 2713
    },
    {
      "epoch": 16.55,
      "learning_rate": 1.7282442748091605e-05,
      "loss": 0.0543,
      "step": 2714
    },
    {
      "epoch": 16.55,
      "learning_rate": 1.7251908396946563e-05,
      "loss": 0.0676,
      "step": 2715
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.7221374045801528e-05,
      "loss": 0.0596,
      "step": 2716
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.719083969465649e-05,
      "loss": 0.0598,
      "step": 2717
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.716030534351145e-05,
      "loss": 0.0548,
      "step": 2718
    },
    {
      "epoch": 16.58,
      "learning_rate": 1.712977099236641e-05,
      "loss": 0.0598,
      "step": 2719
    },
    {
      "epoch": 16.59,
      "learning_rate": 1.7099236641221375e-05,
      "loss": 0.0504,
      "step": 2720
    },
    {
      "epoch": 16.59,
      "learning_rate": 1.7068702290076337e-05,
      "loss": 0.0479,
      "step": 2721
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.7038167938931298e-05,
      "loss": 0.0578,
      "step": 2722
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.700763358778626e-05,
      "loss": 0.0487,
      "step": 2723
    },
    {
      "epoch": 16.61,
      "learning_rate": 1.6977099236641223e-05,
      "loss": 0.0655,
      "step": 2724
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.6946564885496184e-05,
      "loss": 0.0601,
      "step": 2725
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.6916030534351145e-05,
      "loss": 0.054,
      "step": 2726
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.6885496183206107e-05,
      "loss": 0.0601,
      "step": 2727
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.685496183206107e-05,
      "loss": 0.059,
      "step": 2728
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.682442748091603e-05,
      "loss": 0.0584,
      "step": 2729
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.6793893129770993e-05,
      "loss": 0.0526,
      "step": 2730
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.6763358778625954e-05,
      "loss": 0.0564,
      "step": 2731
    },
    {
      "epoch": 16.66,
      "learning_rate": 1.6732824427480916e-05,
      "loss": 0.0671,
      "step": 2732
    },
    {
      "epoch": 16.66,
      "learning_rate": 1.6702290076335877e-05,
      "loss": 0.0655,
      "step": 2733
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.667175572519084e-05,
      "loss": 0.0453,
      "step": 2734
    },
    {
      "epoch": 16.68,
      "learning_rate": 1.6641221374045802e-05,
      "loss": 0.0732,
      "step": 2735
    },
    {
      "epoch": 16.68,
      "learning_rate": 1.6610687022900763e-05,
      "loss": 0.0648,
      "step": 2736
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.6580152671755725e-05,
      "loss": 0.0589,
      "step": 2737
    },
    {
      "epoch": 16.7,
      "learning_rate": 1.654961832061069e-05,
      "loss": 0.0625,
      "step": 2738
    },
    {
      "epoch": 16.7,
      "learning_rate": 1.651908396946565e-05,
      "loss": 0.062,
      "step": 2739
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.648854961832061e-05,
      "loss": 0.0473,
      "step": 2740
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.6458015267175572e-05,
      "loss": 0.0563,
      "step": 2741
    },
    {
      "epoch": 16.72,
      "learning_rate": 1.6427480916030537e-05,
      "loss": 0.0598,
      "step": 2742
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.6396946564885495e-05,
      "loss": 0.0481,
      "step": 2743
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.636641221374046e-05,
      "loss": 0.0515,
      "step": 2744
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.633587786259542e-05,
      "loss": 0.0514,
      "step": 2745
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.630534351145038e-05,
      "loss": 0.0594,
      "step": 2746
    },
    {
      "epoch": 16.75,
      "learning_rate": 1.6274809160305342e-05,
      "loss": 0.0566,
      "step": 2747
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.6244274809160307e-05,
      "loss": 0.0513,
      "step": 2748
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.6213740458015268e-05,
      "loss": 0.0566,
      "step": 2749
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.618320610687023e-05,
      "loss": 0.0491,
      "step": 2750
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.615267175572519e-05,
      "loss": 0.0471,
      "step": 2751
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.6122137404580155e-05,
      "loss": 0.0533,
      "step": 2752
    },
    {
      "epoch": 16.79,
      "learning_rate": 1.6091603053435113e-05,
      "loss": 0.066,
      "step": 2753
    },
    {
      "epoch": 16.79,
      "learning_rate": 1.6061068702290077e-05,
      "loss": 0.0627,
      "step": 2754
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.6030534351145038e-05,
      "loss": 0.0606,
      "step": 2755
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0643,
      "step": 2756
    },
    {
      "epoch": 16.81,
      "learning_rate": 1.596946564885496e-05,
      "loss": 0.0563,
      "step": 2757
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.5938931297709925e-05,
      "loss": 0.0664,
      "step": 2758
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.5908396946564886e-05,
      "loss": 0.0512,
      "step": 2759
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.5877862595419847e-05,
      "loss": 0.0535,
      "step": 2760
    },
    {
      "epoch": 16.84,
      "learning_rate": 1.5847328244274808e-05,
      "loss": 0.0546,
      "step": 2761
    },
    {
      "epoch": 16.84,
      "learning_rate": 1.5816793893129773e-05,
      "loss": 0.0673,
      "step": 2762
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.5786259541984734e-05,
      "loss": 0.0515,
      "step": 2763
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.5755725190839695e-05,
      "loss": 0.0616,
      "step": 2764
    },
    {
      "epoch": 16.86,
      "learning_rate": 1.5725190839694656e-05,
      "loss": 0.0612,
      "step": 2765
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.569465648854962e-05,
      "loss": 0.0518,
      "step": 2766
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.566412213740458e-05,
      "loss": 0.058,
      "step": 2767
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.5633587786259543e-05,
      "loss": 0.06,
      "step": 2768
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.5603053435114504e-05,
      "loss": 0.0642,
      "step": 2769
    },
    {
      "epoch": 16.89,
      "learning_rate": 1.557251908396947e-05,
      "loss": 0.0621,
      "step": 2770
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.5541984732824426e-05,
      "loss": 0.064,
      "step": 2771
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.551145038167939e-05,
      "loss": 0.0643,
      "step": 2772
    },
    {
      "epoch": 16.91,
      "learning_rate": 1.5480916030534352e-05,
      "loss": 0.0539,
      "step": 2773
    },
    {
      "epoch": 16.91,
      "learning_rate": 1.5450381679389313e-05,
      "loss": 0.0537,
      "step": 2774
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.5419847328244274e-05,
      "loss": 0.055,
      "step": 2775
    },
    {
      "epoch": 16.93,
      "learning_rate": 1.538931297709924e-05,
      "loss": 0.0687,
      "step": 2776
    },
    {
      "epoch": 16.93,
      "learning_rate": 1.53587786259542e-05,
      "loss": 0.0686,
      "step": 2777
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.532824427480916e-05,
      "loss": 0.063,
      "step": 2778
    },
    {
      "epoch": 16.95,
      "learning_rate": 1.5297709923664122e-05,
      "loss": 0.0583,
      "step": 2779
    },
    {
      "epoch": 16.95,
      "learning_rate": 1.5267175572519086e-05,
      "loss": 0.0607,
      "step": 2780
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.5236641221374046e-05,
      "loss": 0.0661,
      "step": 2781
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.5206106870229009e-05,
      "loss": 0.0587,
      "step": 2782
    },
    {
      "epoch": 16.97,
      "learning_rate": 1.517557251908397e-05,
      "loss": 0.0608,
      "step": 2783
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.5145038167938933e-05,
      "loss": 0.0579,
      "step": 2784
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.5114503816793894e-05,
      "loss": 0.0539,
      "step": 2785
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.5083969465648856e-05,
      "loss": 0.0539,
      "step": 2786
