{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 19.93527508090615,
  "eval_steps": 500,
  "global_step": 3080,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 2.1505376344086023e-06,
      "loss": 3.4746,
      "step": 1
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.3010752688172045e-06,
      "loss": 3.4688,
      "step": 2
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.451612903225806e-06,
      "loss": 3.4727,
      "step": 3
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.602150537634409e-06,
      "loss": 3.4727,
      "step": 4
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.0752688172043012e-05,
      "loss": 3.4707,
      "step": 5
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.2903225806451613e-05,
      "loss": 3.4707,
      "step": 6
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.5053763440860215e-05,
      "loss": 3.4688,
      "step": 7
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.7204301075268818e-05,
      "loss": 3.4648,
      "step": 8
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.935483870967742e-05,
      "loss": 3.4492,
      "step": 9
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.1505376344086024e-05,
      "loss": 3.4297,
      "step": 10
    },
    {
      "epoch": 0.07,
      "learning_rate": 2.3655913978494626e-05,
      "loss": 3.4355,
      "step": 11
    },
    {
      "epoch": 0.08,
      "learning_rate": 2.5806451612903226e-05,
      "loss": 3.4238,
      "step": 12
    },
    {
      "epoch": 0.08,
      "learning_rate": 2.7956989247311828e-05,
      "loss": 3.4043,
      "step": 13
    },
    {
      "epoch": 0.09,
      "learning_rate": 3.010752688172043e-05,
      "loss": 3.4062,
      "step": 14
    },
    {
      "epoch": 0.1,
      "learning_rate": 3.2258064516129034e-05,
      "loss": 3.4023,
      "step": 15
    },
    {
      "epoch": 0.1,
      "learning_rate": 3.4408602150537636e-05,
      "loss": 3.373,
      "step": 16
    },
    {
      "epoch": 0.11,
      "learning_rate": 3.655913978494624e-05,
      "loss": 3.3496,
      "step": 17
    },
    {
      "epoch": 0.12,
      "learning_rate": 3.870967741935484e-05,
      "loss": 3.3164,
      "step": 18
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.0860215053763444e-05,
      "loss": 3.3086,
      "step": 19
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.301075268817205e-05,
      "loss": 3.2871,
      "step": 20
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.516129032258064e-05,
      "loss": 3.2695,
      "step": 21
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.731182795698925e-05,
      "loss": 3.2168,
      "step": 22
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9462365591397855e-05,
      "loss": 3.1562,
      "step": 23
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.161290322580645e-05,
      "loss": 3.1211,
      "step": 24
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.3763440860215054e-05,
      "loss": 3.0586,
      "step": 25
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.5913978494623656e-05,
      "loss": 2.9707,
      "step": 26
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.8064516129032266e-05,
      "loss": 3.0781,
      "step": 27
    },
    {
      "epoch": 0.18,
      "learning_rate": 6.021505376344086e-05,
      "loss": 3.0664,
      "step": 28
    },
    {
      "epoch": 0.19,
      "learning_rate": 6.236559139784946e-05,
      "loss": 2.9512,
      "step": 29
    },
    {
      "epoch": 0.19,
      "learning_rate": 6.451612903225807e-05,
      "loss": 2.9805,
      "step": 30
    },
    {
      "epoch": 0.2,
      "learning_rate": 6.666666666666667e-05,
      "loss": 2.9922,
      "step": 31
    },
    {
      "epoch": 0.21,
      "learning_rate": 6.881720430107527e-05,
      "loss": 2.8613,
      "step": 32
    },
    {
      "epoch": 0.21,
      "learning_rate": 7.096774193548388e-05,
      "loss": 2.9043,
      "step": 33
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.311827956989248e-05,
      "loss": 2.8555,
      "step": 34
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.526881720430108e-05,
      "loss": 2.7793,
      "step": 35
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.741935483870968e-05,
      "loss": 2.7441,
      "step": 36
    },
    {
      "epoch": 0.24,
      "learning_rate": 7.956989247311829e-05,
      "loss": 2.7773,
      "step": 37
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.172043010752689e-05,
      "loss": 2.7559,
      "step": 38
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.387096774193549e-05,
      "loss": 2.6504,
      "step": 39
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.60215053763441e-05,
      "loss": 2.7148,
      "step": 40
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.81720430107527e-05,
      "loss": 2.6445,
      "step": 41
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.032258064516129e-05,
      "loss": 2.5664,
      "step": 42
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.247311827956989e-05,
      "loss": 2.623,
      "step": 43
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.46236559139785e-05,
      "loss": 2.5039,
      "step": 44
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.677419354838711e-05,
      "loss": 2.6211,
      "step": 45
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.892473118279571e-05,
      "loss": 2.5234,
      "step": 46
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001010752688172043,
      "loss": 2.5078,
      "step": 47
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001032258064516129,
      "loss": 2.623,
      "step": 48
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001053763440860215,
      "loss": 2.5273,
      "step": 49
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00010752688172043011,
      "loss": 2.3887,
      "step": 50
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00010967741935483871,
      "loss": 2.4961,
      "step": 51
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00011182795698924731,
      "loss": 2.5156,
      "step": 52
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00011397849462365593,
      "loss": 2.4375,
      "step": 53
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00011612903225806453,
      "loss": 2.4434,
      "step": 54
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00011827956989247313,
      "loss": 2.5645,
      "step": 55
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012043010752688172,
      "loss": 2.3984,
      "step": 56
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012258064516129034,
      "loss": 2.3906,
      "step": 57
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012473118279569893,
      "loss": 2.3711,
      "step": 58
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012688172043010752,
      "loss": 2.3789,
      "step": 59
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012903225806451613,
      "loss": 2.252,
      "step": 60
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013118279569892472,
      "loss": 2.3379,
      "step": 61
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013333333333333334,
      "loss": 2.3574,
      "step": 62
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013548387096774193,
      "loss": 2.3086,
      "step": 63
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013763440860215055,
      "loss": 2.2539,
      "step": 64
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013978494623655916,
      "loss": 2.3789,
      "step": 65
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00014193548387096775,
      "loss": 2.4512,
      "step": 66
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00014408602150537637,
      "loss": 2.2734,
      "step": 67
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00014623655913978496,
      "loss": 2.29,
      "step": 68
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00014838709677419355,
      "loss": 2.2617,
      "step": 69
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00015053763440860216,
      "loss": 2.1113,
      "step": 70
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00015268817204301075,
      "loss": 2.1582,
      "step": 71
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00015483870967741937,
      "loss": 2.0635,
      "step": 72
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00015698924731182796,
      "loss": 2.125,
      "step": 73
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00015913978494623657,
      "loss": 2.3652,
      "step": 74
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00016129032258064516,
      "loss": 2.1328,
      "step": 75
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00016344086021505378,
      "loss": 2.1816,
      "step": 76
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001655913978494624,
      "loss": 2.1934,
      "step": 77
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00016774193548387098,
      "loss": 2.2881,
      "step": 78
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016989247311827957,
      "loss": 2.1807,
      "step": 79
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001720430107526882,
      "loss": 2.2383,
      "step": 80
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00017419354838709678,
      "loss": 2.1992,
      "step": 81
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001763440860215054,
      "loss": 2.1621,
      "step": 82
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00017849462365591398,
      "loss": 2.2109,
      "step": 83
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018064516129032257,
      "loss": 2.209,
      "step": 84
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001827956989247312,
      "loss": 2.0771,
      "step": 85
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018494623655913978,
      "loss": 2.1934,
      "step": 86
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001870967741935484,
      "loss": 2.1484,
      "step": 87
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000189247311827957,
      "loss": 2.0742,
      "step": 88
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001913978494623656,
      "loss": 2.1719,
      "step": 89
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00019354838709677422,
      "loss": 2.1035,
      "step": 90
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001956989247311828,
      "loss": 2.0234,
      "step": 91
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00019784946236559142,
      "loss": 2.0469,
      "step": 92
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 1.9502,
      "step": 93
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001999999446905596,
      "loss": 1.9512,
      "step": 94
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001999997787622996,
      "loss": 1.9912,
      "step": 95
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00019999950221540348,
      "loss": 1.9854,
      "step": 96
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001999991150501772,
      "loss": 1.8174,
      "step": 97
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00019999861726704908,
      "loss": 2.0107,
      "step": 98
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001999980088665697,
      "loss": 2.0303,
      "step": 99
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001999972898494121,
      "loss": 2.0381,
      "step": 100
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001999964602163716,
      "loss": 1.9268,
      "step": 101
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00019999551996836598,
      "loss": 1.9912,
      "step": 102
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001999944691064353,
      "loss": 1.9219,
      "step": 103
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00019999330763174206,
      "loss": 1.9736,
      "step": 104
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000199992035545571,
      "loss": 1.9922,
      "step": 105
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00019999065284932938,
      "loss": 2.0332,
      "step": 106
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00019998915954454663,
      "loss": 2.0586,
      "step": 107
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00019998755563287467,
      "loss": 2.0088,
      "step": 108
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00019998584111608777,
      "loss": 2.0117,
      "step": 109
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00019998401599608242,
      "loss": 1.9697,
      "step": 110
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00019998208027487756,
      "loss": 1.9395,
      "step": 111
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00019998003395461456,
      "loss": 2.0508,
      "step": 112
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00019997787703755692,
      "loss": 2.0088,
      "step": 113
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001999756095260907,
      "loss": 1.9629,
      "step": 114
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001999732314227241,
      "loss": 2.043,
      "step": 115
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001999707427300878,
      "loss": 2.0566,
      "step": 116
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00019996814345093477,
      "loss": 1.8818,
      "step": 117
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001999654335881403,
      "loss": 1.9209,
      "step": 118
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.000199962613144702,
      "loss": 1.9814,
      "step": 119
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001999596821237398,
      "loss": 1.9102,
      "step": 120
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00019995664052849601,
      "loss": 1.917,
      "step": 121
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00019995348836233516,
      "loss": 1.8623,
      "step": 122
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00019995022562874416,
      "loss": 2.0703,
      "step": 123
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00019994685233133222,
      "loss": 2.041,
      "step": 124
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00019994336847383083,
      "loss": 1.8301,
      "step": 125
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00019993977406009377,
      "loss": 1.9199,
      "step": 126
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001999360690940972,
      "loss": 1.9395,
      "step": 127
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00019993225357993946,
      "loss": 1.8516,
      "step": 128
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00019992832752184123,
      "loss": 1.8574,
      "step": 129
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001999242909241455,
      "loss": 1.9209,
      "step": 130
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00019992014379131753,
      "loss": 1.8633,
      "step": 131
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00019991588612794473,
      "loss": 1.8965,
      "step": 132
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.000199911517938737,
      "loss": 1.7568,
      "step": 133
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00019990703922852636,
      "loss": 1.8203,
      "step": 134
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00019990245000226704,
      "loss": 1.8604,
      "step": 135
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001998977502650356,
      "loss": 1.9932,
      "step": 136
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00019989294002203094,
      "loss": 1.9951,
      "step": 137
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00019988801927857397,
      "loss": 1.8662,
      "step": 138
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00019988298804010803,
      "loss": 1.7705,
      "step": 139
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00019987784631219862,
      "loss": 2.0176,
      "step": 140
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00019987259410053344,
      "loss": 1.7969,
      "step": 141
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00019986723141092246,
      "loss": 2.041,
      "step": 142
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00019986175824929776,
      "loss": 1.915,
      "step": 143
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00019985617462171376,
      "loss": 1.7686,
      "step": 144
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00019985048053434697,
      "loss": 1.8076,
      "step": 145
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00019984467599349612,
      "loss": 1.749,
      "step": 146
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00019983876100558215,
      "loss": 1.7959,
      "step": 147
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00019983273557714818,
      "loss": 1.7695,
      "step": 148
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001998265997148594,
      "loss": 1.6504,
      "step": 149
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001998203534255033,
      "loss": 1.8574,
      "step": 150
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00019981399671598939,
      "loss": 1.917,
      "step": 151
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00019980752959334946,
      "loss": 1.7256,
      "step": 152
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001998009520647373,
      "loss": 1.7725,
      "step": 153
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00019979426413742896,
      "loss": 1.8867,
      "step": 154
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00019978746581882252,
      "loss": 1.625,
      "step": 155
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00019978055711643823,
      "loss": 1.5195,
      "step": 156
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00019977353803791835,
      "loss": 1.7383,
      "step": 157
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00019976640859102737,
      "loss": 1.7217,
      "step": 158
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0001997591687836518,
      "loss": 1.6221,
      "step": 159
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019975181862380022,
      "loss": 1.5703,
      "step": 160
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019974435811960327,
      "loss": 1.5596,
      "step": 161
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019973678727931376,
      "loss": 1.6582,
      "step": 162
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019972910611130637,
      "loss": 1.5361,
      "step": 163
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019972131462407794,
      "loss": 1.5674,
      "step": 164
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001997134128262474,
      "loss": 1.7773,
      "step": 165
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019970540072655556,
      "loss": 1.6162,
      "step": 166
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00019969727833386532,
      "loss": 1.6982,
      "step": 167
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001996890456571616,
      "loss": 1.6064,
      "step": 168
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001996807027055513,
      "loss": 1.6631,
      "step": 169
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001996722494882633,
      "loss": 1.5176,
      "step": 170
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00019966368601464845,
      "loss": 1.5596,
      "step": 171
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00019965501229417953,
      "loss": 1.8301,
      "step": 172
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019964622833645136,
      "loss": 1.6504,
      "step": 173
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00019963733415118062,
      "loss": 1.5449,
      "step": 174
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.000199628329748206,
      "loss": 1.6035,
      "step": 175
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019961921513748804,