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.5053435114503816e-05,
      "loss": 0.062,
      "step": 2787
    },
    {
      "epoch": 17.0,
      "learning_rate": 1.502290076335878e-05,
      "loss": 0.0881,
      "step": 2788
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.499236641221374e-05,
      "loss": 0.0561,
      "step": 2789
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.4961832061068704e-05,
      "loss": 0.049,
      "step": 2790
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.4931297709923664e-05,
      "loss": 0.0557,
      "step": 2791
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.4900763358778627e-05,
      "loss": 0.0525,
      "step": 2792
    },
    {
      "epoch": 17.03,
      "learning_rate": 1.4870229007633588e-05,
      "loss": 0.0666,
      "step": 2793
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.483969465648855e-05,
      "loss": 0.0494,
      "step": 2794
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.4809160305343512e-05,
      "loss": 0.0427,
      "step": 2795
    },
    {
      "epoch": 17.05,
      "learning_rate": 1.4778625954198474e-05,
      "loss": 0.0472,
      "step": 2796
    },
    {
      "epoch": 17.05,
      "learning_rate": 1.4748091603053436e-05,
      "loss": 0.0513,
      "step": 2797
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.4717557251908398e-05,
      "loss": 0.0481,
      "step": 2798
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.468702290076336e-05,
      "loss": 0.0577,
      "step": 2799
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.4656488549618322e-05,
      "loss": 0.0491,
      "step": 2800
    },
    {
      "epoch": 17.08,
      "learning_rate": 1.4625954198473282e-05,
      "loss": 0.0565,
      "step": 2801
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.4595419847328246e-05,
      "loss": 0.0441,
      "step": 2802
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.4564885496183206e-05,
      "loss": 0.058,
      "step": 2803
    },
    {
      "epoch": 17.1,
      "learning_rate": 1.4534351145038168e-05,
      "loss": 0.0569,
      "step": 2804
    },
    {
      "epoch": 17.1,
      "learning_rate": 1.450381679389313e-05,
      "loss": 0.057,
      "step": 2805
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.4473282442748092e-05,
      "loss": 0.0567,
      "step": 2806
    },
    {
      "epoch": 17.12,
      "learning_rate": 1.4442748091603053e-05,
      "loss": 0.051,
      "step": 2807
    },
    {
      "epoch": 17.12,
      "learning_rate": 1.4412213740458016e-05,
      "loss": 0.056,
      "step": 2808
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.4381679389312977e-05,
      "loss": 0.0594,
      "step": 2809
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.435114503816794e-05,
      "loss": 0.0603,
      "step": 2810
    },
    {
      "epoch": 17.14,
      "learning_rate": 1.4320610687022901e-05,
      "loss": 0.0531,
      "step": 2811
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.4290076335877864e-05,
      "loss": 0.0594,
      "step": 2812
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.4259541984732824e-05,
      "loss": 0.0574,
      "step": 2813
    },
    {
      "epoch": 17.16,
      "learning_rate": 1.4229007633587788e-05,
      "loss": 0.0492,
      "step": 2814
    },
    {
      "epoch": 17.16,
      "learning_rate": 1.4198473282442747e-05,
      "loss": 0.0564,
      "step": 2815
    },
    {
      "epoch": 17.17,
      "learning_rate": 1.4167938931297712e-05,
      "loss": 0.0479,
      "step": 2816
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.4137404580152671e-05,
      "loss": 0.0626,
      "step": 2817
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.4106870229007634e-05,
      "loss": 0.0532,
      "step": 2818
    },
    {
      "epoch": 17.19,
      "learning_rate": 1.4076335877862595e-05,
      "loss": 0.0621,
      "step": 2819
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.4045801526717558e-05,
      "loss": 0.0599,
      "step": 2820
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.401526717557252e-05,
      "loss": 0.0443,
      "step": 2821
    },
    {
      "epoch": 17.21,
      "learning_rate": 1.3984732824427482e-05,
      "loss": 0.0618,
      "step": 2822
    },
    {
      "epoch": 17.21,
      "learning_rate": 1.3954198473282443e-05,
      "loss": 0.0615,
      "step": 2823
    },
    {
      "epoch": 17.22,
      "learning_rate": 1.3923664122137406e-05,
      "loss": 0.0659,
      "step": 2824
    },
    {
      "epoch": 17.23,
      "learning_rate": 1.3893129770992367e-05,
      "loss": 0.0604,
      "step": 2825
    },
    {
      "epoch": 17.23,
      "learning_rate": 1.386259541984733e-05,
      "loss": 0.0641,
      "step": 2826
    },
    {
      "epoch": 17.24,
      "learning_rate": 1.383206106870229e-05,
      "loss": 0.0518,
      "step": 2827
    },
    {
      "epoch": 17.24,
      "learning_rate": 1.3801526717557254e-05,
      "loss": 0.0561,
      "step": 2828
    },
    {
      "epoch": 17.25,
      "learning_rate": 1.3770992366412213e-05,
      "loss": 0.0566,
      "step": 2829
    },
    {
      "epoch": 17.26,
      "learning_rate": 1.3740458015267178e-05,
      "loss": 0.051,
      "step": 2830
    },
    {
      "epoch": 17.26,
      "learning_rate": 1.3709923664122137e-05,
      "loss": 0.0526,
      "step": 2831
    },
    {
      "epoch": 17.27,
      "learning_rate": 1.36793893129771e-05,
      "loss": 0.0564,
      "step": 2832
    },
    {
      "epoch": 17.27,
      "learning_rate": 1.3648854961832061e-05,
      "loss": 0.0601,
      "step": 2833
    },
    {
      "epoch": 17.28,
      "learning_rate": 1.3618320610687024e-05,
      "loss": 0.0485,
      "step": 2834
    },
    {
      "epoch": 17.29,
      "learning_rate": 1.3587786259541985e-05,
      "loss": 0.0563,
      "step": 2835
    },
    {
      "epoch": 17.29,
      "learning_rate": 1.3557251908396948e-05,
      "loss": 0.0541,
      "step": 2836
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.3526717557251909e-05,
      "loss": 0.0646,
      "step": 2837
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.3496183206106872e-05,
      "loss": 0.0597,
      "step": 2838
    },
    {
      "epoch": 17.31,
      "learning_rate": 1.3465648854961833e-05,
      "loss": 0.0548,
      "step": 2839
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.3435114503816796e-05,
      "loss": 0.059,
      "step": 2840
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.3404580152671755e-05,
      "loss": 0.05,
      "step": 2841
    },
    {
      "epoch": 17.33,
      "learning_rate": 1.337404580152672e-05,
      "loss": 0.0687,
      "step": 2842
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.3343511450381679e-05,
      "loss": 0.0591,
      "step": 2843
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.3312977099236642e-05,
      "loss": 0.0557,
      "step": 2844
    },
    {
      "epoch": 17.35,
      "learning_rate": 1.3282442748091603e-05,
      "loss": 0.0625,
      "step": 2845
    },
    {
      "epoch": 17.35,
      "learning_rate": 1.3251908396946566e-05,
      "loss": 0.0558,
      "step": 2846
    },
    {
      "epoch": 17.36,
      "learning_rate": 1.3221374045801527e-05,
      "loss": 0.064,
      "step": 2847
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.319083969465649e-05,
      "loss": 0.0441,
      "step": 2848
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.316030534351145e-05,
      "loss": 0.071,
      "step": 2849
    },
    {
      "epoch": 17.38,
      "learning_rate": 1.3129770992366414e-05,
      "loss": 0.0614,
      "step": 2850
    },
    {
      "epoch": 17.38,
      "learning_rate": 1.3099236641221375e-05,
      "loss": 0.0474,
      "step": 2851
    },
    {
      "epoch": 17.39,
      "learning_rate": 1.3068702290076338e-05,
      "loss": 0.0588,
      "step": 2852
    },
    {
      "epoch": 17.4,
      "learning_rate": 1.3038167938931297e-05,
      "loss": 0.0521,
      "step": 2853
    },
    {
      "epoch": 17.4,
      "learning_rate": 1.3007633587786261e-05,
      "loss": 0.0555,
      "step": 2854
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.2977099236641221e-05,
      "loss": 0.0545,
      "step": 2855
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.2946564885496185e-05,
      "loss": 0.0438,
      "step": 2856
    },
    {
      "epoch": 17.42,
      "learning_rate": 1.2916030534351145e-05,
      "loss": 0.0486,
      "step": 2857
    },
    {
      "epoch": 17.43,
      "learning_rate": 1.2885496183206108e-05,
      "loss": 0.0569,