      "loss": 1.6807,
      "step": 176
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00019960999032910922,
      "loss": 1.5918,
      "step": 177
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00019960065533327397,
      "loss": 1.5615,
      "step": 178
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00019959121016030847,
      "loss": 1.4961,
      "step": 179
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00019958165482066094,
      "loss": 1.4404,
      "step": 180
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00019957198932490133,
      "loss": 1.582,
      "step": 181
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00019956221368372157,
      "loss": 1.373,
      "step": 182
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001995523279079353,
      "loss": 1.542,
      "step": 183
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00019954233200847812,
      "loss": 1.4775,
      "step": 184
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001995322259964073,
      "loss": 1.4951,
      "step": 185
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019952200988290208,
      "loss": 1.501,
      "step": 186
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019951168367926333,
      "loss": 1.6367,
      "step": 187
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019950124739691387,
      "loss": 1.6357,
      "step": 188
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019949070104739814,
      "loss": 1.6221,
      "step": 189
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019948004464238238,
      "loss": 1.5947,
      "step": 190
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00019946927819365463,
      "loss": 1.5645,
      "step": 191
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001994584017131246,
      "loss": 1.6299,
      "step": 192
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019944741521282376,
      "loss": 1.5332,
      "step": 193
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00019943631870490518,
      "loss": 1.7236,
      "step": 194
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00019942511220164376,
      "loss": 1.5098,
      "step": 195
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019941379571543596,
      "loss": 1.501,
      "step": 196
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019940236925879997,
      "loss": 1.5908,
      "step": 197
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019939083284437566,
      "loss": 1.6182,
      "step": 198
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019937918648492442,
      "loss": 1.583,
      "step": 199
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001993674301933293,
      "loss": 1.5342,
      "step": 200
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00019935556398259505,
      "loss": 1.3896,
      "step": 201
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00019934358786584787,
      "loss": 1.5176,
      "step": 202
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0001993315018563357,
      "loss": 1.4639,
      "step": 203
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00019931930596742782,
      "loss": 1.4727,
      "step": 204
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001993070002126153,
      "loss": 1.6494,
      "step": 205
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00019929458460551057,
      "loss": 1.5947,
      "step": 206
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00019928205915984768,
      "loss": 1.6641,
      "step": 207
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00019926942388948207,
      "loss": 1.5723,
      "step": 208
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0001992566788083908,
      "loss": 1.4414,
      "step": 209
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.0001992438239306723,
      "loss": 1.584,
      "step": 210
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019923085927054648,
      "loss": 1.6191,
      "step": 211
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019921778484235475,
      "loss": 1.6016,
      "step": 212
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019920460066055986,
      "loss": 1.4551,
      "step": 213
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.000199191306739746,
      "loss": 1.4268,
      "step": 214
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00019917790309461876,
      "loss": 1.5498,
      "step": 215
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019916438974000515,
      "loss": 1.4707,
      "step": 216
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019915076669085343,
      "loss": 1.4717,
      "step": 217
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00019913703396223328,
      "loss": 1.6035,
      "step": 218
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00019912319156933572,
      "loss": 1.3838,
      "step": 219
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.000199109239527473,
      "loss": 1.6279,
      "step": 220
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00019909517785207878,
      "loss": 1.4951,
      "step": 221
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019908100655870788,
      "loss": 1.4424,
      "step": 222
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019906672566303638,
      "loss": 1.5439,
      "step": 223
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00019905233518086176,
      "loss": 1.498,
      "step": 224
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00019903783512810255,
      "loss": 1.5205,
      "step": 225
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00019902322552079854,
      "loss": 1.5117,
      "step": 226
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0001990085063751107,
      "loss": 1.5195,
      "step": 227
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00019899367770732123,
      "loss": 1.4932,
      "step": 228
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0001989787395338334,
      "loss": 1.3613,
      "step": 229
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00019896369187117165,
      "loss": 1.6299,
      "step": 230
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00019894853473598158,
      "loss": 1.4873,
      "step": 231
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00019893326814502982,
      "loss": 1.5234,
      "step": 232
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00019891789211520406,
      "loss": 1.5771,
      "step": 233
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00019890240666351315,
      "loss": 1.5762,
      "step": 234
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001988868118070869,
      "loss": 1.5488,
      "step": 235
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00019887110756317613,
      "loss": 1.7451,
      "step": 236
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00019885529394915276,
      "loss": 1.4893,
      "step": 237
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00019883937098250963,
      "loss": 1.4287,
      "step": 238
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0001988233386808605,
      "loss": 1.5352,
      "step": 239
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.00019880719706194015,
      "loss": 1.5049,
      "step": 240
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.00019879094614360426,
      "loss": 1.5137,
      "step": 241
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0001987745859438294,
      "loss": 1.5859,
      "step": 242
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.00019875811648071307,
      "loss": 1.4824,
      "step": 243
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.00019874153777247355,
      "loss": 1.5645,
      "step": 244
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0001987248498374501,
      "loss": 1.5,
      "step": 245
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.00019870805269410266,
      "loss": 1.585,
      "step": 246
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00019869114636101207,
      "loss": 1.5576,
      "step": 247
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.00019867413085687994,
      "loss": 1.3545,
      "step": 248
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0001986570062005286,
      "loss": 1.5166,
      "step": 249
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00019863977241090113,
      "loss": 1.3745,
      "step": 250
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00019862242950706142,
      "loss": 1.4277,
      "step": 251
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00019860497750819397,
      "loss": 1.4219,
      "step": 252
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00019858741643360398,
      "loss": 1.5059,
      "step": 253
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00019856974630271733,
      "loss": 1.5352,
      "step": 254
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00019855196713508048,
      "loss": 1.5557,
      "step": 255
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0001985340789503606,
      "loss": 1.4844,
      "step": 256
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00019851608176834539,
      "loss": 1.5615,
      "step": 257
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00019849797560894309,
      "loss": 1.5918,
      "step": 258
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00019847976049218254,
      "loss": 1.416,
      "step": 259
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00019846143643821313,
      "loss": 1.4199,
      "step": 260
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0001984430034673047,
      "loss": 1.4551,
      "step": 261
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.00019842446159984762,
      "loss": 1.5713,
      "step": 262
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.00019840581085635268,
      "loss": 1.3359,
      "step": 263
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.00019838705125745112,
      "loss": 1.501,
      "step": 264
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0001983681828238946,
      "loss": 1.5,
      "step": 265
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.00019834920557655516,
      "loss": 1.4814,
      "step": 266
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00019833011953642525,
      "loss": 1.4492,
      "step": 267
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00019831092472461762,
      "loss": 1.5225,
      "step": 268
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.00019829162116236532,
      "loss": 1.375,
      "step": 269
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0001982722088710218,
      "loss": 1.3525,
      "step": 270
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.00019825268787206066,
      "loss": 1.5029,
      "step": 271
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.00019823305818707584,
      "loss": 1.5068,
      "step": 272
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.00019821331983778148,
      "loss": 1.4131,
      "step": 273
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.00019819347284601193,
      "loss": 1.4512,
      "step": 274
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.00019817351723372167,
      "loss": 1.4385,
      "step": 275
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00019815345302298543,
      "loss": 1.4893,
      "step": 276
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00019813328023599795,
      "loss": 1.4707,
      "step": 277
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.0001981129988950742,
      "loss": 1.5391,
      "step": 278
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00019809260902264915,
      "loss": 1.4746,
      "step": 279
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0001980721106412778,
      "loss": 1.3633,
      "step": 280
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00019805150377363533,
      "loss": 1.4209,
      "step": 281
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00019803078844251676,
      "loss": 1.4053,
      "step": 282
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00019800996467083716,
      "loss": 1.2822,
      "step": 283
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0001979890324816316,
      "loss": 1.6426,
      "step": 284
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00019796799189805494,
      "loss": 1.4951,
      "step": 285
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.00019794684294338212,
      "loss": 1.4346,
      "step": 286
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.00019792558564100788,
      "loss": 1.4805,
      "step": 287
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.00019790422001444676,
      "loss": 1.4473,
      "step": 288
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0001978827460873332,
      "loss": 1.4668,
      "step": 289
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0001978611638834214,
      "loss": 1.4639,
      "step": 290
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0001978394734265854,
      "loss": 1.4658,
      "step": 291
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.0001978176747408189,
      "loss": 1.627,
      "step": 292
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00019779576785023537,
      "loss": 1.4307,
      "step": 293
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00019777375277906797,
      "loss": 1.4834,
      "step": 294
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00019775162955166953,
      "loss": 1.4023,
      "step": 295
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0001977293981925125,
      "loss": 1.5,
      "step": 296
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.000197707058726189,
      "loss": 1.4746,
      "step": 297
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.00019768461117741063,
      "loss": 1.3564,
      "step": 298
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.00019766205557100868,
      "loss": 1.3057,
      "step": 299
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.00019763939193193387,
      "loss": 1.3252,
      "step": 300
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00019761662028525652,
      "loss": 1.3916,
      "step": 301
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00019759374065616628,
      "loss": 1.4561,
      "step": 302
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00019757075306997242,
      "loss": 1.3799,
      "step": 303
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00019754765755210354,
      "loss": 1.4531,
      "step": 304
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0001975244541281076,
      "loss": 1.291,
      "step": 305
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00019750114282365199,
      "loss": 1.5225,
      "step": 306
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.00019747772366452343,
      "loss": 1.4307,
      "step": 307
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0001974541966766279,
      "loss": 1.4766,
      "step": 308
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0001974305618859907,
      "loss": 1.4502,
      "step": 309
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0001974068193187564,
      "loss": 1.2827,
      "step": 310
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.00019738296900118877,
      "loss": 1.3398,
      "step": 311
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0001973590109596707,
      "loss": 1.1094,
      "step": 312
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00019733494522070438,
      "loss": 1.2598,
      "step": 313
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00019731077181091097,
      "loss": 1.1362,
      "step": 314
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00019728649075703093,
      "loss": 1.1836,
      "step": 315
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00019726210208592361,
      "loss": 1.1514,
      "step": 316
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00019723760582456755,
      "loss": 1.126,
      "step": 317
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00019721300200006018,
      "loss": 1.1504,
      "step": 318
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.000197188290639618,
      "loss": 1.2729,
      "step": 319
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.00019716347177057644,
      "loss": 1.3032,
      "step": 320
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00019713854542038986,
      "loss": 1.2773,
      "step": 321
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00019711351161663153,
      "loss": 1.2969,
      "step": 322
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0001970883703869935,
      "loss": 1.3008,
      "step": 323
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00019706312175928677,
      "loss": 1.1719,
      "step": 324
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00019703776576144105,
      "loss": 1.2144,
      "step": 325
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00019701230242150496,
      "loss": 1.3398,
      "step": 326
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00019698673176764563,
      "loss": 1.0942,
      "step": 327
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00019696105382814908,
      "loss": 1.2134,
      "step": 328
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00019693526863142003,
      "loss": 1.1475,
      "step": 329
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00019690937620598167,
      "loss": 1.1001,
      "step": 330
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00019688337658047597,
      "loss": 1.2266,
      "step": 331
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0001968572697836634,
      "loss": 1.3145,
      "step": 332
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.00019683105584442304,
      "loss": 1.2075,
      "step": 333
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.00019680473479175242,
      "loss": 1.1445,
      "step": 334
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.00019677830665476757,
      "loss": 1.0654,
      "step": 335
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0001967517714627031,
      "loss": 1.2676,
      "step": 336
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.00019672512924491184,
      "loss": 1.3218,
      "step": 337
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00019669838003086518,
      "loss": 1.2959,
      "step": 338
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00019667152385015275,
      "loss": 1.2373,
      "step": 339
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00019664456073248258,
      "loss": 1.1406,
      "step": 340
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00019661749070768098,
      "loss": 1.1958,
      "step": 341
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00019659031380569252,
      "loss": 1.1855,
      "step": 342
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00019656303005657991,
      "loss": 1.2119,
      "step": 343
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00019653563949052422,
      "loss": 1.2964,
      "step": 344
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00019650814213782454,
      "loss": 1.2139,
      "step": 345
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00019648053802889814,
      "loss": 1.3076,
      "step": 346
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00019645282719428037,
      "loss": 1.2129,
      "step": 347
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00019642500966462464,
      "loss": 1.3252,
      "step": 348
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00019639708547070247,
      "loss": 1.335,
      "step": 349
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00019636905464340317,
      "loss": 1.1094,
      "step": 350
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00019634091721373417,