      "step": 2858
    },
    {
      "epoch": 17.43,
      "learning_rate": 1.2854961832061069e-05,
      "loss": 0.0607,
      "step": 2859
    },
    {
      "epoch": 17.44,
      "learning_rate": 1.2824427480916032e-05,
      "loss": 0.0622,
      "step": 2860
    },
    {
      "epoch": 17.45,
      "learning_rate": 1.2793893129770993e-05,
      "loss": 0.0549,
      "step": 2861
    },
    {
      "epoch": 17.45,
      "learning_rate": 1.2763358778625955e-05,
      "loss": 0.0594,
      "step": 2862
    },
    {
      "epoch": 17.46,
      "learning_rate": 1.2732824427480917e-05,
      "loss": 0.0528,
      "step": 2863
    },
    {
      "epoch": 17.46,
      "learning_rate": 1.270229007633588e-05,
      "loss": 0.0562,
      "step": 2864
    },
    {
      "epoch": 17.47,
      "learning_rate": 1.267175572519084e-05,
      "loss": 0.058,
      "step": 2865
    },
    {
      "epoch": 17.48,
      "learning_rate": 1.2641221374045803e-05,
      "loss": 0.0507,
      "step": 2866
    },
    {
      "epoch": 17.48,
      "learning_rate": 1.2610687022900763e-05,
      "loss": 0.0568,
      "step": 2867
    },
    {
      "epoch": 17.49,
      "learning_rate": 1.2580152671755727e-05,
      "loss": 0.0516,
      "step": 2868
    },
    {
      "epoch": 17.49,
      "learning_rate": 1.2549618320610687e-05,
      "loss": 0.0482,
      "step": 2869
    },
    {
      "epoch": 17.5,
      "learning_rate": 1.251908396946565e-05,
      "loss": 0.0587,
      "step": 2870
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.248854961832061e-05,
      "loss": 0.0654,
      "step": 2871
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.2458015267175573e-05,
      "loss": 0.0627,
      "step": 2872
    },
    {
      "epoch": 17.52,
      "learning_rate": 1.2427480916030535e-05,
      "loss": 0.0603,
      "step": 2873
    },
    {
      "epoch": 17.52,
      "learning_rate": 1.2396946564885496e-05,
      "loss": 0.0496,
      "step": 2874
    },
    {
      "epoch": 17.53,
      "learning_rate": 1.2366412213740458e-05,
      "loss": 0.0497,
      "step": 2875
    },
    {
      "epoch": 17.54,
      "learning_rate": 1.233587786259542e-05,
      "loss": 0.0698,
      "step": 2876
    },
    {
      "epoch": 17.54,
      "learning_rate": 1.2305343511450382e-05,
      "loss": 0.055,
      "step": 2877
    },
    {
      "epoch": 17.55,
      "learning_rate": 1.2274809160305343e-05,
      "loss": 0.0619,
      "step": 2878
    },
    {
      "epoch": 17.55,
      "learning_rate": 1.2244274809160305e-05,
      "loss": 0.0524,
      "step": 2879
    },
    {
      "epoch": 17.56,
      "learning_rate": 1.2213740458015267e-05,
      "loss": 0.0626,
      "step": 2880
    },
    {
      "epoch": 17.57,
      "learning_rate": 1.2183206106870229e-05,
      "loss": 0.0637,
      "step": 2881
    },
    {
      "epoch": 17.57,
      "learning_rate": 1.2152671755725191e-05,
      "loss": 0.0602,
      "step": 2882
    },
    {
      "epoch": 17.58,
      "learning_rate": 1.2122137404580152e-05,
      "loss": 0.0563,
      "step": 2883
    },
    {
      "epoch": 17.59,
      "learning_rate": 1.2091603053435115e-05,
      "loss": 0.0571,
      "step": 2884
    },
    {
      "epoch": 17.59,
      "learning_rate": 1.2061068702290076e-05,
      "loss": 0.059,
      "step": 2885
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.2030534351145037e-05,
      "loss": 0.059,
      "step": 2886
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.2e-05,
      "loss": 0.0496,
      "step": 2887
    },
    {
      "epoch": 17.61,
      "learning_rate": 1.1969465648854961e-05,
      "loss": 0.0625,
      "step": 2888
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.1938931297709924e-05,
      "loss": 0.0515,
      "step": 2889
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.1908396946564885e-05,
      "loss": 0.0491,
      "step": 2890
    },
    {
      "epoch": 17.63,
      "learning_rate": 1.1877862595419848e-05,
      "loss": 0.055,
      "step": 2891
    },
    {
      "epoch": 17.63,
      "learning_rate": 1.184732824427481e-05,
      "loss": 0.0705,
      "step": 2892
    },
    {
      "epoch": 17.64,
      "learning_rate": 1.181679389312977e-05,
      "loss": 0.059,
      "step": 2893
    },
    {
      "epoch": 17.65,
      "learning_rate": 1.1786259541984733e-05,
      "loss": 0.0567,
      "step": 2894
    },
    {
      "epoch": 17.65,
      "learning_rate": 1.1755725190839694e-05,
      "loss": 0.0661,
      "step": 2895
    },
    {
      "epoch": 17.66,
      "learning_rate": 1.1725190839694657e-05,
      "loss": 0.0515,
      "step": 2896
    },
    {
      "epoch": 17.66,
      "learning_rate": 1.1694656488549618e-05,
      "loss": 0.0564,
      "step": 2897
    },
    {
      "epoch": 17.67,
      "learning_rate": 1.1664122137404581e-05,
      "loss": 0.071,
      "step": 2898
    },
    {
      "epoch": 17.68,
      "learning_rate": 1.1633587786259542e-05,
      "loss": 0.0524,
      "step": 2899
    },
    {
      "epoch": 17.68,
      "learning_rate": 1.1603053435114503e-05,
      "loss": 0.0541,
      "step": 2900
    },
    {
      "epoch": 17.69,
      "learning_rate": 1.1572519083969466e-05,
      "loss": 0.0527,
      "step": 2901
    },
    {
      "epoch": 17.7,
      "learning_rate": 1.1541984732824427e-05,
      "loss": 0.047,
      "step": 2902
    },
    {
      "epoch": 17.7,
      "learning_rate": 1.151145038167939e-05,
      "loss": 0.0602,
      "step": 2903
    },
    {
      "epoch": 17.71,
      "learning_rate": 1.1480916030534351e-05,
      "loss": 0.0625,
      "step": 2904
    },
    {
      "epoch": 17.71,
      "learning_rate": 1.1450381679389314e-05,
      "loss": 0.0438,
      "step": 2905
    },
    {
      "epoch": 17.72,
      "learning_rate": 1.1419847328244275e-05,
      "loss": 0.0509,
      "step": 2906
    },
    {
      "epoch": 17.73,
      "learning_rate": 1.1389312977099236e-05,
      "loss": 0.0421,
      "step": 2907
    },
    {
      "epoch": 17.73,
      "learning_rate": 1.1358778625954199e-05,
      "loss": 0.0464,
      "step": 2908
    },
    {
      "epoch": 17.74,
      "learning_rate": 1.132824427480916e-05,
      "loss": 0.0594,
      "step": 2909
    },
    {
      "epoch": 17.74,
      "learning_rate": 1.1297709923664123e-05,
      "loss": 0.0526,
      "step": 2910
    },
    {
      "epoch": 17.75,
      "learning_rate": 1.1267175572519084e-05,
      "loss": 0.0618,
      "step": 2911
    },
    {
      "epoch": 17.76,
      "learning_rate": 1.1236641221374047e-05,
      "loss": 0.0549,
      "step": 2912
    },
    {
      "epoch": 17.76,
      "learning_rate": 1.1206106870229008e-05,
      "loss": 0.0632,
      "step": 2913
    },
    {
      "epoch": 17.77,
      "learning_rate": 1.1175572519083969e-05,
      "loss": 0.0484,
      "step": 2914
    },
    {
      "epoch": 17.77,
      "learning_rate": 1.1145038167938932e-05,
      "loss": 0.0564,
      "step": 2915
    },
    {
      "epoch": 17.78,
      "learning_rate": 1.1114503816793893e-05,
      "loss": 0.0526,
      "step": 2916
    },
    {
      "epoch": 17.79,
      "learning_rate": 1.1083969465648856e-05,
      "loss": 0.0655,
      "step": 2917
    },
    {
      "epoch": 17.79,
      "learning_rate": 1.1053435114503817e-05,
      "loss": 0.0516,
      "step": 2918
    },
    {
      "epoch": 17.8,
      "learning_rate": 1.1022900763358778e-05,
      "loss": 0.0609,
      "step": 2919
    },
    {
      "epoch": 17.8,
      "learning_rate": 1.099236641221374e-05,
      "loss": 0.0611,
      "step": 2920
    },
    {
      "epoch": 17.81,
      "learning_rate": 1.0961832061068702e-05,
      "loss": 0.0492,
      "step": 2921
    },
    {
      "epoch": 17.82,
      "learning_rate": 1.0931297709923665e-05,
      "loss": 0.0489,
      "step": 2922
    },
    {
      "epoch": 17.82,
      "learning_rate": 1.0900763358778626e-05,
      "loss": 0.0699,
      "step": 2923
    },
    {
      "epoch": 17.83,
      "learning_rate": 1.0870229007633589e-05,
      "loss": 0.0408,
      "step": 2924
    },
    {
      "epoch": 17.84,
      "learning_rate": 1.083969465648855e-05,
      "loss": 0.0424,
      "step": 2925
    },
    {
      "epoch": 17.84,
      "learning_rate": 1.0809160305343511e-05,
      "loss": 0.0641,
      "step": 2926
    },
    {
      "epoch": 17.85,
      "learning_rate": 1.0778625954198474e-05,
      "loss": 0.0673,
      "step": 2927
    },
    {
      "epoch": 17.85,
      "learning_rate": 1.0748091603053435e-05,
      "loss": 0.0645,
      "step": 2928
    },
    {
      "epoch": 17.86,
      "learning_rate": 1.0717557251908398e-05,
      "loss": 0.0523,
      "step": 2929
    },
    {
      "epoch": 17.87,
      "learning_rate": 1.0687022900763359e-05,
      "loss": 0.0606,
      "step": 2930
    },
    {
      "epoch": 17.87,
      "learning_rate": 1.0656488549618322e-05,
      "loss": 0.0571,
      "step": 2931
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.0625954198473283e-05,
      "loss": 0.0526,
      "step": 2932
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.0595419847328244e-05,
      "loss": 0.0545,
      "step": 2933
    },
    {
      "epoch": 17.89,
      "learning_rate": 1.0564885496183207e-05,
      "loss": 0.0604,
      "step": 2934
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.0534351145038168e-05,
      "loss": 0.0599,
      "step": 2935
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.050381679389313e-05,
      "loss": 0.06,
      "step": 2936
    },
    {
      "epoch": 17.91,
      "learning_rate": 1.0473282442748092e-05,
      "loss": 0.0476,
      "step": 2937
    },
    {
      "epoch": 17.91,
      "learning_rate": 1.0442748091603054e-05,
      "loss": 0.0688,
      "step": 2938
    },
    {
      "epoch": 17.92,
      "learning_rate": 1.0412213740458016e-05,
      "loss": 0.0523,
      "step": 2939
    },
    {
      "epoch": 17.93,
      "learning_rate": 1.0381679389312977e-05,
      "loss": 0.0493,
      "step": 2940
    },
    {
      "epoch": 17.93,
      "learning_rate": 1.035114503816794e-05,
      "loss": 0.0427,
      "step": 2941
    },
    {
      "epoch": 17.94,
      "learning_rate": 1.03206106870229e-05,
      "loss": 0.0635,
      "step": 2942
    },
    {
      "epoch": 17.95,
      "learning_rate": 1.0290076335877863e-05,
      "loss": 0.0553,
      "step": 2943
    },
    {
      "epoch": 17.95,
      "learning_rate": 1.0259541984732825e-05,
      "loss": 0.0553,
      "step": 2944
    },
    {
      "epoch": 17.96,
      "learning_rate": 1.0229007633587787e-05,
      "loss": 0.0472,
      "step": 2945
    },
    {
      "epoch": 17.96,
      "learning_rate": 1.0198473282442748e-05,
      "loss": 0.0439,
      "step": 2946
    },
    {
      "epoch": 17.97,
      "learning_rate": 1.016793893129771e-05,
      "loss": 0.0651,
      "step": 2947
    },
    {
      "epoch": 17.98,
      "learning_rate": 1.0137404580152672e-05,
      "loss": 0.0459,
      "step": 2948
    },
    {
      "epoch": 17.98,
      "learning_rate": 1.0106870229007634e-05,
      "loss": 0.0627,
      "step": 2949
    },
    {
      "epoch": 17.99,
      "learning_rate": 1.0076335877862596e-05,
      "loss": 0.059,
      "step": 2950
    },
    {
      "epoch": 17.99,
      "learning_rate": 1.0045801526717557e-05,
      "loss": 0.0708,
      "step": 2951
    },
    {
      "epoch": 18.0,
      "learning_rate": 1.001526717557252e-05,
      "loss": 0.0393,
      "step": 2952
    },
    {
      "epoch": 18.01,
      "learning_rate": 9.984732824427481e-06,
      "loss": 0.05,
      "step": 2953
    },
    {
      "epoch": 18.01,
      "learning_rate": 9.954198473282442e-06,
      "loss": 0.0542,
      "step": 2954
    },
    {
      "epoch": 18.02,
      "learning_rate": 9.923664122137405e-06,
      "loss": 0.0593,
      "step": 2955
    },
    {
      "epoch": 18.02,
      "learning_rate": 9.893129770992366e-06,
      "loss": 0.0621,
      "step": 2956
    },
    {
      "epoch": 18.03,
      "learning_rate": 9.86259541984733e-06,
      "loss": 0.0495,
      "step": 2957
    },
    {
      "epoch": 18.04,
      "learning_rate": 9.83206106870229e-06,
      "loss": 0.0508,
      "step": 2958
    },
    {
      "epoch": 18.04,
      "learning_rate": 9.801526717557251e-06,
      "loss": 0.0665,
      "step": 2959
    },
    {
      "epoch": 18.05,
      "learning_rate": 9.770992366412214e-06,
      "loss": 0.0542,
      "step": 2960
    },
    {
      "epoch": 18.05,
      "learning_rate": 9.740458015267175e-06,
      "loss": 0.0721,
      "step": 2961
    },
    {
      "epoch": 18.06,
      "learning_rate": 9.709923664122138e-06,
      "loss": 0.0555,
      "step": 2962
    },
    {
      "epoch": 18.07,
      "learning_rate": 9.6793893129771e-06,
      "loss": 0.0595,
      "step": 2963
    },
    {
      "epoch": 18.07,
      "learning_rate": 9.648854961832062e-06,
      "loss": 0.0498,
      "step": 2964
    },
    {
      "epoch": 18.08,
      "learning_rate": 9.618320610687023e-06,
      "loss": 0.0616,
      "step": 2965
    },
    {
      "epoch": 18.09,
      "learning_rate": 9.587786259541984e-06,
      "loss": 0.0586,
      "step": 2966
    },
    {
      "epoch": 18.09,
      "learning_rate": 9.557251908396947e-06,
      "loss": 0.057,
      "step": 2967
    },
    {
      "epoch": 18.1,
      "learning_rate": 9.526717557251908e-06,
      "loss": 0.0551,
      "step": 2968
    },
    {
      "epoch": 18.1,
      "learning_rate": 9.496183206106871e-06,
      "loss": 0.0458,
      "step": 2969
    },
    {
      "epoch": 18.11,
      "learning_rate": 9.465648854961832e-06,
      "loss": 0.048,
      "step": 2970
    },
    {
      "epoch": 18.12,
      "learning_rate": 9.435114503816795e-06,
      "loss": 0.0568,
      "step": 2971
    },
    {
      "epoch": 18.12,
      "learning_rate": 9.404580152671756e-06,
      "loss": 0.0508,
      "step": 2972
    },
    {
      "epoch": 18.13,
      "learning_rate": 9.374045801526717e-06,
      "loss": 0.0481,
      "step": 2973
    },
    {
      "epoch": 18.13,
      "learning_rate": 9.34351145038168e-06,
      "loss": 0.058,
      "step": 2974
    },
    {
      "epoch": 18.14,
      "learning_rate": 9.312977099236641e-06,
      "loss": 0.0667,
      "step": 2975
    },
    {
      "epoch": 18.15,
      "learning_rate": 9.282442748091604e-06,
      "loss": 0.0568,
      "step": 2976
    },
    {
      "epoch": 18.15,
      "learning_rate": 9.251908396946565e-06,
      "loss": 0.0585,
      "step": 2977
    },
    {
      "epoch": 18.16,
      "learning_rate": 9.221374045801528e-06,
      "loss": 0.0502,
      "step": 2978
    },
    {
      "epoch": 18.16,
      "learning_rate": 9.190839694656489e-06,
      "loss": 0.059,
      "step": 2979
    },
    {
      "epoch": 18.17,
      "learning_rate": 9.16030534351145e-06,
      "loss": 0.0615,
      "step": 2980
    },
    {
      "epoch": 18.18,
      "learning_rate": 9.129770992366413e-06,
      "loss": 0.0536,
      "step": 2981
    },
    {
      "epoch": 18.18,
      "learning_rate": 9.099236641221374e-06,
      "loss": 0.0499,
      "step": 2982
    },
    {
      "epoch": 18.19,
      "learning_rate": 9.068702290076337e-06,
      "loss": 0.0489,
      "step": 2983
    },
    {
      "epoch": 18.2,
      "learning_rate": 9.038167938931298e-06,
      "loss": 0.0525,
      "step": 2984
    },
    {
      "epoch": 18.2,
      "learning_rate": 9.00763358778626e-06,
      "loss": 0.0612,
      "step": 2985
    },
    {
      "epoch": 18.21,
      "learning_rate": 8.977099236641222e-06,
      "loss": 0.0503,
      "step": 2986
    },
    {
      "epoch": 18.21,
      "learning_rate": 8.946564885496183e-06,
      "loss": 0.0502,
      "step": 2987
    },
    {
      "epoch": 18.22,
      "learning_rate": 8.916030534351146e-06,
      "loss": 0.0501,
      "step": 2988
    },
    {
      "epoch": 18.23,
      "learning_rate": 8.885496183206107e-06,
      "loss": 0.0515,
      "step": 2989
    },
    {
      "epoch": 18.23,
      "learning_rate": 8.85496183206107e-06,
      "loss": 0.0456,
      "step": 2990
    },
    {
      "epoch": 18.24,
      "learning_rate": 8.824427480916031e-06,
      "loss": 0.0443,
      "step": 2991
    },
    {
      "epoch": 18.24,
      "learning_rate": 8.793893129770992e-06,
      "loss": 0.0508,
      "step": 2992
    },
    {
      "epoch": 18.25,
      "learning_rate": 8.763358778625955e-06,
      "loss": 0.0474,
      "step": 2993
    },
    {
      "epoch": 18.26,
      "learning_rate": 8.732824427480916e-06,
      "loss": 0.059,
      "step": 2994
    },
    {
      "epoch": 18.26,
      "learning_rate": 8.702290076335879e-06,
      "loss": 0.063,
      "step": 2995
    },
    {
      "epoch": 18.27,
      "learning_rate": 8.67175572519084e-06,
      "loss": 0.0503,
      "step": 2996
    },
    {
      "epoch": 18.27,
      "learning_rate": 8.641221374045803e-06,
      "loss": 0.0542,
      "step": 2997
    },
    {
      "epoch": 18.28,
      "learning_rate": 8.610687022900764e-06,
      "loss": 0.0591,
      "step": 2998
    },
    {
      "epoch": 18.29,
      "learning_rate": 8.580152671755725e-06,
      "loss": 0.0588,
      "step": 2999
    },
    {
      "epoch": 18.29,
      "learning_rate": 8.549618320610688e-06,