      "loss": 1.3223,
      "step": 351
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00019631267321282083,
      "loss": 1.1309,
      "step": 352
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00019628432267190625,
      "loss": 1.2412,
      "step": 353
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0001962558656223516,
      "loss": 1.3408,
      "step": 354
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0001962273020956357,
      "loss": 1.1826,
      "step": 355
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00019619863212335515,
      "loss": 1.2661,
      "step": 356
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0001961698557372244,
      "loss": 1.1514,
      "step": 357
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00019614097296907556,
      "loss": 1.1826,
      "step": 358
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00019611198385085846,
      "loss": 1.1904,
      "step": 359
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.00019608288841464048,
      "loss": 1.1631,
      "step": 360
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00019605368669260668,
      "loss": 1.1445,
      "step": 361
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00019602437871705968,
      "loss": 1.1553,
      "step": 362
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0001959949645204197,
      "loss": 1.2832,
      "step": 363
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0001959654441352243,
      "loss": 1.1475,
      "step": 364
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.00019593581759412866,
      "loss": 1.1709,
      "step": 365
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0001959060849299053,
      "loss": 1.3613,
      "step": 366
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00019587624617544414,
      "loss": 1.2285,
      "step": 367
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00019584630136375254,
      "loss": 1.2539,
      "step": 368
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00019581625052795504,
      "loss": 1.3066,
      "step": 369
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0001957860937012936,
      "loss": 1.1167,
      "step": 370
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00019575583091712735,
      "loss": 1.1816,
      "step": 371
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00019572546220893261,
      "loss": 1.2183,
      "step": 372
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00019569498761030297,
      "loss": 1.3105,
      "step": 373
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.000195664407154949,
      "loss": 1.1523,
      "step": 374
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00019563372087669855,
      "loss": 1.1094,
      "step": 375
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00019560292880949634,
      "loss": 1.4746,
      "step": 376
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00019557203098740432,
      "loss": 1.3125,
      "step": 377
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.00019554102744460123,
      "loss": 1.2305,
      "step": 378
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0001955099182153829,
      "loss": 1.1958,
      "step": 379
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00019547870333416195,
      "loss": 1.2656,
      "step": 380
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.00019544738283546799,
      "loss": 1.0889,
      "step": 381
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.00019541595675394736,
      "loss": 1.335,
      "step": 382
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00019538442512436328,
      "loss": 1.2988,
      "step": 383
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00019535278798159565,
      "loss": 1.3477,
      "step": 384
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00019532104536064114,
      "loss": 1.2822,
      "step": 385
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.0001952891972966131,
      "loss": 1.5068,
      "step": 386
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00019525724382474143,
      "loss": 1.2563,
      "step": 387
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.00019522518498037283,
      "loss": 1.2051,
      "step": 388
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.00019519302079897032,
      "loss": 1.1836,
      "step": 389
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0001951607513161136,
      "loss": 1.0493,
      "step": 390
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.00019512837656749881,
      "loss": 1.2041,
      "step": 391
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00019509589658893854,
      "loss": 1.1997,
      "step": 392
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00019506331141636176,
      "loss": 1.3262,
      "step": 393
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00019503062108581384,
      "loss": 1.2856,
      "step": 394
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00019499782563345646,
      "loss": 1.3252,
      "step": 395
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00019496492509556756,
      "loss": 1.105,
      "step": 396
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00019493191950854138,
      "loss": 1.1953,
      "step": 397
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0001948988089088883,
      "loss": 1.2866,
      "step": 398
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.00019486559333323493,
      "loss": 1.3516,
      "step": 399
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00019483227281832393,
      "loss": 1.2402,
      "step": 400
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.00019479884740101412,
      "loss": 1.2627,
      "step": 401
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.00019476531711828027,
      "loss": 1.2012,
      "step": 402
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00019473168200721323,
      "loss": 1.3535,
      "step": 403
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00019469794210501978,
      "loss": 1.1758,
      "step": 404
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.00019466409744902267,
      "loss": 1.3418,
      "step": 405
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0001946301480766604,
      "loss": 1.165,
      "step": 406
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00019459609402548743,
      "loss": 1.2734,
      "step": 407
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.00019456193533317397,
      "loss": 1.3555,
      "step": 408
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.000194527672037506,
      "loss": 1.2363,
      "step": 409
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.00019449330417638513,
      "loss": 1.249,
      "step": 410
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00019445883178782877,
      "loss": 1.1436,
      "step": 411
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.00019442425490996988,
      "loss": 1.1807,
      "step": 412
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.00019438957358105697,
      "loss": 1.0889,
      "step": 413
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.00019435478783945417,
      "loss": 1.1294,
      "step": 414
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0001943198977236411,
      "loss": 1.2241,
      "step": 415
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.00019428490327221278,
      "loss": 1.166,
      "step": 416
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.0001942498045238797,
      "loss": 1.2144,
      "step": 417
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00019421460151746768,
      "loss": 1.2529,
      "step": 418
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00019417929429191793,
      "loss": 1.1641,
      "step": 419
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00019414388288628684,
      "loss": 1.1973,
      "step": 420
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.0001941083673397462,
      "loss": 1.2393,
      "step": 421
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00019407274769158285,
      "loss": 1.1816,
      "step": 422
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00019403702398119888,
      "loss": 1.1416,
      "step": 423
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.0001940011962481114,
      "loss": 1.251,
      "step": 424
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0001939652645319527,
      "loss": 1.1509,
      "step": 425
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.00019392922887247004,
      "loss": 1.1875,
      "step": 426
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.00019389308930952564,
      "loss": 1.1274,
      "step": 427
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0001938568458830967,
      "loss": 1.1299,
      "step": 428
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0001938204986332753,
      "loss": 1.3174,
      "step": 429
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.00019378404760026836,
      "loss": 1.1392,
      "step": 430
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.00019374749282439756,
      "loss": 1.165,
      "step": 431
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0001937108343460994,
      "loss": 1.2358,
      "step": 432
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00019367407220592515,
      "loss": 1.165,
      "step": 433
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00019363720644454058,
      "loss": 1.1064,
      "step": 434
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00019360023710272622,
      "loss": 1.0928,
      "step": 435
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00019356316422137717,
      "loss": 1.1597,
      "step": 436
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.000193525987841503,
      "loss": 1.2012,
      "step": 437
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00019348870800422782,
      "loss": 1.2158,
      "step": 438
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00019345132475079017,
      "loss": 1.1748,
      "step": 439
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.000193413838122543,
      "loss": 1.1982,
      "step": 440
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.00019337624816095358,
      "loss": 1.2798,
      "step": 441
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00019333855490760347,
      "loss": 1.2065,
      "step": 442
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00019330075840418858,
      "loss": 1.1357,
      "step": 443
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00019326285869251898,
      "loss": 1.1216,
      "step": 444
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.00019322485581451887,
      "loss": 1.3838,
      "step": 445
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.0001931867498122266,
      "loss": 1.252,
      "step": 446
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00019314854072779468,
      "loss": 1.1416,
      "step": 447
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.00019311022860348952,
      "loss": 1.2129,
      "step": 448
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.00019307181348169152,
      "loss": 1.3184,
      "step": 449
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.00019303329540489513,
      "loss": 1.1973,
      "step": 450
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0001929946744157086,
      "loss": 1.1226,
      "step": 451
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.00019295595055685398,
      "loss": 1.1978,
      "step": 452
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.00019291712387116725,
      "loss": 1.3682,
      "step": 453
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.000192878194401598,
      "loss": 1.2305,
      "step": 454
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.00019283916219120962,
      "loss": 1.313,
      "step": 455
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.00019280002728317905,
      "loss": 1.2695,
      "step": 456
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0001927607897207969,
      "loss": 1.3447,
      "step": 457
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0001927214495474674,
      "loss": 1.1074,
      "step": 458
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.00019268200680670807,
      "loss": 1.1973,
      "step": 459
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00019264246154215012,
      "loss": 1.3945,
      "step": 460
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00019260281379753812,
      "loss": 1.2402,
      "step": 461
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00019256306361672984,
      "loss": 1.2842,
      "step": 462
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00019252321104369657,
      "loss": 1.2393,
      "step": 463
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00019248325612252276,
      "loss": 1.043,
      "step": 464
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00019244319889740607,
      "loss": 1.0928,
      "step": 465
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.00019240303941265742,
      "loss": 1.1133,
      "step": 466
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0001923627777127007,
      "loss": 1.0635,
      "step": 467
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.00019232241384207303,
      "loss": 1.0371,
      "step": 468
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00019228194784542445,
      "loss": 1.1055,
      "step": 469
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00019224137976751795,
      "loss": 1.0811,
      "step": 470
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0001922007096532295,
      "loss": 1.0444,
      "step": 471
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00019215993754754797,
      "loss": 1.0332,
      "step": 472
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00019211906349557496,
      "loss": 0.8345,
      "step": 473
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0001920780875425249,
      "loss": 0.9038,
      "step": 474
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0001920370097337249,
      "loss": 1.0166,
      "step": 475
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00019199583011461489,
      "loss": 1.0312,
      "step": 476
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00019195454873074716,
      "loss": 0.9893,
      "step": 477
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00019191316562778679,
      "loss": 1.0015,
      "step": 478
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.00019187168085151127,
      "loss": 1.0254,
      "step": 479
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.00019183009444781067,
      "loss": 1.0669,
      "step": 480
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0001917884064626873,
      "loss": 0.9678,
      "step": 481
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.000191746616942256,
      "loss": 0.9922,
      "step": 482
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.00019170472593274388,
      "loss": 1.1436,
      "step": 483
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0001916627334804903,
      "loss": 1.0005,
      "step": 484
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00019162063963194679,
      "loss": 1.1182,
      "step": 485
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00019157844443367714,
      "loss": 0.9844,
      "step": 486
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00019153614793235725,
      "loss": 1.0088,
      "step": 487
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00019149375017477494,
      "loss": 1.0498,
      "step": 488
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00019145125120783015,
      "loss": 1.1411,
      "step": 489
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00019140865107853484,
      "loss": 1.0361,
      "step": 490
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00019136594983401275,
      "loss": 1.1431,
      "step": 491
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00019132314752149952,
      "loss": 1.0957,
      "step": 492
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.00019128024418834256,
      "loss": 1.1499,
      "step": 493
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0001912372398820011,
      "loss": 1.084,
      "step": 494
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.000191194134650046,
      "loss": 1.0327,
      "step": 495
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.00019115092854015976,
      "loss": 1.2075,
      "step": 496
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.00019110762160013658,
      "loss": 1.0068,
      "step": 497
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.00019106421387788202,
      "loss": 1.1938,
      "step": 498
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00019102070542141328,
      "loss": 1.0308,
      "step": 499
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.00019097709627885888,
      "loss": 0.9863,
      "step": 500
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.00019093338649845882,
      "loss": 1.123,
      "step": 501
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.0001908895761285643,
      "loss": 0.8833,
      "step": 502
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.00019084566521763793,
      "loss": 1.0361,
      "step": 503
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0001908016538142534,
      "loss": 0.9902,
      "step": 504
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00019075754196709572,
      "loss": 1.0239,
      "step": 505
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.00019071332972496086,
      "loss": 1.1318,
      "step": 506
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.00019066901713675585,
      "loss": 1.1465,
      "step": 507
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.00019062460425149893,
      "loss": 0.9761,
      "step": 508
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.000190580091118319,
      "loss": 0.9707,
      "step": 509
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.00019053547778645607,
      "loss": 1.0601,
      "step": 510
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00019049076430526087,
      "loss": 0.9517,
      "step": 511
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.000190445950724195,
      "loss": 1.0205,
      "step": 512
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00019040103709283066,
      "loss": 1.0322,
      "step": 513
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0001903560234608509,
      "loss": 1.0654,
      "step": 514
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0001903109098780492,
      "loss": 1.0117,
      "step": 515
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00019026569639432977,
      "loss": 1.1572,
      "step": 516
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.00019022038305970722,
      "loss": 1.0625,
      "step": 517
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.00019017496992430668,
      "loss": 0.9741,
      "step": 518
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.00019012945703836365,
      "loss": 1.1001,
      "step": 519
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.00019008384445222398,
      "loss": 0.9834,
      "step": 520
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.00019003813221634376,
      "loss": 1.0728,
      "step": 521
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.0001899923203812894,
      "loss": 1.1284,
      "step": 522
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.00018994640899773745,
      "loss": 1.1035,
      "step": 523
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0001899003981164745,
      "loss": 1.0576,
      "step": 524
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00018985428778839735,
      "loss": 0.998,
      "step": 525
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00018980807806451267,