      "loss": 0.0514,
      "step": 3000
    },
    {
      "epoch": 18.3,
      "learning_rate": 8.519083969465649e-06,
      "loss": 0.0427,
      "step": 3001
    },
    {
      "epoch": 18.3,
      "learning_rate": 8.488549618320612e-06,
      "loss": 0.0671,
      "step": 3002
    },
    {
      "epoch": 18.31,
      "learning_rate": 8.458015267175573e-06,
      "loss": 0.0544,
      "step": 3003
    },
    {
      "epoch": 18.32,
      "learning_rate": 8.427480916030536e-06,
      "loss": 0.065,
      "step": 3004
    },
    {
      "epoch": 18.32,
      "learning_rate": 8.396946564885497e-06,
      "loss": 0.0621,
      "step": 3005
    },
    {
      "epoch": 18.33,
      "learning_rate": 8.366412213740458e-06,
      "loss": 0.0731,
      "step": 3006
    },
    {
      "epoch": 18.34,
      "learning_rate": 8.33587786259542e-06,
      "loss": 0.0498,
      "step": 3007
    },
    {
      "epoch": 18.34,
      "learning_rate": 8.305343511450382e-06,
      "loss": 0.0599,
      "step": 3008
    },
    {
      "epoch": 18.35,
      "learning_rate": 8.274809160305345e-06,
      "loss": 0.0559,
      "step": 3009
    },
    {
      "epoch": 18.35,
      "learning_rate": 8.244274809160306e-06,
      "loss": 0.0538,
      "step": 3010
    },
    {
      "epoch": 18.36,
      "learning_rate": 8.213740458015268e-06,
      "loss": 0.061,
      "step": 3011
    },
    {
      "epoch": 18.37,
      "learning_rate": 8.18320610687023e-06,
      "loss": 0.0542,
      "step": 3012
    },
    {
      "epoch": 18.37,
      "learning_rate": 8.15267175572519e-06,
      "loss": 0.0451,
      "step": 3013
    },
    {
      "epoch": 18.38,
      "learning_rate": 8.122137404580153e-06,
      "loss": 0.0509,
      "step": 3014
    },
    {
      "epoch": 18.38,
      "learning_rate": 8.091603053435115e-06,
      "loss": 0.0795,
      "step": 3015
    },
    {
      "epoch": 18.39,
      "learning_rate": 8.061068702290077e-06,
      "loss": 0.0478,
      "step": 3016
    },
    {
      "epoch": 18.4,
      "learning_rate": 8.030534351145039e-06,
      "loss": 0.0656,
      "step": 3017
    },
    {
      "epoch": 18.4,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0556,
      "step": 3018
    },
    {
      "epoch": 18.41,
      "learning_rate": 7.969465648854962e-06,
      "loss": 0.0534,
      "step": 3019
    },
    {
      "epoch": 18.41,
      "learning_rate": 7.938931297709924e-06,
      "loss": 0.0572,
      "step": 3020
    },
    {
      "epoch": 18.42,
      "learning_rate": 7.908396946564886e-06,
      "loss": 0.064,
      "step": 3021
    },
    {
      "epoch": 18.43,
      "learning_rate": 7.877862595419847e-06,
      "loss": 0.0602,
      "step": 3022
    },
    {
      "epoch": 18.43,
      "learning_rate": 7.84732824427481e-06,
      "loss": 0.0609,
      "step": 3023
    },
    {
      "epoch": 18.44,
      "learning_rate": 7.816793893129771e-06,
      "loss": 0.0613,
      "step": 3024
    },
    {
      "epoch": 18.45,
      "learning_rate": 7.786259541984734e-06,
      "loss": 0.0504,
      "step": 3025
    },
    {
      "epoch": 18.45,
      "learning_rate": 7.755725190839695e-06,
      "loss": 0.0558,
      "step": 3026
    },
    {
      "epoch": 18.46,
      "learning_rate": 7.725190839694656e-06,
      "loss": 0.0496,
      "step": 3027
    },
    {
      "epoch": 18.46,
      "learning_rate": 7.69465648854962e-06,
      "loss": 0.0705,
      "step": 3028
    },
    {
      "epoch": 18.47,
      "learning_rate": 7.66412213740458e-06,
      "loss": 0.0523,
      "step": 3029
    },
    {
      "epoch": 18.48,
      "learning_rate": 7.633587786259543e-06,
      "loss": 0.0575,
      "step": 3030
    },
    {
      "epoch": 18.48,
      "learning_rate": 7.603053435114504e-06,
      "loss": 0.0528,
      "step": 3031
    },
    {
      "epoch": 18.49,
      "learning_rate": 7.572519083969466e-06,
      "loss": 0.047,
      "step": 3032
    },
    {
      "epoch": 18.49,
      "learning_rate": 7.541984732824428e-06,
      "loss": 0.053,
      "step": 3033
    },
    {
      "epoch": 18.5,
      "learning_rate": 7.51145038167939e-06,
      "loss": 0.0583,
      "step": 3034
    },
    {
      "epoch": 18.51,
      "learning_rate": 7.480916030534352e-06,
      "loss": 0.0471,
      "step": 3035
    },
    {
      "epoch": 18.51,
      "learning_rate": 7.450381679389313e-06,
      "loss": 0.0593,
      "step": 3036
    },
    {
      "epoch": 18.52,
      "learning_rate": 7.419847328244275e-06,
      "loss": 0.0563,
      "step": 3037
    },
    {
      "epoch": 18.52,
      "learning_rate": 7.389312977099237e-06,
      "loss": 0.053,
      "step": 3038
    },
    {
      "epoch": 18.53,
      "learning_rate": 7.358778625954199e-06,
      "loss": 0.0566,
      "step": 3039
    },
    {
      "epoch": 18.54,
      "learning_rate": 7.328244274809161e-06,
      "loss": 0.05,
      "step": 3040
    },
    {
      "epoch": 18.54,
      "learning_rate": 7.297709923664123e-06,
      "loss": 0.0539,
      "step": 3041
    },
    {
      "epoch": 18.55,
      "learning_rate": 7.267175572519084e-06,
      "loss": 0.0516,
      "step": 3042
    },
    {
      "epoch": 18.55,
      "learning_rate": 7.236641221374046e-06,
      "loss": 0.07,
      "step": 3043
    },
    {
      "epoch": 18.56,
      "learning_rate": 7.206106870229008e-06,
      "loss": 0.0504,
      "step": 3044
    },
    {
      "epoch": 18.57,
      "learning_rate": 7.17557251908397e-06,
      "loss": 0.0514,
      "step": 3045
    },
    {
      "epoch": 18.57,
      "learning_rate": 7.145038167938932e-06,
      "loss": 0.0461,
      "step": 3046
    },
    {
      "epoch": 18.58,
      "learning_rate": 7.114503816793894e-06,
      "loss": 0.0532,
      "step": 3047
    },
    {
      "epoch": 18.59,
      "learning_rate": 7.083969465648856e-06,
      "loss": 0.0636,
      "step": 3048
    },
    {
      "epoch": 18.59,
      "learning_rate": 7.053435114503817e-06,
      "loss": 0.0593,
      "step": 3049
    },
    {
      "epoch": 18.6,
      "learning_rate": 7.022900763358779e-06,
      "loss": 0.07,
      "step": 3050
    },
    {
      "epoch": 18.6,
      "learning_rate": 6.992366412213741e-06,
      "loss": 0.0542,
      "step": 3051
    },
    {
      "epoch": 18.61,
      "learning_rate": 6.961832061068703e-06,
      "loss": 0.0545,
      "step": 3052
    },
    {
      "epoch": 18.62,
      "learning_rate": 6.931297709923665e-06,
      "loss": 0.0511,
      "step": 3053
    },
    {
      "epoch": 18.62,
      "learning_rate": 6.900763358778627e-06,
      "loss": 0.0497,
      "step": 3054
    },
    {
      "epoch": 18.63,
      "learning_rate": 6.870229007633589e-06,
      "loss": 0.0524,
      "step": 3055
    },
    {
      "epoch": 18.63,
      "learning_rate": 6.83969465648855e-06,
      "loss": 0.0576,
      "step": 3056
    },
    {
      "epoch": 18.64,
      "learning_rate": 6.809160305343512e-06,
      "loss": 0.058,
      "step": 3057
    },
    {
      "epoch": 18.65,
      "learning_rate": 6.778625954198474e-06,
      "loss": 0.0461,
      "step": 3058
    },
    {
      "epoch": 18.65,
      "learning_rate": 6.748091603053436e-06,
      "loss": 0.0499,
      "step": 3059
    },
    {
      "epoch": 18.66,
      "learning_rate": 6.717557251908398e-06,
      "loss": 0.0572,
      "step": 3060
    },
    {
      "epoch": 18.66,
      "learning_rate": 6.68702290076336e-06,
      "loss": 0.0683,
      "step": 3061
    },
    {
      "epoch": 18.67,
      "learning_rate": 6.656488549618321e-06,
      "loss": 0.0523,
      "step": 3062
    },
    {
      "epoch": 18.68,
      "learning_rate": 6.625954198473283e-06,
      "loss": 0.0518,
      "step": 3063
    },
    {
      "epoch": 18.68,
      "learning_rate": 6.595419847328245e-06,
      "loss": 0.0667,
      "step": 3064
    },
    {
      "epoch": 18.69,
      "learning_rate": 6.564885496183207e-06,
      "loss": 0.0588,
      "step": 3065
    },
    {
      "epoch": 18.7,
      "learning_rate": 6.534351145038169e-06,
      "loss": 0.0468,
      "step": 3066
    },
    {
      "epoch": 18.7,
      "learning_rate": 6.503816793893131e-06,
      "loss": 0.0553,
      "step": 3067
    },
    {
      "epoch": 18.71,
      "learning_rate": 6.473282442748093e-06,
      "loss": 0.0466,
      "step": 3068
    },
    {
      "epoch": 18.71,
      "learning_rate": 6.442748091603054e-06,
      "loss": 0.0425,
      "step": 3069
    },
    {
      "epoch": 18.72,
      "learning_rate": 6.412213740458016e-06,
      "loss": 0.0477,
      "step": 3070
    },
    {
      "epoch": 18.73,
      "learning_rate": 6.381679389312978e-06,