      "loss": 0.9766,
      "step": 526
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00018976176899593717,
      "loss": 0.9814,
      "step": 527
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00018971536063389744,
      "loss": 0.9014,
      "step": 528
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00018966885302972982,
      "loss": 1.0859,
      "step": 529
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0001896222462348806,
      "loss": 1.0078,
      "step": 530
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.00018957554030090562,
      "loss": 1.146,
      "step": 531
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.00018952873527947048,
      "loss": 0.9102,
      "step": 532
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00018948183122235036,
      "loss": 0.9448,
      "step": 533
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00018943482818143003,
      "loss": 0.9873,
      "step": 534
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00018938772620870372,
      "loss": 1.0635,
      "step": 535
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.00018934052535627514,
      "loss": 1.0303,
      "step": 536
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00018929322567635726,
      "loss": 0.9956,
      "step": 537
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00018924582722127251,
      "loss": 0.9585,
      "step": 538
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00018919833004345252,
      "loss": 1.0391,
      "step": 539
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00018915073419543813,
      "loss": 0.9888,
      "step": 540
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00018910303972987936,
      "loss": 1.0562,
      "step": 541
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0001890552466995353,
      "loss": 0.9634,
      "step": 542
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.00018900735515727402,
      "loss": 1.1011,
      "step": 543
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.00018895936515607263,
      "loss": 1.0688,
      "step": 544
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.00018891127674901714,
      "loss": 1.0688,
      "step": 545
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.00018886308998930238,
      "loss": 0.9883,
      "step": 546
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00018881480493023204,
      "loss": 1.0679,
      "step": 547
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0001887664216252185,
      "loss": 0.999,
      "step": 548
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.00018871794012778284,
      "loss": 0.998,
      "step": 549
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00018866936049155472,
      "loss": 1.0098,
      "step": 550
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00018862068277027241,
      "loss": 1.0488,
      "step": 551
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00018857190701778267,
      "loss": 0.9751,
      "step": 552
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.00018852303328804066,
      "loss": 1.0283,
      "step": 553
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00018847406163511,
      "loss": 0.9653,
      "step": 554
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00018842499211316254,
      "loss": 0.9702,
      "step": 555
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00018837582477647846,
      "loss": 1.0625,
      "step": 556
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00018832655967944607,
      "loss": 1.0586,
      "step": 557
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00018827719687656197,
      "loss": 0.9531,
      "step": 558
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.00018822773642243064,
      "loss": 1.1206,
      "step": 559
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.00018817817837176476,
      "loss": 1.1318,
      "step": 560
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0001881285227793848,
      "loss": 1.0137,
      "step": 561
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0001880787697002193,
      "loss": 1.0767,
      "step": 562
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.00018802891918930456,
      "loss": 0.9507,
      "step": 563
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.00018797897130178458,
      "loss": 1.1992,
      "step": 564
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.00018792892609291123,
      "loss": 1.1172,
      "step": 565
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0001878787836180439,
      "loss": 0.9199,
      "step": 566
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00018782854393264967,
      "loss": 1.0703,
      "step": 567
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.00018777820709230315,
      "loss": 1.0586,
      "step": 568
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.00018772777315268635,
      "loss": 1.0215,
      "step": 569
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00018767724216958872,
      "loss": 1.1494,
      "step": 570
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.00018762661419890705,
      "loss": 1.0469,
      "step": 571
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.00018757588929664554,
      "loss": 0.9917,
      "step": 572
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0001875250675189154,
      "loss": 0.9009,
      "step": 573
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.00018747414892193511,
      "loss": 0.9707,
      "step": 574
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.00018742313356203032,
      "loss": 1.1885,
      "step": 575
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00018737202149563364,
      "loss": 1.0293,
      "step": 576
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00018732081277928462,
      "loss": 1.1055,
      "step": 577
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0001872695074696298,
      "loss": 1.1089,
      "step": 578
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00018721810562342253,
      "loss": 1.103,
      "step": 579
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00018716660729752298,
      "loss": 0.9707,
      "step": 580
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.000187115012548898,
      "loss": 1.0244,
      "step": 581
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0001870633214346211,
      "loss": 0.9678,
      "step": 582
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.00018701153401187243,
      "loss": 1.002,
      "step": 583
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0001869596503379387,
      "loss": 1.0303,
      "step": 584
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.000186907670470213,
      "loss": 1.0415,
      "step": 585
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0001868555944661949,
      "loss": 0.9722,
      "step": 586
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.00018680342238349028,
      "loss": 0.918,
      "step": 587
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0001867511542798113,
      "loss": 0.999,
      "step": 588
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0001866987902129764,
      "loss": 0.9854,
      "step": 589
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.00018664633024091006,
      "loss": 0.9233,
      "step": 590
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.00018659377442164298,
      "loss": 1.0557,
      "step": 591
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0001865411228133118,
      "loss": 1.1328,
      "step": 592
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0001864883754741591,
      "loss": 1.2676,
      "step": 593
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00018643553246253344,
      "loss": 1.0371,
      "step": 594
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.00018638259383688913,
      "loss": 1.106,
      "step": 595
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00018632955965578637,
      "loss": 1.0269,
      "step": 596
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00018627642997789083,
      "loss": 0.9839,
      "step": 597
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0001862232048619741,
      "loss": 1.1992,
      "step": 598
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.00018616988436691317,
      "loss": 0.9932,
      "step": 599
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0001861164685516905,
      "loss": 1.062,
      "step": 600
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00018606295747539419,
      "loss": 0.9067,
      "step": 601
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0001860093511972175,
      "loss": 1.0347,
      "step": 602
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.00018595564977645914,
      "loss": 0.9604,
      "step": 603
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00018590185327252304,
      "loss": 1.0845,
      "step": 604
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0001858479617449182,
      "loss": 1.042,
      "step": 605
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.00018579397525325892,
      "loss": 1.0596,
      "step": 606
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00018573989385726442,
      "loss": 1.0464,
      "step": 607
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.00018568571761675893,
      "loss": 1.1021,
      "step": 608
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0001856314465916716,
      "loss": 1.1299,
      "step": 609
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0001855770808420364,
      "loss": 0.9854,
      "step": 610
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0001855226204279922,
      "loss": 1.1187,
      "step": 611
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.00018546806540978246,
      "loss": 1.1562,
      "step": 612
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00018541341584775526,
      "loss": 1.167,
      "step": 613
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00018535867180236344,
      "loss": 0.9482,
      "step": 614
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.00018530383333416418,
      "loss": 1.0898,
      "step": 615
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0001852489005038192,
      "loss": 1.104,
      "step": 616
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.00018519387337209456,
      "loss": 1.1748,
      "step": 617
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00018513875199986074,
      "loss": 0.9141,
      "step": 618
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.00018508353644809228,
      "loss": 0.8579,
      "step": 619
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.00018502822677786805,
      "loss": 0.8462,
      "step": 620
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.000184972823050371,
      "loss": 0.8032,
      "step": 621
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.00018491732532688814,
      "loss": 0.8574,
      "step": 622
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.00018486173366881036,
      "loss": 0.9146,
      "step": 623
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00018480604813763258,
      "loss": 0.9316,
      "step": 624
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.00018475026879495346,
      "loss": 0.9692,
      "step": 625
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.00018469439570247556,
      "loss": 0.9136,
      "step": 626
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00018463842892200503,
      "loss": 0.7979,
      "step": 627
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00018458236851545166,
      "loss": 0.8237,
      "step": 628
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0001845262145448289,
      "loss": 0.9634,
      "step": 629
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00018446996707225367,
      "loss": 0.9199,
      "step": 630
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00018441362615994618,
      "loss": 0.9907,
      "step": 631
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.00018435719187023022,
      "loss": 1.0322,
      "step": 632
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.00018430066426553272,
      "loss": 0.9492,
      "step": 633
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0001842440434083839,
      "loss": 0.7708,
      "step": 634
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.00018418732936141714,
      "loss": 0.812,
      "step": 635
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.00018413052218736883,
      "loss": 0.8535,
      "step": 636
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.00018407362194907846,
      "loss": 0.8481,
      "step": 637
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00018401662870948848,
      "loss": 0.7432,
      "step": 638
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0001839595425316441,
      "loss": 0.8521,
      "step": 639
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.00018390236347869343,
      "loss": 0.8555,
      "step": 640
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00018384509161388732,
      "loss": 0.8042,
      "step": 641
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00018378772700057926,
      "loss": 0.8486,
      "step": 642
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00018373026970222535,
      "loss": 0.8574,
      "step": 643
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.0001836727197823842,
      "loss": 0.9321,
      "step": 644
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.0001836150773047169,
      "loss": 0.8257,
      "step": 645
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.00018355734233298687,
      "loss": 0.8184,
      "step": 646
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00018349951493105998,
      "loss": 0.8682,
      "step": 647
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00018344159516290413,
      "loss": 0.7832,
      "step": 648
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0001833835830925896,
      "loss": 0.9404,
      "step": 649
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0001833254787842887,
      "loss": 0.8955,
      "step": 650
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.00018326728230227572,
      "loss": 0.9282,
      "step": 651
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.00018320899371092702,
      "loss": 0.9487,
      "step": 652
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0001831506130747207,
      "loss": 0.9829,
      "step": 653
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.00018309214045823683,
      "loss": 0.749,
      "step": 654
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00018303357592615717,
      "loss": 0.8174,
      "step": 655
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0001829749195432651,
      "loss": 0.9917,
      "step": 656
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0001829161713744457,
      "loss": 0.8652,
      "step": 657
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.00018285733148468554,
      "loss": 0.7964,
      "step": 658
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0001827983999390726,
      "loss": 0.7603,
      "step": 659
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0001827393768027963,
      "loss": 0.8765,
      "step": 660
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00018268026214114742,
      "loss": 0.9092,
      "step": 661
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00018262105601951793,
      "loss": 0.9014,
      "step": 662
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.00018256175850340096,
      "loss": 0.8999,
      "step": 663
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00018250236965839073,
      "loss": 0.8594,
      "step": 664
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00018244288955018256,
      "loss": 0.8945,
      "step": 665
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.00018238331824457267,
      "loss": 0.8882,
      "step": 666
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0001823236558074582,
      "loss": 0.8618,
      "step": 667
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.000182263902304837,
      "loss": 0.8271,
      "step": 668
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.00018220405780280775,
      "loss": 0.9399,
      "step": 669
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00018214412236756982,
      "loss": 0.9531,
      "step": 670
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00018208409606542306,
      "loss": 0.9404,
      "step": 671
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00018202397896276793,
      "loss": 0.9712,
      "step": 672
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00018196377112610526,
      "loss": 0.9463,
      "step": 673
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00018190347262203634,
      "loss": 0.8711,
      "step": 674
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00018184308351726263,
      "loss": 0.9702,
      "step": 675
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.00018178260387858593,
      "loss": 0.7798,
      "step": 676
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0001817220337729081,
      "loss": 0.8569,
      "step": 677
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.00018166137326723115,
      "loss": 0.875,
      "step": 678
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0001816006224286571,
      "loss": 0.8906,
      "step": 679
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.00018153978132438774,
      "loss": 0.9492,
      "step": 680
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.00018147885002172485,
      "loss": 0.9785,
      "step": 681
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.00018141782858807002,
      "loss": 0.7607,
      "step": 682
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0001813567170909244,
      "loss": 0.998,
      "step": 683
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0001812955155978889,
      "loss": 0.7573,
      "step": 684
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.00018123422417666387,
      "loss": 0.832,
      "step": 685
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00018117284289504926,
      "loss": 0.8223,
      "step": 686
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.00018111137182094428,
      "loss": 0.917,
      "step": 687
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0001810498110223476,
      "loss": 0.894,
      "step": 688
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0001809881605673571,
      "loss": 0.9595,
      "step": 689
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.00018092642052416978,
      "loss": 0.8696,
      "step": 690
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0001808645909610818,
      "loss": 0.9854,
      "step": 691
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.00018080267194648831,
      "loss": 0.8755,
      "step": 692
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0001807406635488835,
      "loss": 0.9595,
      "step": 693
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.00018067856583686028,
      "loss": 0.9521,
      "step": 694
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00018061637887911052,
      "loss": 0.938,
      "step": 695
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00018055410274442466,
      "loss": 0.98,
      "step": 696
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0001804917375016919,
      "loss": 0.9761,
      "step": 697
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.0001804292832199,
      "loss": 0.8848,
      "step": 698
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00018036673996813516,
      "loss": 0.8306,
      "step": 699
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.000180304107815582,
      "loss": 0.8711,
      "step": 700
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00018024138683152358,
      "loss": 0.9404,
      "step": 701
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00018017857708534107,
      "loss": 0.9287,
      "step": 702
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.00018011567864651396,
      "loss": 0.8262,
      "step": 703
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00018005269158461978,
      "loss": 0.9492,
      "step": 704
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00017998961596933413,
      "loss": 0.9248,
      "step": 705
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0001799264518704305,
      "loss": 0.8696,
      "step": 706
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.00017986319935778036,
      "loss": 0.915,
      "step": 707
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0001797998585013529,
      "loss": 0.9609,
      "step": 708
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0001797364293712151,
      "loss": 0.7847,
      "step": 709
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.00017967291203753157,
      "loss": 0.959,
      "step": 710
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.00017960930657056438,
      "loss": 0.8608,
      "step": 711
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00017954561304067327,
      "loss": 0.8267,
      "step": 712
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00017948183151831527,
      "loss": 0.9277,
      "step": 713
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.00017941796207404483,
      "loss": 0.9009,
      "step": 714
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00017935400477851358,
      "loss": 0.8218,
      "step": 715
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0001792899597024704,
      "loss": 0.9956,
      "step": 716
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00017922582691676112,
      "loss": 1.0249,
      "step": 717
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.00017916160649232885,
      "loss": 0.9663,
      "step": 718
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.00017909729850021345,
      "loss": 0.9355,
      "step": 719
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0001790329030115517,
      "loss": 0.8232,
      "step": 720
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.0001789684200975772,
      "loss": 0.9546,
      "step": 721
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00017890384982962015,
      "loss": 0.9092,
      "step": 722
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.00017883919227910754,
      "loss": 0.9956,
      "step": 723
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00017877444751756278,
      "loss": 0.8945,
      "step": 724
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00017870961561660583,
      "loss": 0.8828,
      "step": 725
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.00017864469664795297,
      "loss": 0.9448,
      "step": 726
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.0001785796906834169,
      "loss": 0.9106,
      "step": 727
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00017851459779490644,
      "loss": 0.9917,
      "step": 728
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.00017844941805442662,
      "loss": 0.9121,
      "step": 729
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.00017838415153407857,
      "loss": 1.0591,
      "step": 730
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00017831879830605937,
      "loss": 0.9697,
      "step": 731
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.000178253358442662,
      "loss": 0.9541,
      "step": 732
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00017818783201627535,
      "loss": 0.9219,
      "step": 733
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.00017812221909938397,
      "loss": 0.9487,
      "step": 734
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0001780565197645682,
      "loss": 0.8486,
      "step": 735
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.00017799073408450387,
      "loss": 0.8662,
      "step": 736
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00017792486213196232,
      "loss": 1.0293,
      "step": 737
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00017785890397981044,
      "loss": 0.8833,
      "step": 738
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00017779285970101038,
      "loss": 0.8667,
      "step": 739
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00017772672936861955,
      "loss": 1.0205,
      "step": 740
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00017766051305579064,
      "loss": 0.7397,
      "step": 741
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00017759421083577133,
      "loss": 0.8843,
      "step": 742
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00017752782278190443,
      "loss": 0.8066,
      "step": 743
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.00017746134896762765,
      "loss": 0.9639,
      "step": 744
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0001773947894664736,
      "loss": 0.9331,
      "step": 745
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00017732814435206967,
      "loss": 0.9717,
      "step": 746
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00017726141369813788,
      "loss": 0.8838,
      "step": 747
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00017719459757849495,
      "loss": 1.0117,
      "step": 748
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00017712769606705216,
      "loss": 0.8833,
      "step": 749
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00017706070923781517,
      "loss": 0.8916,
      "step": 750
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00017699363716488408,
      "loss": 0.9229,
      "step": 751
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.0001769264799224533,
      "loss": 0.9834,
      "step": 752
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00017685923758481136,
      "loss": 0.9751,
      "step": 753
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.000176791910226341,
      "loss": 0.7949,
      "step": 754
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00017672449792151902,
      "loss": 1.0664,
      "step": 755
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00017665700074491613,
      "loss": 0.8921,
      "step": 756
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00017658941877119695,
      "loss": 0.9185,
      "step": 757
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00017652175207511994,
      "loss": 0.8984,
      "step": 758
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00017645400073153716,
      "loss": 1.0117,
      "step": 759
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.0001763861648153945,
      "loss": 0.9351,
      "step": 760
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00017631824440173118,
      "loss": 0.8716,
      "step": 761
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.0001762502395656801,
      "loss": 0.8867,
      "step": 762
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00017618215038246737,
      "loss": 0.8145,
      "step": 763
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00017611397692741254,
      "loss": 0.9697,
      "step": 764
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00017604571927592826,
      "loss": 0.812,
      "step": 765
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00017597737750352045,
      "loss": 0.8291,
      "step": 766
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00017590895168578797,
      "loss": 0.8335,
      "step": 767
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.0001758404418984227,
      "loss": 0.8867,
      "step": 768
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.0001757718482172094,
      "loss": 0.8555,
      "step": 769
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00017570317071802563,
      "loss": 0.7988,
      "step": 770
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.0001756344094768417,
      "loss": 0.8711,
      "step": 771
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0001755655645697205,
      "loss": 0.9546,
      "step": 772
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0001754966360728175,
      "loss": 0.7158,
      "step": 773
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00017542762406238063,
      "loss": 0.7605,
      "step": 774
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.0001753585286147502,
      "loss": 0.7158,
      "step": 775
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00017528934980635881,
      "loss": 0.6694,
      "step": 776
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00017522008771373132,
      "loss": 0.7463,
      "step": 777
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00017515074241348467,
      "loss": 0.7881,
      "step": 778
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00017508131398232782,
      "loss": 0.7095,
      "step": 779
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00017501180249706176,
      "loss": 0.6941,
      "step": 780
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.00017494220803457932,
      "loss": 0.6851,
      "step": 781
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0001748725306718651,
      "loss": 0.7593,
      "step": 782
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00017480277048599542,
      "loss": 0.7812,
      "step": 783
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.0001747329275541382,
      "loss": 0.7546,
      "step": 784
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00017466300195355294,
      "loss": 0.6543,
      "step": 785
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00017459299376159054,
      "loss": 0.6714,
      "step": 786
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.0001745229030556933,
      "loss": 0.7783,
      "step": 787
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00017445272991339474,
      "loss": 0.7466,
      "step": 788
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0001743824744123196,
      "loss": 0.7559,
      "step": 789
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.00017431213663018376,
      "loss": 0.7646,
      "step": 790
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00017424171664479413,
      "loss": 0.7739,
      "step": 791
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0001741712145340484,
      "loss": 0.7231,
      "step": 792
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0001741006303759353,
      "loss": 0.7075,
      "step": 793
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0001740299642485342,
      "loss": 0.8291,
      "step": 794
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00017395921623001518,
      "loss": 0.7847,
      "step": 795
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00017388838639863891,
      "loss": 0.769,
      "step": 796
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.00017381747483275658,
      "loss": 0.6846,
      "step": 797
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00017374648161080979,
      "loss": 0.6865,
      "step": 798
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00017367540681133034,
      "loss": 0.7598,
      "step": 799
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.00017360425051294047,
      "loss": 0.8486,
      "step": 800
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.00017353301279435246,
      "loss": 0.8032,
      "step": 801
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00017346169373436867,
      "loss": 0.7949,
      "step": 802
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00017339029341188145,
      "loss": 0.7456,
      "step": 803
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00017331881190587307,
      "loss": 0.8159,
      "step": 804
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0001732472492954155,
      "loss": 0.6841,
      "step": 805
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00017317560565967055,
      "loss": 0.7642,
      "step": 806
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00017310388107788963,
      "loss": 0.8569,
      "step": 807
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.00017303207562941356,
      "loss": 0.8906,
      "step": 808
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.0001729601893936729,
      "loss": 0.7217,
      "step": 809
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.00017288822245018723,
      "loss": 0.7075,
      "step": 810
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.00017281617487856568,
      "loss": 0.6594,
      "step": 811
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00017274404675850642,
      "loss": 0.7578,
      "step": 812
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00017267183816979678,
      "loss": 0.8643,
      "step": 813
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.0001725995491923131,
      "loss": 0.812,
      "step": 814
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00017252717990602062,
      "loss": 0.729,
      "step": 815
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00017245473039097348,
      "loss": 0.8755,
      "step": 816
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.00017238220072731447,
      "loss": 0.7773,
      "step": 817
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.00017230959099527512,
      "loss": 0.7793,
      "step": 818
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0001722369012751755,
      "loss": 0.7422,
      "step": 819
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00017216413164742415,
      "loss": 0.8667,
      "step": 820
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00017209128219251805,
      "loss": 0.812,
      "step": 821
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00017201835299104242,
      "loss": 0.7817,
      "step": 822
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.00017194534412367073,
      "loss": 0.8022,
      "step": 823
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.00017187225567116458,
      "loss": 0.7148,
      "step": 824
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00017179908771437362,
      "loss": 0.7471,
      "step": 825
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0001717258403342354,
      "loss": 0.6919,
      "step": 826
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.00017165251361177532,
      "loss": 0.729,
      "step": 827
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.00017157910762810664,
      "loss": 0.8237,
      "step": 828
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.0001715056224644302,
      "loss": 0.9004,
      "step": 829
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.0001714320582020345,
      "loss": 0.7925,
      "step": 830
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.00017135841492229545,
      "loss": 0.6699,
      "step": 831
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00017128469270667648,
      "loss": 0.7422,
      "step": 832
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00017121089163672825,
      "loss": 0.7456,
      "step": 833
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.00017113701179408868,
      "loss": 0.7681,
      "step": 834
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0001710630532604828,
      "loss": 0.8267,
      "step": 835
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00017098901611772274,
      "loss": 0.7681,
      "step": 836
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0001709149004477076,
      "loss": 0.9194,
      "step": 837
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00017084070633242325,
      "loss": 0.813,
      "step": 838
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00017076643385394237,
      "loss": 0.7007,
      "step": 839
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.0001706920830944244,
      "loss": 0.8901,
      "step": 840
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00017061765413611525,
      "loss": 0.8003,
      "step": 841
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.0001705431470613475,
      "loss": 0.7344,
      "step": 842
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00017046856195253998,
      "loss": 0.8506,
      "step": 843
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00017039389889219787,
      "loss": 0.8359,
      "step": 844
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.00017031915796291268,
      "loss": 0.7939,
      "step": 845
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00017024433924736195,
      "loss": 0.8286,
      "step": 846
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00017016944282830933,
      "loss": 0.7759,
      "step": 847
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00017009446878860435,
      "loss": 0.8481,
      "step": 848
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.00017001941721118255,
      "loss": 0.7969,
      "step": 849
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.00016994428817906505,
      "loss": 0.8325,
      "step": 850
    },
    {
      "epoch": 5.51,
      "learning_rate": 0.00016986908177535877,
      "loss": 0.9209,
      "step": 851
    },
    {
      "epoch": 5.51,
      "learning_rate": 0.00016979379808325625,
      "loss": 0.8115,
      "step": 852
    },
    {
      "epoch": 5.52,
      "learning_rate": 0.00016971843718603536,
      "loss": 0.7356,
      "step": 853
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.00016964299916705957,
      "loss": 0.7969,
      "step": 854
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.00016956748410977755,
      "loss": 0.7241,
      "step": 855
    },
    {
      "epoch": 5.54,
      "learning_rate": 0.0001694918920977232,
      "loss": 0.8384,
      "step": 856
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.0001694162232145155,
      "loss": 0.8774,
      "step": 857
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.00016934047754385865,
      "loss": 0.8228,
      "step": 858
    },
    {
      "epoch": 5.56,
      "learning_rate": 0.0001692646551695416,
      "loss": 0.7964,
      "step": 859
    },
    {
      "epoch": 5.57,
      "learning_rate": 0.00016918875617543818,
      "loss": 0.8132,
      "step": 860
    },
    {
      "epoch": 5.57,
      "learning_rate": 0.00016911278064550708,
      "loss": 0.7681,
      "step": 861
    },
    {
      "epoch": 5.58,
      "learning_rate": 0.00016903672866379148,
      "loss": 0.7129,
      "step": 862
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.0001689606003144193,
      "loss": 0.7817,
      "step": 863
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.00016888439568160288,
      "loss": 0.9058,
      "step": 864
    },
    {
      "epoch": 5.6,
      "learning_rate": 0.0001688081148496389,
      "loss": 0.8242,
      "step": 865
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.00016873175790290836,
      "loss": 0.7173,
      "step": 866
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.0001686553249258765,
      "loss": 0.8198,
      "step": 867
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.00016857881600309255,
      "loss": 0.7495,
      "step": 868
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.00016850223121918988,
      "loss": 0.6855,
      "step": 869
    },
    {
      "epoch": 5.63,
      "learning_rate": 0.00016842557065888575,
      "loss": 0.8013,
      "step": 870
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.0001683488344069811,
      "loss": 0.7734,
      "step": 871
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.00016827202254836085,
      "loss": 0.7559,
      "step": 872
    },
    {
      "epoch": 5.65,
      "learning_rate": 0.00016819513516799333,
      "loss": 0.792,
      "step": 873
    },
    {
      "epoch": 5.66,
      "learning_rate": 0.00016811817235093052,
      "loss": 0.7769,
      "step": 874
    },
    {
      "epoch": 5.66,