      "loss": 0.0502,
      "step": 3071
    },
    {
      "epoch": 18.73,
      "learning_rate": 6.35114503816794e-06,
      "loss": 0.0566,
      "step": 3072
    },
    {
      "epoch": 18.74,
      "learning_rate": 6.320610687022902e-06,
      "loss": 0.0667,
      "step": 3073
    },
    {
      "epoch": 18.74,
      "learning_rate": 6.290076335877864e-06,
      "loss": 0.0562,
      "step": 3074
    },
    {
      "epoch": 18.75,
      "learning_rate": 6.259541984732825e-06,
      "loss": 0.0584,
      "step": 3075
    },
    {
      "epoch": 18.76,
      "learning_rate": 6.229007633587787e-06,
      "loss": 0.043,
      "step": 3076
    },
    {
      "epoch": 18.76,
      "learning_rate": 6.198473282442748e-06,
      "loss": 0.0477,
      "step": 3077
    },
    {
      "epoch": 18.77,
      "learning_rate": 6.16793893129771e-06,
      "loss": 0.0576,
      "step": 3078
    },
    {
      "epoch": 18.77,
      "learning_rate": 6.137404580152672e-06,
      "loss": 0.0508,
      "step": 3079
    },
    {
      "epoch": 18.78,
      "learning_rate": 6.106870229007634e-06,
      "loss": 0.0556,
      "step": 3080
    },
    {
      "epoch": 18.79,
      "learning_rate": 6.076335877862596e-06,
      "loss": 0.0636,
      "step": 3081
    },
    {
      "epoch": 18.79,
      "learning_rate": 6.045801526717558e-06,
      "loss": 0.0553,
      "step": 3082
    },
    {
      "epoch": 18.8,
      "learning_rate": 6.015267175572519e-06,
      "loss": 0.0565,
      "step": 3083
    },
    {
      "epoch": 18.8,
      "learning_rate": 5.984732824427481e-06,
      "loss": 0.0503,
      "step": 3084
    },
    {
      "epoch": 18.81,
      "learning_rate": 5.954198473282443e-06,
      "loss": 0.0569,
      "step": 3085
    },
    {
      "epoch": 18.82,
      "learning_rate": 5.923664122137405e-06,
      "loss": 0.0541,
      "step": 3086
    },
    {
      "epoch": 18.82,
      "learning_rate": 5.893129770992367e-06,
      "loss": 0.0551,
      "step": 3087
    },
    {
      "epoch": 18.83,
      "learning_rate": 5.8625954198473286e-06,
      "loss": 0.0508,
      "step": 3088
    },
    {
      "epoch": 18.84,
      "learning_rate": 5.8320610687022905e-06,
      "loss": 0.061,
      "step": 3089
    },
    {
      "epoch": 18.84,
      "learning_rate": 5.801526717557252e-06,
      "loss": 0.0631,
      "step": 3090
    },
    {
      "epoch": 18.85,
      "learning_rate": 5.770992366412214e-06,
      "loss": 0.0503,
      "step": 3091
    },
    {
      "epoch": 18.85,
      "learning_rate": 5.7404580152671756e-06,
      "loss": 0.0545,
      "step": 3092
    },
    {
      "epoch": 18.86,
      "learning_rate": 5.7099236641221375e-06,
      "loss": 0.0549,
      "step": 3093
    },
    {
      "epoch": 18.87,
      "learning_rate": 5.6793893129770995e-06,
      "loss": 0.0521,
      "step": 3094
    },
    {
      "epoch": 18.87,
      "learning_rate": 5.6488549618320615e-06,
      "loss": 0.0461,
      "step": 3095
    },
    {
      "epoch": 18.88,
      "learning_rate": 5.618320610687023e-06,
      "loss": 0.0512,
      "step": 3096
    },
    {
      "epoch": 18.88,
      "learning_rate": 5.5877862595419845e-06,
      "loss": 0.0464,
      "step": 3097
    },
    {
      "epoch": 18.89,
      "learning_rate": 5.5572519083969465e-06,
      "loss": 0.0586,
      "step": 3098
    },
    {
      "epoch": 18.9,
      "learning_rate": 5.5267175572519085e-06,
      "loss": 0.052,
      "step": 3099
    },
    {
      "epoch": 18.9,
      "learning_rate": 5.49618320610687e-06,
      "loss": 0.0538,
      "step": 3100
    },
    {
      "epoch": 18.91,
      "learning_rate": 5.465648854961832e-06,
      "loss": 0.0553,
      "step": 3101
    },
    {
      "epoch": 18.91,
      "learning_rate": 5.435114503816794e-06,
      "loss": 0.0537,
      "step": 3102
    },
    {
      "epoch": 18.92,
      "learning_rate": 5.4045801526717555e-06,
      "loss": 0.0534,
      "step": 3103
    },
    {
      "epoch": 18.93,
      "learning_rate": 5.3740458015267174e-06,
      "loss": 0.0695,
      "step": 3104
    },
    {
      "epoch": 18.93,
      "learning_rate": 5.343511450381679e-06,
      "loss": 0.0554,
      "step": 3105
    },
    {
      "epoch": 18.94,
      "learning_rate": 5.312977099236641e-06,
      "loss": 0.0508,
      "step": 3106
    },
    {
      "epoch": 18.95,
      "learning_rate": 5.282442748091603e-06,
      "loss": 0.0504,
      "step": 3107
    },
    {
      "epoch": 18.95,
      "learning_rate": 5.251908396946565e-06,
      "loss": 0.0664,
      "step": 3108
    },
    {
      "epoch": 18.96,
      "learning_rate": 5.221374045801527e-06,
      "loss": 0.0508,
      "step": 3109
    },
    {
      "epoch": 18.96,
      "learning_rate": 5.190839694656488e-06,
      "loss": 0.0666,
      "step": 3110
    },
    {
      "epoch": 18.97,
      "learning_rate": 5.16030534351145e-06,
      "loss": 0.0529,
      "step": 3111
    },
    {
      "epoch": 18.98,
      "learning_rate": 5.129770992366412e-06,
      "loss": 0.0535,
      "step": 3112
    },
    {
      "epoch": 18.98,
      "learning_rate": 5.099236641221374e-06,
      "loss": 0.0428,
      "step": 3113
    },
    {
      "epoch": 18.99,
      "learning_rate": 5.068702290076336e-06,
      "loss": 0.0528,
      "step": 3114
    },
    {
      "epoch": 18.99,
      "learning_rate": 5.038167938931298e-06,
      "loss": 0.0529,
      "step": 3115
    },
    {
      "epoch": 19.0,
      "learning_rate": 5.00763358778626e-06,
      "loss": 0.0585,
      "step": 3116
    },
    {
      "epoch": 19.01,
      "learning_rate": 4.977099236641221e-06,
      "loss": 0.0579,
      "step": 3117
    },
    {
      "epoch": 19.01,
      "learning_rate": 4.946564885496183e-06,
      "loss": 0.049,
      "step": 3118
    },
    {
      "epoch": 19.02,
      "learning_rate": 4.916030534351145e-06,
      "loss": 0.052,
      "step": 3119
    },
    {
      "epoch": 19.02,
      "learning_rate": 4.885496183206107e-06,
      "loss": 0.0556,
      "step": 3120
    },
    {
      "epoch": 19.03,
      "learning_rate": 4.854961832061069e-06,
      "loss": 0.0577,
      "step": 3121
    },
    {
      "epoch": 19.04,
      "learning_rate": 4.824427480916031e-06,
      "loss": 0.0473,
      "step": 3122
    },
    {
      "epoch": 19.04,
      "learning_rate": 4.793893129770992e-06,
      "loss": 0.0502,
      "step": 3123
    },
    {
      "epoch": 19.05,
      "learning_rate": 4.763358778625954e-06,
      "loss": 0.0615,
      "step": 3124
    },
    {
      "epoch": 19.05,
      "learning_rate": 4.732824427480916e-06,
      "loss": 0.0508,
      "step": 3125
    },
    {
      "epoch": 19.06,
      "learning_rate": 4.702290076335878e-06,
      "loss": 0.0568,
      "step": 3126
    },
    {
      "epoch": 19.07,
      "learning_rate": 4.67175572519084e-06,
      "loss": 0.0593,
      "step": 3127
    },
    {
      "epoch": 19.07,
      "learning_rate": 4.641221374045802e-06,
      "loss": 0.0519,
      "step": 3128
    },
    {
      "epoch": 19.08,
      "learning_rate": 4.610687022900764e-06,
      "loss": 0.0582,
      "step": 3129
    },
    {
      "epoch": 19.09,
      "learning_rate": 4.580152671755725e-06,
      "loss": 0.0584,
      "step": 3130
    },
    {
      "epoch": 19.09,
      "learning_rate": 4.549618320610687e-06,
      "loss": 0.0532,
      "step": 3131
    },
    {
      "epoch": 19.1,
      "learning_rate": 4.519083969465649e-06,
      "loss": 0.0543,
      "step": 3132
    },
    {
      "epoch": 19.1,
      "learning_rate": 4.488549618320611e-06,
      "loss": 0.0564,
      "step": 3133
    },
    {
      "epoch": 19.11,
      "learning_rate": 4.458015267175573e-06,
      "loss": 0.0522,
      "step": 3134
    },
    {
      "epoch": 19.12,
      "learning_rate": 4.427480916030535e-06,
      "loss": 0.0512,
      "step": 3135
    },
    {
      "epoch": 19.12,
      "learning_rate": 4.396946564885496e-06,
      "loss": 0.0483,
      "step": 3136
    },
    {
      "epoch": 19.13,
      "learning_rate": 4.366412213740458e-06,
      "loss": 0.0675,
      "step": 3137
    },
    {
      "epoch": 19.13,
      "learning_rate": 4.33587786259542e-06,
      "loss": 0.0568,
      "step": 3138
    },
    {
      "epoch": 19.14,
      "learning_rate": 4.305343511450382e-06,
      "loss": 0.0496,
      "step": 3139
    },
    {
      "epoch": 19.15,
      "learning_rate": 4.274809160305344e-06,
      "loss": 0.0629,
      "step": 3140
    },
    {
      "epoch": 19.15,
      "learning_rate": 4.244274809160306e-06,
      "loss": 0.0535,
      "step": 3141
    },
    {
      "epoch": 19.16,
      "learning_rate": 4.213740458015268e-06,