      "learning_rate": 0.00016804113418230786,
      "loss": 0.8833,
      "step": 875
    },
    {
      "epoch": 5.67,
      "learning_rate": 0.00016796402074734402,
      "loss": 0.8423,
      "step": 876
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.00016788683213134114,
      "loss": 0.8418,
      "step": 877
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.0001678095684196843,
      "loss": 0.7319,
      "step": 878
    },
    {
      "epoch": 5.69,
      "learning_rate": 0.00016773222969784186,
      "loss": 0.7349,
      "step": 879
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.0001676548160513649,
      "loss": 0.7358,
      "step": 880
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.00016757732756588766,
      "loss": 0.9116,
      "step": 881
    },
    {
      "epoch": 5.71,
      "learning_rate": 0.000167499764327127,
      "loss": 0.7007,
      "step": 882
    },
    {
      "epoch": 5.72,
      "learning_rate": 0.0001674221264208825,
      "loss": 0.7153,
      "step": 883
    },
    {
      "epoch": 5.72,
      "learning_rate": 0.0001673444139330363,
      "loss": 0.9238,
      "step": 884
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.00016726662694955313,
      "loss": 0.7832,
      "step": 885
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.00016718876555648008,
      "loss": 0.8252,
      "step": 886
    },
    {
      "epoch": 5.74,
      "learning_rate": 0.00016711082983994656,
      "loss": 0.7427,
      "step": 887
    },
    {
      "epoch": 5.75,
      "learning_rate": 0.00016703281988616415,
      "loss": 0.79,
      "step": 888
    },
    {
      "epoch": 5.75,
      "learning_rate": 0.00016695473578142665,
      "loss": 0.7693,
      "step": 889
    },
    {
      "epoch": 5.76,
      "learning_rate": 0.00016687657761210975,
      "loss": 0.9307,
      "step": 890
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.00016679834546467118,
      "loss": 0.8696,
      "step": 891
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.00016672003942565044,
      "loss": 0.7778,
      "step": 892
    },
    {
      "epoch": 5.78,
      "learning_rate": 0.00016664165958166883,
      "loss": 0.6453,
      "step": 893
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.00016656320601942927,
      "loss": 0.769,
      "step": 894
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.00016648467882571618,
      "loss": 0.7715,
      "step": 895
    },
    {
      "epoch": 5.8,
      "learning_rate": 0.00016640607808739543,
      "loss": 0.7749,
      "step": 896
    },
    {
      "epoch": 5.81,
      "learning_rate": 0.0001663274038914143,
      "loss": 0.7476,
      "step": 897
    },
    {
      "epoch": 5.81,
      "learning_rate": 0.0001662486563248014,
      "loss": 0.8247,
      "step": 898
    },
    {
      "epoch": 5.82,
      "learning_rate": 0.00016616983547466626,
      "loss": 0.7749,
      "step": 899
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.0001660909414281997,
      "loss": 0.8474,
      "step": 900
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.00016601197427267342,
      "loss": 0.7549,
      "step": 901
    },
    {
      "epoch": 5.84,
      "learning_rate": 0.00016593293409544,
      "loss": 0.8691,
      "step": 902
    },
    {
      "epoch": 5.84,
      "learning_rate": 0.0001658538209839328,
      "loss": 0.8115,
      "step": 903
    },
    {
      "epoch": 5.85,
      "learning_rate": 0.00016577463502566588,
      "loss": 0.7563,
      "step": 904
    },
    {
      "epoch": 5.86,
      "learning_rate": 0.00016569537630823383,
      "loss": 0.793,
      "step": 905
    },
    {
      "epoch": 5.86,
      "learning_rate": 0.00016561604491931178,
      "loss": 0.8511,
      "step": 906
    },
    {
      "epoch": 5.87,
      "learning_rate": 0.0001655366409466552,
      "loss": 0.7539,
      "step": 907
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.00016545716447809988,
      "loss": 0.7803,
      "step": 908
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.0001653776156015618,
      "loss": 0.6904,
      "step": 909
    },
    {
      "epoch": 5.89,
      "learning_rate": 0.00016529799440503706,
      "loss": 0.8267,
      "step": 910
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.00016521830097660173,
      "loss": 0.7817,
      "step": 911
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.00016513853540441175,
      "loss": 0.8296,
      "step": 912
    },
    {
      "epoch": 5.91,
      "learning_rate": 0.00016505869777670297,
      "loss": 0.8472,
      "step": 913
    },
    {
      "epoch": 5.92,
      "learning_rate": 0.00016497878818179084,
      "loss": 0.7749,
      "step": 914
    },
    {
      "epoch": 5.92,
      "learning_rate": 0.00016489880670807043,
      "loss": 0.9438,
      "step": 915
    },
    {
      "epoch": 5.93,
      "learning_rate": 0.00016481875344401644,
      "loss": 0.7178,
      "step": 916
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.00016473862847818277,
      "loss": 0.8442,
      "step": 917
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.00016465843189920288,
      "loss": 0.7373,
      "step": 918
    },
    {
      "epoch": 5.95,
      "learning_rate": 0.00016457816379578926,
      "loss": 0.6606,
      "step": 919
    },
    {
      "epoch": 5.95,
      "learning_rate": 0.00016449782425673358,
      "loss": 0.833,
      "step": 920
    },
    {
      "epoch": 5.96,
      "learning_rate": 0.0001644174133709066,
      "loss": 0.8701,
      "step": 921
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.0001643369312272579,
      "loss": 0.8599,
      "step": 922
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.00016425637791481592,
      "loss": 0.7314,
      "step": 923
    },
    {
      "epoch": 5.98,
      "learning_rate": 0.00016417575352268787,
      "loss": 0.8154,
      "step": 924
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.0001640950581400595,
      "loss": 0.8823,
      "step": 925
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.00016401429185619517,
      "loss": 0.7671,
      "step": 926
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.00016393345476043762,
      "loss": 0.8855,
      "step": 927
    },
    {
      "epoch": 6.01,
      "learning_rate": 0.000163852546942208,
      "loss": 0.6558,
      "step": 928
    },
    {
      "epoch": 6.01,
      "learning_rate": 0.00016377156849100555,
      "loss": 0.6982,
      "step": 929
    },
    {
      "epoch": 6.02,
      "learning_rate": 0.00016369051949640775,
      "loss": 0.6797,
      "step": 930
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.00016360940004807014,
      "loss": 0.6978,
      "step": 931
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.0001635282102357261,
      "loss": 0.6958,
      "step": 932
    },
    {
      "epoch": 6.04,
      "learning_rate": 0.00016344695014918688,
      "loss": 0.7964,
      "step": 933
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.00016336561987834153,
      "loss": 0.6035,
      "step": 934
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.00016328421951315665,
      "loss": 0.6514,
      "step": 935
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.0001632027491436764,
      "loss": 0.5264,
      "step": 936
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.00016312120886002242,
      "loss": 0.707,
      "step": 937
    },
    {
      "epoch": 6.07,
      "learning_rate": 0.00016303959875239366,
      "loss": 0.6216,
      "step": 938
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.00016295791891106628,
      "loss": 0.6614,
      "step": 939
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.00016287616942639365,
      "loss": 0.6636,
      "step": 940
    },
    {
      "epoch": 6.09,
      "learning_rate": 0.0001627943503888061,
      "loss": 0.6401,
      "step": 941
    },
    {
      "epoch": 6.1,
      "learning_rate": 0.00016271246188881091,
      "loss": 0.5952,
      "step": 942
    },
    {
      "epoch": 6.1,
      "learning_rate": 0.00016263050401699227,
      "loss": 0.6968,
      "step": 943
    },
    {
      "epoch": 6.11,
      "learning_rate": 0.00016254847686401105,
      "loss": 0.5569,
      "step": 944
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.00016246638052060477,
      "loss": 0.7478,
      "step": 945
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.00016238421507758748,
      "loss": 0.6587,
      "step": 946
    },
    {
      "epoch": 6.13,
      "learning_rate": 0.00016230198062584962,
      "loss": 0.6768,
      "step": 947
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.00016221967725635813,
      "loss": 0.6943,
      "step": 948
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.000162137305060156,
      "loss": 0.6787,
      "step": 949
    },
    {
      "epoch": 6.15,
      "learning_rate": 0.00016205486412836247,
      "loss": 0.6338,
      "step": 950
    },
    {
      "epoch": 6.16,
      "learning_rate": 0.00016197235455217275,
      "loss": 0.6533,
      "step": 951
    },
    {
      "epoch": 6.16,
      "learning_rate": 0.000161889776422858,
      "loss": 0.7402,
      "step": 952
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.00016180712983176524,
      "loss": 0.6733,
      "step": 953
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.0001617244148703172,
      "loss": 0.8081,
      "step": 954
    },
    {
      "epoch": 6.18,
      "learning_rate": 0.00016164163163001226,
      "loss": 0.6814,
      "step": 955
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.00016155878020242427,
      "loss": 0.6221,
      "step": 956
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.00016147586067920264,
      "loss": 0.5928,
      "step": 957
    },
    {
      "epoch": 6.2,
      "learning_rate": 0.00016139287315207192,
      "loss": 0.6135,
      "step": 958
    },
    {
      "epoch": 6.21,
      "learning_rate": 0.00016130981771283204,
      "loss": 0.7905,
      "step": 959
    },
    {
      "epoch": 6.21,
      "learning_rate": 0.00016122669445335798,
      "loss": 0.645,
      "step": 960
    },
    {
      "epoch": 6.22,
      "learning_rate": 0.00016114350346559976,
      "loss": 0.6338,
      "step": 961
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.00016106024484158232,
      "loss": 0.6873,
      "step": 962
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.00016097691867340545,
      "loss": 0.6338,
      "step": 963
    },
    {
      "epoch": 6.24,
      "learning_rate": 0.00016089352505324356,
      "loss": 0.7432,
      "step": 964
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.00016081006407334578,
      "loss": 0.6914,
      "step": 965
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.0001607265358260357,
      "loss": 0.6377,
      "step": 966
    },
    {
      "epoch": 6.26,
      "learning_rate": 0.00016064294040371135,
      "loss": 0.6797,
      "step": 967
    },
    {
      "epoch": 6.27,
      "learning_rate": 0.00016055927789884503,
      "loss": 0.6279,
      "step": 968
    },
    {
      "epoch": 6.27,
      "learning_rate": 0.00016047554840398326,
      "loss": 0.6973,
      "step": 969
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.00016039175201174673,
      "loss": 0.6152,
      "step": 970
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.00016030788881483,
      "loss": 0.6958,
      "step": 971
    },
    {
      "epoch": 6.29,
      "learning_rate": 0.00016022395890600161,
      "loss": 0.7866,
      "step": 972
    },
    {
      "epoch": 6.3,
      "learning_rate": 0.00016013996237810392,
      "loss": 0.7358,
      "step": 973
    },
    {
      "epoch": 6.3,
      "learning_rate": 0.00016005589932405293,
      "loss": 0.7212,
      "step": 974
    },
    {
      "epoch": 6.31,
      "learning_rate": 0.00015997176983683826,
      "loss": 0.7441,
      "step": 975
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.00015988757400952296,
      "loss": 0.7715,
      "step": 976
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.00015980331193524361,
      "loss": 0.748,
      "step": 977
    },
    {
      "epoch": 6.33,
      "learning_rate": 0.0001597189837072099,
      "loss": 0.6174,
      "step": 978
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.00015963458941870473,
      "loss": 0.665,
      "step": 979
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.0001595501291630842,
      "loss": 0.7964,
      "step": 980
    },
    {
      "epoch": 6.35,
      "learning_rate": 0.00015946560303377728,
      "loss": 0.6357,
      "step": 981
    },
    {
      "epoch": 6.36,
      "learning_rate": 0.00015938101112428582,
      "loss": 0.5706,
      "step": 982
    },
    {
      "epoch": 6.36,
      "learning_rate": 0.00015929635352818447,
      "loss": 0.6948,
      "step": 983
    },
    {
      "epoch": 6.37,
      "learning_rate": 0.00015921163033912045,
      "loss": 0.5872,
      "step": 984
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.00015912684165081365,
      "loss": 0.6284,
      "step": 985
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.0001590419875570564,
      "loss": 0.5869,
      "step": 986
    },
    {
      "epoch": 6.39,
      "learning_rate": 0.00015895706815171327,
      "loss": 0.7285,
      "step": 987
    },
    {
      "epoch": 6.39,
      "learning_rate": 0.0001588720835287212,
      "loss": 0.6909,
      "step": 988
    },
    {
      "epoch": 6.4,
      "learning_rate": 0.00015878703378208924,
      "loss": 0.7056,
      "step": 989
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.00015870191900589842,
      "loss": 0.7671,
      "step": 990
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.0001586167392943018,
      "loss": 0.6719,
      "step": 991
    },
    {
      "epoch": 6.42,
      "learning_rate": 0.00015853149474152423,
      "loss": 0.6729,
      "step": 992
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.00015844618544186222,
      "loss": 0.7544,
      "step": 993
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.000158360811489684,
      "loss": 0.7881,
      "step": 994
    },
    {
      "epoch": 6.44,
      "learning_rate": 0.00015827537297942934,
      "loss": 0.603,
      "step": 995
    },
    {
      "epoch": 6.45,
      "learning_rate": 0.00015818987000560926,
      "loss": 0.7031,
      "step": 996
    },
    {
      "epoch": 6.45,
      "learning_rate": 0.00015810430266280625,
      "loss": 0.7866,
      "step": 997
    },
    {
      "epoch": 6.46,
      "learning_rate": 0.00015801867104567395,
      "loss": 0.7827,
      "step": 998
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.00015793297524893703,
      "loss": 0.7134,
      "step": 999
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.00015784721536739134,
      "loss": 0.7349,
      "step": 1000
    },
    {
      "epoch": 6.48,
      "learning_rate": 0.0001577613914959034,
      "loss": 0.6836,
      "step": 1001
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.00015767550372941064,
      "loss": 0.7339,
      "step": 1002
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.00015758955216292115,
      "loss": 0.7437,
      "step": 1003
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.00015750353689151364,
      "loss": 0.7119,
      "step": 1004
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.00015741745801033718,
      "loss": 0.7651,
      "step": 1005
    },
    {
      "epoch": 6.51,
      "learning_rate": 0.00015733131561461127,
      "loss": 0.6426,
      "step": 1006
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.00015724510979962567,
      "loss": 0.6045,
      "step": 1007
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.0001571588406607403,
      "loss": 0.7388,
      "step": 1008
    },
    {
      "epoch": 6.53,
      "learning_rate": 0.00015707250829338513,
      "loss": 0.6941,
      "step": 1009
    },
    {
      "epoch": 6.54,
      "learning_rate": 0.00015698611279306004,
      "loss": 0.75,
      "step": 1010
    },
    {
      "epoch": 6.54,
      "learning_rate": 0.00015689965425533475,
      "loss": 0.6621,
      "step": 1011
    },
    {
      "epoch": 6.55,
      "learning_rate": 0.00015681313277584872,
      "loss": 0.7749,
      "step": 1012
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.0001567265484503111,
      "loss": 0.6572,
      "step": 1013
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.00015663990137450043,
      "loss": 0.7058,
      "step": 1014
    },
    {
      "epoch": 6.57,
      "learning_rate": 0.0001565531916442648,
      "loss": 0.802,
      "step": 1015
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.00015646641935552152,
      "loss": 0.6978,
      "step": 1016
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.0001563795846042571,
      "loss": 0.7471,
      "step": 1017
    },
    {
      "epoch": 6.59,
      "learning_rate": 0.00015629268748652717,
      "loss": 0.7192,
      "step": 1018
    },
    {
      "epoch": 6.6,
      "learning_rate": 0.0001562057280984564,
      "loss": 0.6304,
      "step": 1019
    },
    {
      "epoch": 6.6,
      "learning_rate": 0.00015611870653623825,
      "loss": 0.7036,
      "step": 1020
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.00015603162289613503,
      "loss": 0.7437,
      "step": 1021
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.00015594447727447763,
      "loss": 0.752,
      "step": 1022
    },
    {
      "epoch": 6.62,
      "learning_rate": 0.00015585726976766559,
      "loss": 0.832,
      "step": 1023
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.0001557700004721669,
      "loss": 0.6777,
      "step": 1024
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.00015568266948451788,
      "loss": 0.6724,
      "step": 1025
    },
    {
      "epoch": 6.64,
      "learning_rate": 0.00015559527690132304,
      "loss": 0.6816,
      "step": 1026
    },
    {
      "epoch": 6.65,
      "learning_rate": 0.00015550782281925515,
      "loss": 0.6318,
      "step": 1027
    },
    {
      "epoch": 6.65,
      "learning_rate": 0.0001554203073350549,
      "loss": 0.7153,
      "step": 1028
    },
    {
      "epoch": 6.66,
      "learning_rate": 0.0001553327305455309,
      "loss": 0.6689,
      "step": 1029
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.00015524509254755969,
      "loss": 0.7378,
      "step": 1030
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.00015515739343808534,
      "loss": 0.6782,
      "step": 1031
    },
    {
      "epoch": 6.68,
      "learning_rate": 0.00015506963331411976,
      "loss": 0.7583,
      "step": 1032
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.00015498181227274212,
      "loss": 0.665,
      "step": 1033
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.0001548939304110991,
      "loss": 0.6709,
      "step": 1034
    },
    {
      "epoch": 6.7,
      "learning_rate": 0.0001548059878264046,
      "loss": 0.7402,
      "step": 1035
    },
    {
      "epoch": 6.71,
      "learning_rate": 0.0001547179846159398,
      "loss": 0.7969,
      "step": 1036
    },
    {
      "epoch": 6.71,
      "learning_rate": 0.0001546299208770528,
      "loss": 0.7505,
      "step": 1037
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.00015454179670715872,
      "loss": 0.6655,
      "step": 1038
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.00015445361220373956,
      "loss": 0.665,
      "step": 1039
    },
    {
      "epoch": 6.73,
      "learning_rate": 0.00015436536746434406,
      "loss": 0.6904,
      "step": 1040
    },
    {
      "epoch": 6.74,
      "learning_rate": 0.00015427706258658747,
      "loss": 0.6067,
      "step": 1041
    },
    {
      "epoch": 6.74,
      "learning_rate": 0.00015418869766815177,
      "loss": 0.6504,
      "step": 1042
    },
    {
      "epoch": 6.75,
      "learning_rate": 0.00015410027280678517,
      "loss": 0.6624,
      "step": 1043
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.00015401178810030226,
      "loss": 0.6479,
      "step": 1044