      "loss": 0.0512,
      "step": 3142
    },
    {
      "epoch": 19.16,
      "learning_rate": 4.183206106870229e-06,
      "loss": 0.0562,
      "step": 3143
    },
    {
      "epoch": 19.17,
      "learning_rate": 4.152671755725191e-06,
      "loss": 0.0552,
      "step": 3144
    },
    {
      "epoch": 19.18,
      "learning_rate": 4.122137404580153e-06,
      "loss": 0.0542,
      "step": 3145
    },
    {
      "epoch": 19.18,
      "learning_rate": 4.091603053435115e-06,
      "loss": 0.0512,
      "step": 3146
    },
    {
      "epoch": 19.19,
      "learning_rate": 4.061068702290077e-06,
      "loss": 0.0571,
      "step": 3147
    },
    {
      "epoch": 19.2,
      "learning_rate": 4.030534351145039e-06,
      "loss": 0.0631,
      "step": 3148
    },
    {
      "epoch": 19.2,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0484,
      "step": 3149
    },
    {
      "epoch": 19.21,
      "learning_rate": 3.969465648854962e-06,
      "loss": 0.0449,
      "step": 3150
    },
    {
      "epoch": 19.21,
      "learning_rate": 3.938931297709924e-06,
      "loss": 0.057,
      "step": 3151
    },
    {
      "epoch": 19.22,
      "learning_rate": 3.908396946564886e-06,
      "loss": 0.0573,
      "step": 3152
    },
    {
      "epoch": 19.23,
      "learning_rate": 3.877862595419848e-06,
      "loss": 0.0481,
      "step": 3153
    },
    {
      "epoch": 19.23,
      "learning_rate": 3.84732824427481e-06,
      "loss": 0.0616,
      "step": 3154
    },
    {
      "epoch": 19.24,
      "learning_rate": 3.816793893129772e-06,
      "loss": 0.052,
      "step": 3155
    },
    {
      "epoch": 19.24,
      "learning_rate": 3.786259541984733e-06,
      "loss": 0.0466,
      "step": 3156
    },
    {
      "epoch": 19.25,
      "learning_rate": 3.755725190839695e-06,
      "loss": 0.0573,
      "step": 3157
    },
    {
      "epoch": 19.26,
      "learning_rate": 3.7251908396946566e-06,
      "loss": 0.046,
      "step": 3158
    },
    {
      "epoch": 19.26,
      "learning_rate": 3.6946564885496186e-06,
      "loss": 0.0628,
      "step": 3159
    },
    {
      "epoch": 19.27,
      "learning_rate": 3.6641221374045806e-06,
      "loss": 0.0478,
      "step": 3160
    },
    {
      "epoch": 19.27,
      "learning_rate": 3.633587786259542e-06,
      "loss": 0.0711,
      "step": 3161
    },
    {
      "epoch": 19.28,
      "learning_rate": 3.603053435114504e-06,
      "loss": 0.0562,
      "step": 3162
    },
    {
      "epoch": 19.29,
      "learning_rate": 3.572519083969466e-06,
      "loss": 0.0527,
      "step": 3163
    },
    {
      "epoch": 19.29,
      "learning_rate": 3.541984732824428e-06,
      "loss": 0.0455,
      "step": 3164
    },
    {
      "epoch": 19.3,
      "learning_rate": 3.5114503816793895e-06,
      "loss": 0.0507,
      "step": 3165
    },
    {
      "epoch": 19.3,
      "learning_rate": 3.4809160305343515e-06,
      "loss": 0.0438,
      "step": 3166
    },
    {
      "epoch": 19.31,
      "learning_rate": 3.4503816793893135e-06,
      "loss": 0.0444,
      "step": 3167
    },
    {
      "epoch": 19.32,
      "learning_rate": 3.419847328244275e-06,
      "loss": 0.0499,
      "step": 3168
    },
    {
      "epoch": 19.32,
      "learning_rate": 3.389312977099237e-06,
      "loss": 0.056,
      "step": 3169
    },
    {
      "epoch": 19.33,
      "learning_rate": 3.358778625954199e-06,
      "loss": 0.0553,
      "step": 3170
    },
    {
      "epoch": 19.34,
      "learning_rate": 3.3282442748091605e-06,
      "loss": 0.0547,
      "step": 3171
    },
    {
      "epoch": 19.34,
      "learning_rate": 3.2977099236641224e-06,
      "loss": 0.0541,
      "step": 3172
    },
    {
      "epoch": 19.35,
      "learning_rate": 3.2671755725190844e-06,
      "loss": 0.0529,
      "step": 3173
    },
    {
      "epoch": 19.35,
      "learning_rate": 3.2366412213740463e-06,
      "loss": 0.052,
      "step": 3174
    },
    {
      "epoch": 19.36,
      "learning_rate": 3.206106870229008e-06,
      "loss": 0.0573,
      "step": 3175
    },
    {
      "epoch": 19.37,
      "learning_rate": 3.17557251908397e-06,
      "loss": 0.0603,
      "step": 3176
    },
    {
      "epoch": 19.37,
      "learning_rate": 3.145038167938932e-06,
      "loss": 0.043,
      "step": 3177
    },
    {
      "epoch": 19.38,
      "learning_rate": 3.1145038167938934e-06,
      "loss": 0.0534,
      "step": 3178
    },
    {
      "epoch": 19.38,
      "learning_rate": 3.083969465648855e-06,
      "loss": 0.0668,
      "step": 3179
    },
    {
      "epoch": 19.39,
      "learning_rate": 3.053435114503817e-06,
      "loss": 0.0479,
      "step": 3180
    },
    {
      "epoch": 19.4,
      "learning_rate": 3.022900763358779e-06,
      "loss": 0.0537,
      "step": 3181
    },
    {
      "epoch": 19.4,
      "learning_rate": 2.9923664122137404e-06,
      "loss": 0.0565,
      "step": 3182
    },
    {
      "epoch": 19.41,
      "learning_rate": 2.9618320610687023e-06,
      "loss": 0.0639,
      "step": 3183
    },
    {
      "epoch": 19.41,
      "learning_rate": 2.9312977099236643e-06,
      "loss": 0.05,
      "step": 3184
    },
    {
      "epoch": 19.42,
      "learning_rate": 2.900763358778626e-06,
      "loss": 0.0492,
      "step": 3185
    },
    {
      "epoch": 19.43,
      "learning_rate": 2.8702290076335878e-06,
      "loss": 0.0561,
      "step": 3186
    },
    {
      "epoch": 19.43,
      "learning_rate": 2.8396946564885497e-06,
      "loss": 0.0748,
      "step": 3187
    },
    {
      "epoch": 19.44,
      "learning_rate": 2.8091603053435117e-06,
      "loss": 0.0475,
      "step": 3188
    },
    {
      "epoch": 19.45,
      "learning_rate": 2.7786259541984732e-06,
      "loss": 0.0501,
      "step": 3189
    },
    {
      "epoch": 19.45,
      "learning_rate": 2.748091603053435e-06,
      "loss": 0.0578,
      "step": 3190
    },
    {
      "epoch": 19.46,
      "learning_rate": 2.717557251908397e-06,
      "loss": 0.0587,
      "step": 3191
    },
    {
      "epoch": 19.46,
      "learning_rate": 2.6870229007633587e-06,
      "loss": 0.0591,
      "step": 3192
    },
    {
      "epoch": 19.47,
      "learning_rate": 2.6564885496183207e-06,
      "loss": 0.0585,
      "step": 3193
    },
    {
      "epoch": 19.48,
      "learning_rate": 2.6259541984732826e-06,
      "loss": 0.0447,
      "step": 3194
    },
    {
      "epoch": 19.48,
      "learning_rate": 2.595419847328244e-06,
      "loss": 0.0514,
      "step": 3195
    },
    {
      "epoch": 19.49,
      "learning_rate": 2.564885496183206e-06,
      "loss": 0.0514,
      "step": 3196
    },
    {
      "epoch": 19.49,
      "learning_rate": 2.534351145038168e-06,
      "loss": 0.051,
      "step": 3197
    },
    {
      "epoch": 19.5,
      "learning_rate": 2.50381679389313e-06,
      "loss": 0.0493,
      "step": 3198
    },
    {
      "epoch": 19.51,
      "learning_rate": 2.4732824427480916e-06,
      "loss": 0.0548,
      "step": 3199
    },
    {
      "epoch": 19.51,
      "learning_rate": 2.4427480916030536e-06,
      "loss": 0.0609,
      "step": 3200
    },
    {
      "epoch": 19.52,
      "learning_rate": 2.4122137404580155e-06,
      "loss": 0.0453,
      "step": 3201
    },
    {
      "epoch": 19.52,
      "learning_rate": 2.381679389312977e-06,
      "loss": 0.0548,
      "step": 3202
    },
    {
      "epoch": 19.53,
      "learning_rate": 2.351145038167939e-06,
      "loss": 0.0452,
      "step": 3203
    },
    {
      "epoch": 19.54,
      "learning_rate": 2.320610687022901e-06,
      "loss": 0.052,
      "step": 3204
    },
    {
      "epoch": 19.54,
      "learning_rate": 2.2900763358778625e-06,
      "loss": 0.0488,
      "step": 3205
    },
    {
      "epoch": 19.55,
      "learning_rate": 2.2595419847328245e-06,
      "loss": 0.0578,
      "step": 3206
    },
    {
      "epoch": 19.55,
      "learning_rate": 2.2290076335877865e-06,
      "loss": 0.0592,
      "step": 3207
    },
    {
      "epoch": 19.56,
      "learning_rate": 2.198473282442748e-06,
      "loss": 0.044,
      "step": 3208
    },
    {
      "epoch": 19.57,
      "learning_rate": 2.16793893129771e-06,
      "loss": 0.0481,
      "step": 3209
    },
    {
      "epoch": 19.57,
      "learning_rate": 2.137404580152672e-06,
      "loss": 0.0606,
      "step": 3210
    },
    {
      "epoch": 19.58,
      "learning_rate": 2.106870229007634e-06,
      "loss": 0.0584,
      "step": 3211
    },
    {
      "epoch": 19.59,
      "learning_rate": 2.0763358778625954e-06,
      "loss": 0.0378,
      "step": 3212
    },
    {
      "epoch": 19.59,