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.00015392324364658385,
      "loss": 0.7769,
      "step": 1045
    },
    {
      "epoch": 6.77,
      "learning_rate": 0.00015383463954357683,
      "loss": 0.8486,
      "step": 1046
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.00015374597588929407,
      "loss": 0.7095,
      "step": 1047
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.0001536572527818143,
      "loss": 0.6182,
      "step": 1048
    },
    {
      "epoch": 6.79,
      "learning_rate": 0.00015356847031928203,
      "loss": 0.6743,
      "step": 1049
    },
    {
      "epoch": 6.8,
      "learning_rate": 0.00015347962859990744,
      "loss": 0.7603,
      "step": 1050
    },
    {
      "epoch": 6.8,
      "learning_rate": 0.0001533907277219662,
      "loss": 0.6812,
      "step": 1051
    },
    {
      "epoch": 6.81,
      "learning_rate": 0.00015330176778379954,
      "loss": 0.6689,
      "step": 1052
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.00015321274888381388,
      "loss": 0.6233,
      "step": 1053
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.00015312367112048098,
      "loss": 0.6899,
      "step": 1054
    },
    {
      "epoch": 6.83,
      "learning_rate": 0.00015303453459233763,
      "loss": 0.7456,
      "step": 1055
    },
    {
      "epoch": 6.83,
      "learning_rate": 0.0001529453393979857,
      "loss": 0.7559,
      "step": 1056
    },
    {
      "epoch": 6.84,
      "learning_rate": 0.00015285608563609187,
      "loss": 0.656,
      "step": 1057
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.00015276677340538767,
      "loss": 0.6807,
      "step": 1058
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.0001526774028046693,
      "loss": 0.729,
      "step": 1059
    },
    {
      "epoch": 6.86,
      "learning_rate": 0.0001525879739327975,
      "loss": 0.7607,
      "step": 1060
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.0001524984868886975,
      "loss": 0.8281,
      "step": 1061
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.00015240894177135888,
      "loss": 0.7593,
      "step": 1062
    },
    {
      "epoch": 6.88,
      "learning_rate": 0.00015231933867983541,
      "loss": 0.6987,
      "step": 1063
    },
    {
      "epoch": 6.89,
      "learning_rate": 0.00015222967771324504,
      "loss": 0.7141,
      "step": 1064
    },
    {
      "epoch": 6.89,
      "learning_rate": 0.00015213995897076977,
      "loss": 0.6394,
      "step": 1065
    },
    {
      "epoch": 6.9,
      "learning_rate": 0.00015205018255165542,
      "loss": 0.752,
      "step": 1066
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.00015196034855521169,
      "loss": 0.6816,
      "step": 1067
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.00015187045708081191,
      "loss": 0.6978,
      "step": 1068
    },
    {
      "epoch": 6.92,
      "learning_rate": 0.00015178050822789303,
      "loss": 0.7227,
      "step": 1069
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.00015169050209595545,
      "loss": 0.7417,
      "step": 1070
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.00015160043878456302,
      "loss": 0.874,
      "step": 1071
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.0001515103183933427,
      "loss": 0.769,
      "step": 1072
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.00015142014102198465,
      "loss": 0.7471,
      "step": 1073
    },
    {
      "epoch": 6.95,
      "learning_rate": 0.00015132990677024213,
      "loss": 0.7119,
      "step": 1074
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.00015123961573793122,
      "loss": 0.7334,
      "step": 1075
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.0001511492680249308,
      "loss": 0.7097,
      "step": 1076
    },
    {
      "epoch": 6.97,
      "learning_rate": 0.00015105886373118264,
      "loss": 0.7021,
      "step": 1077
    },
    {
      "epoch": 6.98,
      "learning_rate": 0.00015096840295669082,
      "loss": 0.7563,
      "step": 1078
    },
    {
      "epoch": 6.98,
      "learning_rate": 0.00015087788580152206,
      "loss": 0.6211,
      "step": 1079
    },
    {
      "epoch": 6.99,
      "learning_rate": 0.00015078731236580547,
      "loss": 0.6838,
      "step": 1080
    },
    {
      "epoch": 7.0,
      "learning_rate": 0.00015069668274973234,
      "loss": 0.6895,
      "step": 1081
    },
    {
      "epoch": 7.0,
      "learning_rate": 0.00015060599705355617,
      "loss": 0.6807,
      "step": 1082
    },
    {
      "epoch": 7.01,
      "learning_rate": 0.00015051525537759246,
      "loss": 0.6577,
      "step": 1083
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.00015042445782221858,
      "loss": 0.5938,
      "step": 1084
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.00015033360448787382,
      "loss": 0.5151,
      "step": 1085
    },
    {
      "epoch": 7.03,
      "learning_rate": 0.00015024269547505908,
      "loss": 0.5618,
      "step": 1086
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.00015015173088433695,
      "loss": 0.6309,
      "step": 1087
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.0001500607108163314,
      "loss": 0.5955,
      "step": 1088
    },
    {
      "epoch": 7.05,
      "learning_rate": 0.00014996963537172782,
      "loss": 0.6028,
      "step": 1089
    },
    {
      "epoch": 7.06,
      "learning_rate": 0.00014987850465127285,
      "loss": 0.6587,
      "step": 1090
    },
    {
      "epoch": 7.06,
      "learning_rate": 0.00014978731875577427,
      "loss": 0.5704,
      "step": 1091
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0001496960777861009,
      "loss": 0.6309,
      "step": 1092
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0001496047818431825,
      "loss": 0.519,
      "step": 1093
    },
    {
      "epoch": 7.08,
      "learning_rate": 0.00014951343102800955,
      "loss": 0.5908,
      "step": 1094
    },
    {
      "epoch": 7.09,
      "learning_rate": 0.0001494220254416334,
      "loss": 0.5813,
      "step": 1095
    },
    {
      "epoch": 7.09,
      "learning_rate": 0.00014933056518516582,
      "loss": 0.5261,
      "step": 1096
    },
    {
      "epoch": 7.1,
      "learning_rate": 0.00014923905035977913,
      "loss": 0.5679,
      "step": 1097
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.00014914748106670603,
      "loss": 0.5291,
      "step": 1098
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.0001490558574072394,
      "loss": 0.5123,
      "step": 1099
    },
    {
      "epoch": 7.12,
      "learning_rate": 0.0001489641794827324,
      "loss": 0.5632,
      "step": 1100
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.000148872447394598,
      "loss": 0.6006,
      "step": 1101
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.00014878066124430928,
      "loss": 0.5825,
      "step": 1102
    },
    {
      "epoch": 7.14,
      "learning_rate": 0.00014868882113339905,
      "loss": 0.5654,
      "step": 1103
    },
    {
      "epoch": 7.15,
      "learning_rate": 0.0001485969271634598,
      "loss": 0.6294,
      "step": 1104
    },
    {
      "epoch": 7.15,
      "learning_rate": 0.00014850497943614364,
      "loss": 0.6042,
      "step": 1105
    },
    {
      "epoch": 7.16,
      "learning_rate": 0.0001484129780531621,
      "loss": 0.5662,
      "step": 1106
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.00014832092311628604,
      "loss": 0.6458,
      "step": 1107
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.00014822881472734562,
      "loss": 0.7178,
      "step": 1108
    },
    {
      "epoch": 7.18,
      "learning_rate": 0.00014813665298823016,
      "loss": 0.7205,
      "step": 1109
    },
    {
      "epoch": 7.18,
      "learning_rate": 0.0001480444380008879,
      "loss": 0.5903,
      "step": 1110
    },
    {
      "epoch": 7.19,
      "learning_rate": 0.000147952169867326,
      "loss": 0.5598,
      "step": 1111
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.00014785984868961053,
      "loss": 0.5898,
      "step": 1112
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.00014776747456986603,
      "loss": 0.6277,
      "step": 1113
    },
    {
      "epoch": 7.21,
      "learning_rate": 0.00014767504761027583,
      "loss": 0.6216,
      "step": 1114
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0001475825679130815,
      "loss": 0.5938,
      "step": 1115
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0001474900355805831,
      "loss": 0.585,
      "step": 1116
    },
    {
      "epoch": 7.23,
      "learning_rate": 0.0001473974507151388,
      "loss": 0.5562,
      "step": 1117
    },
    {
      "epoch": 7.24,
      "learning_rate": 0.00014730481341916507,
      "loss": 0.7104,
      "step": 1118
    },
    {
      "epoch": 7.24,
      "learning_rate": 0.00014721212379513612,
      "loss": 0.6003,
      "step": 1119
    },
    {
      "epoch": 7.25,
      "learning_rate": 0.0001471193819455842,
      "loss": 0.7344,
      "step": 1120
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.00014702658797309938,
      "loss": 0.646,
      "step": 1121
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.00014693374198032925,
      "loss": 0.5581,
      "step": 1122
    },
    {
      "epoch": 7.27,
      "learning_rate": 0.00014684084406997903,
      "loss": 0.5359,
      "step": 1123
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.00014674789434481134,
      "loss": 0.5566,
      "step": 1124
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.0001466548929076461,
      "loss": 0.6008,
      "step": 1125
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.0001465618398613605,
      "loss": 0.6089,
      "step": 1126
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.00014646873530888875,
      "loss": 0.7446,
      "step": 1127
    },
    {
      "epoch": 7.3,
      "learning_rate": 0.0001463755793532221,
      "loss": 0.634,
      "step": 1128
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.0001462823720974086,
      "loss": 0.6792,
      "step": 1129
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.00014618911364455304,
      "loss": 0.6543,
      "step": 1130
    },
    {
      "epoch": 7.32,
      "learning_rate": 0.00014609580409781698,
      "loss": 0.6309,
      "step": 1131
    },
    {
      "epoch": 7.33,
      "learning_rate": 0.00014600244356041827,
      "loss": 0.5679,
      "step": 1132
    },
    {
      "epoch": 7.33,
      "learning_rate": 0.0001459090321356314,
      "loss": 0.5776,
      "step": 1133
    },
    {
      "epoch": 7.34,
      "learning_rate": 0.00014581556992678696,
      "loss": 0.5651,
      "step": 1134
    },
    {
      "epoch": 7.35,
      "learning_rate": 0.00014572205703727184,
      "loss": 0.6025,
      "step": 1135
    },
    {
      "epoch": 7.35,
      "learning_rate": 0.00014562849357052893,
      "loss": 0.6602,
      "step": 1136
    },
    {
      "epoch": 7.36,
      "learning_rate": 0.0001455348796300571,
      "loss": 0.6423,
      "step": 1137
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.0001454412153194111,
      "loss": 0.6895,
      "step": 1138
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.00014534750074220125,
      "loss": 0.5964,
      "step": 1139
    },
    {
      "epoch": 7.38,
      "learning_rate": 0.0001452537360020936,
      "loss": 0.5327,
      "step": 1140
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.00014515992120280968,
      "loss": 0.6194,
      "step": 1141
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.00014506605644812635,
      "loss": 0.7134,
      "step": 1142
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.00014497214184187572,
      "loss": 0.5647,
      "step": 1143
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.00014487817748794516,
      "loss": 0.6831,
      "step": 1144
    },
    {
      "epoch": 7.41,
      "learning_rate": 0.00014478416349027691,
      "loss": 0.6089,
      "step": 1145
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.00014469009995286825,
      "loss": 0.6487,
      "step": 1146
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.00014459598697977118,
      "loss": 0.5457,
      "step": 1147
    },
    {
      "epoch": 7.43,
      "learning_rate": 0.00014450182467509243,
      "loss": 0.5671,
      "step": 1148
    },
    {
      "epoch": 7.44,
      "learning_rate": 0.00014440761314299326,
      "loss": 0.5229,
      "step": 1149
    },
    {
      "epoch": 7.44,
      "learning_rate": 0.00014431335248768948,
      "loss": 0.6982,
      "step": 1150
    },
    {
      "epoch": 7.45,
      "learning_rate": 0.00014421904281345114,
      "loss": 0.5947,
      "step": 1151
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.0001441246842246025,
      "loss": 0.604,
      "step": 1152
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.00014403027682552205,
      "loss": 0.6106,
      "step": 1153
    },
    {
      "epoch": 7.47,
      "learning_rate": 0.00014393582072064215,
      "loss": 0.6763,
      "step": 1154
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.00014384131601444903,
      "loss": 0.5239,
      "step": 1155
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.0001437467628114829,
      "loss": 0.623,
      "step": 1156
    },
    {
      "epoch": 7.49,
      "learning_rate": 0.0001436521612163373,
      "loss": 0.6021,
      "step": 1157
    },
    {
      "epoch": 7.5,
      "learning_rate": 0.00014355751133365954,
      "loss": 0.6763,
      "step": 1158
    },
    {
      "epoch": 7.5,
      "learning_rate": 0.00014346281326815024,
      "loss": 0.6372,
      "step": 1159
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.00014336806712456329,
      "loss": 0.6841,
      "step": 1160
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.00014327327300770593,
      "loss": 0.6743,
      "step": 1161
    },
    {
      "epoch": 7.52,
      "learning_rate": 0.00014317843102243824,
      "loss": 0.7559,
      "step": 1162
    },
    {
      "epoch": 7.53,
      "learning_rate": 0.0001430835412736734,
      "loss": 0.7075,
      "step": 1163
    },
    {
      "epoch": 7.53,
      "learning_rate": 0.0001429886038663774,
      "loss": 0.6804,
      "step": 1164
    },
    {
      "epoch": 7.54,
      "learning_rate": 0.00014289361890556894,
      "loss": 0.5276,
      "step": 1165
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.0001427985864963193,
      "loss": 0.5771,
      "step": 1166
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.00014270350674375229,
      "loss": 0.6516,
      "step": 1167
    },
    {
      "epoch": 7.56,
      "learning_rate": 0.00014260837975304405,
      "loss": 0.678,
      "step": 1168
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.00014251320562942298,
      "loss": 0.6875,
      "step": 1169
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.00014241798447816968,
      "loss": 0.6055,
      "step": 1170
    },
    {
      "epoch": 7.58,
      "learning_rate": 0.00014232271640461668,
      "loss": 0.7056,
      "step": 1171
    },
    {
      "epoch": 7.59,
      "learning_rate": 0.00014222740151414843,
      "loss": 0.5781,
      "step": 1172
    },
    {
      "epoch": 7.59,
      "learning_rate": 0.0001421320399122013,
      "loss": 0.658,
      "step": 1173
    },
    {
      "epoch": 7.6,
      "learning_rate": 0.00014203663170426312,
      "loss": 0.6433,
      "step": 1174
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.0001419411769958734,
      "loss": 0.5247,
      "step": 1175
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.00014184567589262313,
      "loss": 0.6077,
      "step": 1176
    },
    {
      "epoch": 7.62,
      "learning_rate": 0.0001417501285001545,
      "loss": 0.5676,
      "step": 1177
    },
    {
      "epoch": 7.62,
      "learning_rate": 0.00014165453492416102,
      "loss": 0.7454,
      "step": 1178
    },
    {
      "epoch": 7.63,
      "learning_rate": 0.00014155889527038716,
      "loss": 0.606,
      "step": 1179
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.0001414632096446285,
      "loss": 0.6152,
      "step": 1180
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.00014136747815273137,
      "loss": 0.6838,
      "step": 1181
    },
    {
      "epoch": 7.65,
      "learning_rate": 0.0001412717009005929,
      "loss": 0.6143,
      "step": 1182
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0001411758779941608,
      "loss": 0.583,
      "step": 1183
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0001410800095394333,
      "loss": 0.623,
      "step": 1184
    },
    {
      "epoch": 7.67,
      "learning_rate": 0.00014098409564245904,
      "loss": 0.6694,
      "step": 1185
    },
    {
      "epoch": 7.68,
      "learning_rate": 0.00014088813640933688,
      "loss": 0.6985,
      "step": 1186
    },
    {
      "epoch": 7.68,
      "learning_rate": 0.0001407921319462158,
      "loss": 0.6782,
      "step": 1187
    },
    {
      "epoch": 7.69,
      "learning_rate": 0.00014069608235929492,
      "loss": 0.6079,
      "step": 1188
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.0001405999877548232,
      "loss": 0.5486,
      "step": 1189
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.00014050384823909942,
      "loss": 0.5693,
      "step": 1190
    },
    {
      "epoch": 7.71,
      "learning_rate": 0.000140407663918472,
      "loss": 0.7603,
      "step": 1191
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.00014031143489933907,
      "loss": 0.6355,
      "step": 1192
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.00014021516128814797,
      "loss": 0.6484,
      "step": 1193
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.00014011884319139556,
      "loss": 0.6743,
      "step": 1194
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.0001400224807156278,
      "loss": 0.6343,
      "step": 1195
    },
    {
      "epoch": 7.74,
      "learning_rate": 0.00013992607396743983,
      "loss": 0.6001,
      "step": 1196
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.00013982962305347565,
      "loss": 0.6438,
      "step": 1197
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.00013973312808042825,
      "loss": 0.6689,
      "step": 1198
    },
    {
      "epoch": 7.76,
      "learning_rate": 0.0001396365891550392,
      "loss": 0.7773,
      "step": 1199
    },
    {
      "epoch": 7.77,
      "learning_rate": 0.00013954000638409893,
      "loss": 0.7646,
      "step": 1200
    },
    {
      "epoch": 7.77,
      "learning_rate": 0.00013944337987444608,
      "loss": 0.7017,
      "step": 1201
    },
    {
      "epoch": 7.78,
      "learning_rate": 0.00013934670973296786,
      "loss": 0.584,
      "step": 1202
    },
    {
      "epoch": 7.79,
      "learning_rate": 0.0001392499960665997,
      "loss": 0.5649,
      "step": 1203
    },
    {
      "epoch": 7.79,
      "learning_rate": 0.00013915323898232518,
      "loss": 0.8218,
      "step": 1204
    },
    {
      "epoch": 7.8,
      "learning_rate": 0.0001390564385871759,
      "loss": 0.658,
      "step": 1205
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.00013895959498823137,
      "loss": 0.6724,
      "step": 1206
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.0001388627082926189,
      "loss": 0.7036,
      "step": 1207
    },
    {
      "epoch": 7.82,
      "learning_rate": 0.00013876577860751345,
      "loss": 0.6465,
      "step": 1208
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.0001386688060401376,
      "loss": 0.6851,
      "step": 1209
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.00013857179069776127,
      "loss": 0.6499,
      "step": 1210
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.0001384747326877018,
      "loss": 0.6846,
      "step": 1211
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.00013837763211732357,
      "loss": 0.6355,
      "step": 1212
    },
    {
      "epoch": 7.85,
      "learning_rate": 0.00013828048909403826,