      "learning_rate": 2.0458015267175574e-06,
      "loss": 0.0471,
      "step": 3213
    },
    {
      "epoch": 19.6,
      "learning_rate": 2.0152671755725194e-06,
      "loss": 0.0525,
      "step": 3214
    },
    {
      "epoch": 19.6,
      "learning_rate": 1.984732824427481e-06,
      "loss": 0.0569,
      "step": 3215
    },
    {
      "epoch": 19.61,
      "learning_rate": 1.954198473282443e-06,
      "loss": 0.0582,
      "step": 3216
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.923664122137405e-06,
      "loss": 0.0558,
      "step": 3217
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.8931297709923666e-06,
      "loss": 0.0507,
      "step": 3218
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.8625954198473283e-06,
      "loss": 0.0622,
      "step": 3219
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.8320610687022903e-06,
      "loss": 0.0481,
      "step": 3220
    },
    {
      "epoch": 19.64,
      "learning_rate": 1.801526717557252e-06,
      "loss": 0.0482,
      "step": 3221
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.770992366412214e-06,
      "loss": 0.0557,
      "step": 3222
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.7404580152671757e-06,
      "loss": 0.0624,
      "step": 3223
    },
    {
      "epoch": 19.66,
      "learning_rate": 1.7099236641221375e-06,
      "loss": 0.0522,
      "step": 3224
    },
    {
      "epoch": 19.66,
      "learning_rate": 1.6793893129770995e-06,
      "loss": 0.0437,
      "step": 3225
    },
    {
      "epoch": 19.67,
      "learning_rate": 1.6488549618320612e-06,
      "loss": 0.0626,
      "step": 3226
    },
    {
      "epoch": 19.68,
      "learning_rate": 1.6183206106870232e-06,
      "loss": 0.0473,
      "step": 3227
    },
    {
      "epoch": 19.68,
      "learning_rate": 1.587786259541985e-06,
      "loss": 0.0561,
      "step": 3228
    },
    {
      "epoch": 19.69,
      "learning_rate": 1.5572519083969467e-06,
      "loss": 0.0537,
      "step": 3229
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.5267175572519084e-06,
      "loss": 0.0687,
      "step": 3230
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.4961832061068702e-06,
      "loss": 0.0547,
      "step": 3231
    },
    {
      "epoch": 19.71,
      "learning_rate": 1.4656488549618321e-06,
      "loss": 0.0576,
      "step": 3232
    },
    {
      "epoch": 19.71,
      "learning_rate": 1.4351145038167939e-06,
      "loss": 0.0649,
      "step": 3233
    },
    {
      "epoch": 19.72,
      "learning_rate": 1.4045801526717559e-06,
      "loss": 0.045,
      "step": 3234
    },
    {
      "epoch": 19.73,
      "learning_rate": 1.3740458015267176e-06,
      "loss": 0.0518,
      "step": 3235
    },
    {
      "epoch": 19.73,
      "learning_rate": 1.3435114503816794e-06,
      "loss": 0.0535,
      "step": 3236
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.3129770992366413e-06,
      "loss": 0.0529,
      "step": 3237
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.282442748091603e-06,
      "loss": 0.0497,
      "step": 3238
    },
    {
      "epoch": 19.75,
      "learning_rate": 1.251908396946565e-06,
      "loss": 0.0601,
      "step": 3239
    },
    {
      "epoch": 19.76,
      "learning_rate": 1.2213740458015268e-06,
      "loss": 0.0563,
      "step": 3240
    },
    {
      "epoch": 19.76,
      "learning_rate": 1.1908396946564885e-06,
      "loss": 0.0489,
      "step": 3241
    },
    {
      "epoch": 19.77,
      "learning_rate": 1.1603053435114505e-06,
      "loss": 0.0469,
      "step": 3242
    },
    {
      "epoch": 19.77,
      "learning_rate": 1.1297709923664122e-06,
      "loss": 0.0485,
      "step": 3243
    },
    {
      "epoch": 19.78,
      "learning_rate": 1.099236641221374e-06,
      "loss": 0.0437,
      "step": 3244
    },
    {
      "epoch": 19.79,
      "learning_rate": 1.068702290076336e-06,
      "loss": 0.0571,
      "step": 3245
    },
    {
      "epoch": 19.79,
      "learning_rate": 1.0381679389312977e-06,
      "loss": 0.0531,
      "step": 3246
    },
    {
      "epoch": 19.8,
      "learning_rate": 1.0076335877862597e-06,
      "loss": 0.0461,
      "step": 3247
    },
    {
      "epoch": 19.8,
      "learning_rate": 9.770992366412214e-07,
      "loss": 0.0515,
      "step": 3248
    },
    {
      "epoch": 19.81,
      "learning_rate": 9.465648854961833e-07,
      "loss": 0.0569,
      "step": 3249
    },
    {
      "epoch": 19.82,
      "learning_rate": 9.160305343511451e-07,
      "loss": 0.0697,
      "step": 3250
    },
    {
      "epoch": 19.82,
      "learning_rate": 8.85496183206107e-07,
      "loss": 0.0541,
      "step": 3251
    },
    {
      "epoch": 19.83,
      "learning_rate": 8.549618320610687e-07,
      "loss": 0.0554,
      "step": 3252
    },
    {
      "epoch": 19.84,
      "learning_rate": 8.244274809160306e-07,
      "loss": 0.0468,
      "step": 3253
    },
    {
      "epoch": 19.84,
      "learning_rate": 7.938931297709925e-07,
      "loss": 0.0567,
      "step": 3254
    },
    {
      "epoch": 19.85,
      "learning_rate": 7.633587786259542e-07,
      "loss": 0.0687,
      "step": 3255
    },
    {
      "epoch": 19.85,
      "learning_rate": 7.328244274809161e-07,
      "loss": 0.0627,
      "step": 3256
    },
    {
      "epoch": 19.86,
      "learning_rate": 7.022900763358779e-07,
      "loss": 0.0597,
      "step": 3257
    },
    {
      "epoch": 19.87,
      "learning_rate": 6.717557251908397e-07,
      "loss": 0.0429,
      "step": 3258
    },
    {
      "epoch": 19.87,
      "learning_rate": 6.412213740458015e-07,
      "loss": 0.0553,
      "step": 3259
    },
    {
      "epoch": 19.88,
      "learning_rate": 6.106870229007634e-07,
      "loss": 0.0552,
      "step": 3260
    },
    {
      "epoch": 19.88,
      "learning_rate": 5.801526717557252e-07,
      "loss": 0.0537,
      "step": 3261
    },
    {
      "epoch": 19.89,
      "learning_rate": 5.49618320610687e-07,
      "loss": 0.0446,
      "step": 3262
    },
    {
      "epoch": 19.9,
      "learning_rate": 5.190839694656489e-07,
      "loss": 0.0594,
      "step": 3263
    },
    {
      "epoch": 19.9,
      "learning_rate": 4.885496183206107e-07,
      "loss": 0.0476,
      "step": 3264
    },
    {
      "epoch": 19.91,
      "learning_rate": 4.5801526717557257e-07,
      "loss": 0.056,
      "step": 3265
    },
    {
      "epoch": 19.91,
      "learning_rate": 4.274809160305344e-07,
      "loss": 0.0555,
      "step": 3266
    },
    {
      "epoch": 19.92,
      "learning_rate": 3.9694656488549623e-07,
      "loss": 0.056,
      "step": 3267
    },
    {
      "epoch": 19.93,
      "learning_rate": 3.6641221374045804e-07,
      "loss": 0.0561,
      "step": 3268
    },
    {
      "epoch": 19.93,
      "learning_rate": 3.3587786259541984e-07,
      "loss": 0.0521,
      "step": 3269
    },
    {
      "epoch": 19.94,
      "learning_rate": 3.053435114503817e-07,
      "loss": 0.0649,
      "step": 3270
    },
    {
      "epoch": 19.95,
      "learning_rate": 2.748091603053435e-07,
      "loss": 0.058,
      "step": 3271
    },
    {
      "epoch": 19.95,
      "learning_rate": 2.4427480916030536e-07,
      "loss": 0.0592,
      "step": 3272
    },
    {
      "epoch": 19.96,
      "learning_rate": 2.137404580152672e-07,
      "loss": 0.0618,
      "step": 3273
    },
    {
      "epoch": 19.96,
      "learning_rate": 1.8320610687022902e-07,
      "loss": 0.0537,
      "step": 3274
    },
    {
      "epoch": 19.97,
      "learning_rate": 1.5267175572519085e-07,
      "loss": 0.0637,
      "step": 3275
    },
    {
      "epoch": 19.98,
      "learning_rate": 1.2213740458015268e-07,
      "loss": 0.0539,
      "step": 3276
    },
    {
      "epoch": 19.98,
      "learning_rate": 9.160305343511451e-08,
      "loss": 0.0578,
      "step": 3277
    },
    {
      "epoch": 19.99,
      "learning_rate": 6.106870229007634e-08,
      "loss": 0.052,
      "step": 3278
    },
    {
      "epoch": 19.99,
      "learning_rate": 3.053435114503817e-08,
      "loss": 0.0467,
      "step": 3279
    },
    {
      "epoch": 20.0,
      "learning_rate": 0.0,
      "loss": 0.0476,
      "step": 3280
    },
    {
      "epoch": 20.0,
      "step": 3280,
      "total_flos": 1.4741018341490688e+17,
      "train_loss": 0.13857511495262748,
      "train_runtime": 5194.9579,
      "train_samples_per_second": 2.514,
      "train_steps_per_second": 0.631
    }
  ],
  "logging_steps": 1,
  "max_steps": 3280,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 1.4741018341490688e+17,
  "trial_name": null,
  "trial_params": null
}