      "loss": 0.6577,
      "step": 1213
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.00013818330372530435,
      "loss": 0.6487,
      "step": 1214
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.00013808607611862715,
      "loss": 0.6006,
      "step": 1215
    },
    {
      "epoch": 7.87,
      "learning_rate": 0.00013798880638155888,
      "loss": 0.6572,
      "step": 1216
    },
    {
      "epoch": 7.88,
      "learning_rate": 0.00013789149462169813,
      "loss": 0.6445,
      "step": 1217
    },
    {
      "epoch": 7.88,
      "learning_rate": 0.00013779414094669009,
      "loss": 0.658,
      "step": 1218
    },
    {
      "epoch": 7.89,
      "learning_rate": 0.00013769674546422635,
      "loss": 0.6602,
      "step": 1219
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.0001375993082820447,
      "loss": 0.7251,
      "step": 1220
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.00013750182950792895,
      "loss": 0.6465,
      "step": 1221
    },
    {
      "epoch": 7.91,
      "learning_rate": 0.00013740430924970916,
      "loss": 0.6494,
      "step": 1222
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.00013730674761526114,
      "loss": 0.5669,
      "step": 1223
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.00013720914471250644,
      "loss": 0.6692,
      "step": 1224
    },
    {
      "epoch": 7.93,
      "learning_rate": 0.00013711150064941226,
      "loss": 0.6274,
      "step": 1225
    },
    {
      "epoch": 7.94,
      "learning_rate": 0.00013701381553399145,
      "loss": 0.5967,
      "step": 1226
    },
    {
      "epoch": 7.94,
      "learning_rate": 0.00013691608947430216,
      "loss": 0.6948,
      "step": 1227
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.0001368183225784478,
      "loss": 0.7412,
      "step": 1228
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.00013672051495457711,
      "loss": 0.6948,
      "step": 1229
    },
    {
      "epoch": 7.96,
      "learning_rate": 0.00013662266671088375,
      "loss": 0.6821,
      "step": 1230
    },
    {
      "epoch": 7.97,
      "learning_rate": 0.00013652477795560633,
      "loss": 0.6763,
      "step": 1231
    },
    {
      "epoch": 7.97,
      "learning_rate": 0.0001364268487970284,
      "loss": 0.667,
      "step": 1232
    },
    {
      "epoch": 7.98,
      "learning_rate": 0.00013632887934347796,
      "loss": 0.7236,
      "step": 1233
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.00013623086970332776,
      "loss": 0.6377,
      "step": 1234
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.00013613281998499504,
      "loss": 0.5906,
      "step": 1235
    },
    {
      "epoch": 8.0,
      "learning_rate": 0.00013603473029694119,
      "loss": 0.6475,
      "step": 1236
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.000135936600747672,
      "loss": 0.4602,
      "step": 1237
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.0001358384314457373,
      "loss": 0.5415,
      "step": 1238
    },
    {
      "epoch": 8.02,
      "learning_rate": 0.0001357402224997308,
      "loss": 0.624,
      "step": 1239
    },
    {
      "epoch": 8.03,
      "learning_rate": 0.0001356419740182902,
      "loss": 0.4651,
      "step": 1240
    },
    {
      "epoch": 8.03,
      "learning_rate": 0.0001355436861100968,
      "loss": 0.6274,
      "step": 1241
    },
    {
      "epoch": 8.04,
      "learning_rate": 0.0001354453588838757,
      "loss": 0.6096,
      "step": 1242
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.00013534699244839525,
      "loss": 0.4961,
      "step": 1243
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.00013524858691246741,
      "loss": 0.4741,
      "step": 1244
    },
    {
      "epoch": 8.06,
      "learning_rate": 0.00013515014238494715,
      "loss": 0.4786,
      "step": 1245
    },
    {
      "epoch": 8.06,
      "learning_rate": 0.00013505165897473282,
      "loss": 0.5271,
      "step": 1246
    },
    {
      "epoch": 8.07,
      "learning_rate": 0.00013495313679076567,
      "loss": 0.4187,
      "step": 1247
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.00013485457594202972,
      "loss": 0.5347,
      "step": 1248
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.00013475597653755204,
      "loss": 0.5432,
      "step": 1249
    },
    {
      "epoch": 8.09,
      "learning_rate": 0.00013465733868640205,
      "loss": 0.5574,
      "step": 1250
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.00013455866249769186,
      "loss": 0.5161,
      "step": 1251
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.00013445994808057604,
      "loss": 0.6206,
      "step": 1252
    },
    {
      "epoch": 8.11,
      "learning_rate": 0.00013436119554425133,
      "loss": 0.4917,
      "step": 1253
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.00013426240499795665,
      "loss": 0.5449,
      "step": 1254
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.00013416357655097306,
      "loss": 0.5146,
      "step": 1255
    },
    {
      "epoch": 8.13,
      "learning_rate": 0.0001340647103126234,
      "loss": 0.4771,
      "step": 1256
    },
    {
      "epoch": 8.14,
      "learning_rate": 0.00013396580639227248,
      "loss": 0.5251,
      "step": 1257
    },
    {
      "epoch": 8.14,
      "learning_rate": 0.00013386686489932665,
      "loss": 0.5312,
      "step": 1258
    },
    {
      "epoch": 8.15,
      "learning_rate": 0.00013376788594323392,
      "loss": 0.5063,
      "step": 1259
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.00013366886963348367,
      "loss": 0.5977,
      "step": 1260
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.00013356981607960666,
      "loss": 0.3987,
      "step": 1261
    },
    {
      "epoch": 8.17,
      "learning_rate": 0.00013347072539117482,
      "loss": 0.5603,
      "step": 1262
    },
    {
      "epoch": 8.17,
      "learning_rate": 0.00013337159767780115,
      "loss": 0.4639,
      "step": 1263
    },
    {
      "epoch": 8.18,
      "learning_rate": 0.00013327243304913963,
      "loss": 0.5767,
      "step": 1264
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.00013317323161488506,
      "loss": 0.5198,
      "step": 1265
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.00013307399348477293,
      "loss": 0.5476,
      "step": 1266
    },
    {
      "epoch": 8.2,
      "learning_rate": 0.00013297471876857938,
      "loss": 0.5371,
      "step": 1267
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.00013287540757612098,
      "loss": 0.5854,
      "step": 1268
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.00013277606001725465,
      "loss": 0.522,
      "step": 1269
    },
    {
      "epoch": 8.22,
      "learning_rate": 0.0001326766762018776,
      "loss": 0.5403,
      "step": 1270
    },
    {
      "epoch": 8.23,
      "learning_rate": 0.000132577256239927,
      "loss": 0.4895,
      "step": 1271
    },
    {
      "epoch": 8.23,
      "learning_rate": 0.00013247780024138015,
      "loss": 0.4391,
      "step": 1272
    },
    {
      "epoch": 8.24,
      "learning_rate": 0.0001323783083162542,
      "loss": 0.5613,
      "step": 1273
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.00013227878057460595,
      "loss": 0.5386,
      "step": 1274
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.00013217921712653188,
      "loss": 0.6116,
      "step": 1275
    },
    {
      "epoch": 8.26,
      "learning_rate": 0.000132079618082168,
      "loss": 0.5925,
      "step": 1276
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.0001319799835516896,
      "loss": 0.5962,
      "step": 1277
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.00013188031364531132,
      "loss": 0.5486,
      "step": 1278
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.0001317806084732869,
      "loss": 0.5159,
      "step": 1279
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.00013168086814590904,
      "loss": 0.4626,
      "step": 1280
    },
    {
      "epoch": 8.29,
      "learning_rate": 0.00013158109277350947,
      "loss": 0.6172,
      "step": 1281
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0001314812824664585,
      "loss": 0.4951,
      "step": 1282
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0001313814373351652,
      "loss": 0.5374,
      "step": 1283
    },
    {
      "epoch": 8.31,
      "learning_rate": 0.00013128155749007712,
      "loss": 0.5596,
      "step": 1284
    },
    {
      "epoch": 8.32,
      "learning_rate": 0.00013118164304168028,
      "loss": 0.5933,
      "step": 1285
    },
    {
      "epoch": 8.32,
      "learning_rate": 0.00013108169410049885,
      "loss": 0.4988,
      "step": 1286
    },
    {
      "epoch": 8.33,
      "learning_rate": 0.00013098171077709534,
      "loss": 0.5813,
      "step": 1287
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.00013088169318207005,
      "loss": 0.5918,
      "step": 1288
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.00013078164142606142,
      "loss": 0.5652,
      "step": 1289
    },
    {
      "epoch": 8.35,
      "learning_rate": 0.00013068155561974556,
      "loss": 0.6313,
      "step": 1290
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0001305814358738362,
      "loss": 0.5024,
      "step": 1291
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0001304812822990848,
      "loss": 0.5632,
      "step": 1292
    },
    {
      "epoch": 8.37,
      "learning_rate": 0.00013038109500628008,
      "loss": 0.4773,
      "step": 1293
    },
    {
      "epoch": 8.38,
      "learning_rate": 0.00013028087410624803,
      "loss": 0.6152,
      "step": 1294
    },
    {
      "epoch": 8.38,
      "learning_rate": 0.00013018061970985196,
      "loss": 0.5488,
      "step": 1295
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.00013008033192799217,
      "loss": 0.5786,
      "step": 1296
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.00012998001087160586,
      "loss": 0.5762,
      "step": 1297
    },
    {
      "epoch": 8.4,
      "learning_rate": 0.00012987965665166705,
      "loss": 0.5132,
      "step": 1298
    },
    {
      "epoch": 8.41,
      "learning_rate": 0.00012977926937918647,
      "loss": 0.5398,
      "step": 1299
    },
    {
      "epoch": 8.41,
      "learning_rate": 0.00012967884916521136,
      "loss": 0.5933,
      "step": 1300
    },
    {
      "epoch": 8.42,
      "learning_rate": 0.0001295783961208255,
      "loss": 0.5808,
      "step": 1301
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.0001294779103571489,
      "loss": 0.5715,
      "step": 1302
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.00012937739198533773,
      "loss": 0.5256,
      "step": 1303
    },
    {
      "epoch": 8.44,
      "learning_rate": 0.00012927684111658437,
      "loss": 0.6333,
      "step": 1304
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.00012917625786211702,
      "loss": 0.5632,
      "step": 1305
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.0001290756423331998,
      "loss": 0.6436,
      "step": 1306
    },
    {
      "epoch": 8.46,
      "learning_rate": 0.00012897499464113238,
      "loss": 0.552,
      "step": 1307
    },
    {
      "epoch": 8.47,
      "learning_rate": 0.00012887431489725024,
      "loss": 0.6096,
      "step": 1308
    },
    {
      "epoch": 8.47,
      "learning_rate": 0.00012877360321292408,
      "loss": 0.5942,
      "step": 1309
    },
    {
      "epoch": 8.48,
      "learning_rate": 0.00012867285969956008,
      "loss": 0.6113,
      "step": 1310
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.0001285720844685996,
      "loss": 0.5303,
      "step": 1311
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.00012847127763151902,
      "loss": 0.5623,
      "step": 1312
    },
    {
      "epoch": 8.5,
      "learning_rate": 0.00012837043929982985,
      "loss": 0.6038,
      "step": 1313
    },
    {
      "epoch": 8.5,
      "learning_rate": 0.0001282695695850782,
      "loss": 0.4351,
      "step": 1314
    },
    {
      "epoch": 8.51,
      "learning_rate": 0.00012816866859884505,
      "loss": 0.5254,
      "step": 1315
    },
    {
      "epoch": 8.52,
      "learning_rate": 0.00012806773645274597,
      "loss": 0.6548,
      "step": 1316
    },
    {
      "epoch": 8.52,
      "learning_rate": 0.00012796677325843097,
      "loss": 0.5381,
      "step": 1317
    },
    {
      "epoch": 8.53,
      "learning_rate": 0.0001278657791275844,
      "loss": 0.5811,
      "step": 1318
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.00012776475417192483,
      "loss": 0.5906,
      "step": 1319
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.00012766369850320493,
      "loss": 0.5444,
      "step": 1320
    },
    {
      "epoch": 8.55,
      "learning_rate": 0.00012756261223321138,
      "loss": 0.5359,
      "step": 1321
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.00012746149547376463,
      "loss": 0.5933,
      "step": 1322
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.00012736034833671893,
      "loss": 0.6631,
      "step": 1323
    },
    {
      "epoch": 8.57,
      "learning_rate": 0.00012725917093396212,
      "loss": 0.6104,
      "step": 1324
    },
    {
      "epoch": 8.58,
      "learning_rate": 0.0001271579633774155,
      "loss": 0.5959,
      "step": 1325
    },
    {
      "epoch": 8.58,
      "learning_rate": 0.00012705672577903375,
      "loss": 0.5596,
      "step": 1326
    },
    {
      "epoch": 8.59,
      "learning_rate": 0.00012695545825080475,
      "loss": 0.626,
      "step": 1327
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.00012685416090474954,
      "loss": 0.6072,
      "step": 1328
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.00012675283385292212,
      "loss": 0.5339,
      "step": 1329
    },
    {
      "epoch": 8.61,
      "learning_rate": 0.00012665147720740925,
      "loss": 0.6199,
      "step": 1330
    },
    {
      "epoch": 8.61,
      "learning_rate": 0.0001265500910803306,
      "loss": 0.5325,
      "step": 1331
    },
    {
      "epoch": 8.62,
      "learning_rate": 0.00012644867558383834,
      "loss": 0.6194,
      "step": 1332
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.00012634723083011713,
      "loss": 0.5986,
      "step": 1333
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.0001262457569313841,
      "loss": 0.5862,
      "step": 1334
    },
    {
      "epoch": 8.64,
      "learning_rate": 0.00012614425399988844,
      "loss": 0.6035,
      "step": 1335
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.00012604272214791168,
      "loss": 0.6094,
      "step": 1336
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.0001259411614877671,
      "loss": 0.5764,
      "step": 1337
    },
    {
      "epoch": 8.66,
      "learning_rate": 0.00012583957213180004,
      "loss": 0.6235,
      "step": 1338
    },
    {
      "epoch": 8.67,
      "learning_rate": 0.00012573795419238746,
      "loss": 0.5249,
      "step": 1339
    },
    {
      "epoch": 8.67,
      "learning_rate": 0.00012563630778193805,
      "loss": 0.6562,
      "step": 1340
    },
    {
      "epoch": 8.68,
      "learning_rate": 0.00012553463301289184,
      "loss": 0.6045,
      "step": 1341
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.00012543292999772038,
      "loss": 0.6177,
      "step": 1342
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.00012533119884892643,
      "loss": 0.5605,
      "step": 1343
    },
    {
      "epoch": 8.7,
      "learning_rate": 0.0001252294396790438,
      "loss": 0.6431,
      "step": 1344
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0001251276526006374,
      "loss": 0.5752,
      "step": 1345
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0001250258377263029,
      "loss": 0.5806,
      "step": 1346
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.00012492399516866678,
      "loss": 0.627,
      "step": 1347
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.00012482212504038616,
      "loss": 0.6062,
      "step": 1348
    },
    {
      "epoch": 8.73,
      "learning_rate": 0.00012472022745414865,
      "loss": 0.6802,
      "step": 1349
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.00012461830252267215,
      "loss": 0.5012,
      "step": 1350
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.00012451635035870494,
      "loss": 0.4924,
      "step": 1351
    },
    {
      "epoch": 8.75,
      "learning_rate": 0.00012441437107502538,
      "loss": 0.5222,
      "step": 1352
    },
    {
      "epoch": 8.76,
      "learning_rate": 0.00012431236478444176,
      "loss": 0.5918,
      "step": 1353
    },
    {
      "epoch": 8.76,
      "learning_rate": 0.0001242103315997923,
      "loss": 0.5864,
      "step": 1354
    },
    {
      "epoch": 8.77,
      "learning_rate": 0.00012410827163394504,
      "loss": 0.5957,
      "step": 1355
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.00012400618499979748,
      "loss": 0.6169,
      "step": 1356
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.00012390407181027676,
      "loss": 0.574,
      "step": 1357
    },
    {
      "epoch": 8.79,
      "learning_rate": 0.00012380193217833934,
      "loss": 0.5349,
      "step": 1358
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.0001236997662169709,
      "loss": 0.6147,
      "step": 1359
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.00012359757403918636,
      "loss": 0.5754,
      "step": 1360
    },
    {
      "epoch": 8.81,
      "learning_rate": 0.0001234953557580295,
      "loss": 0.635,
      "step": 1361
    },
    {
      "epoch": 8.82,
      "learning_rate": 0.00012339311148657305,
      "loss": 0.5278,
      "step": 1362
    },
    {
      "epoch": 8.82,
      "learning_rate": 0.00012329084133791852,
      "loss": 0.6072,
      "step": 1363
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.00012318854542519596,
      "loss": 0.5613,
      "step": 1364
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.00012308622386156395,
      "loss": 0.509,
      "step": 1365
    },
    {
      "epoch": 8.84,
      "learning_rate": 0.0001229838767602095,
      "loss": 0.6392,
      "step": 1366
    },
    {
      "epoch": 8.85,
      "learning_rate": 0.0001228815042343478,
      "loss": 0.7495,
      "step": 1367
    },
    {
      "epoch": 8.85,
      "learning_rate": 0.00012277910639722225,
      "loss": 0.6064,
      "step": 1368
    },
    {
      "epoch": 8.86,
      "learning_rate": 0.00012267668336210413,
      "loss": 0.5667,
      "step": 1369
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.00012257423524229263,
      "loss": 0.73,
      "step": 1370
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.0001224717621511148,
      "loss": 0.5471,
      "step": 1371
    },
    {
      "epoch": 8.88,
      "learning_rate": 0.00012236926420192517,
      "loss": 0.4932,
      "step": 1372
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0001222667415081058,
      "loss": 0.5593,
      "step": 1373
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0001221641941830662,
      "loss": 0.5542,
      "step": 1374
    },
    {
      "epoch": 8.9,
      "learning_rate": 0.0001220616223402431,
      "loss": 0.5635,
      "step": 1375
    },
    {
      "epoch": 8.91,
      "learning_rate": 0.00012195902609310023,
      "loss": 0.5533,
      "step": 1376
    },
    {
      "epoch": 8.91,
      "learning_rate": 0.0001218564055551285,
      "loss": 0.5525,
      "step": 1377
    },
    {
      "epoch": 8.92,
      "learning_rate": 0.00012175376083984554,
      "loss": 0.707,
      "step": 1378
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.00012165109206079579,
      "loss": 0.5703,
      "step": 1379
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.00012154839933155034,
      "loss": 0.5439,
      "step": 1380
    },
    {
      "epoch": 8.94,
      "learning_rate": 0.00012144568276570671,
      "loss": 0.5591,
      "step": 1381
    },
    {
      "epoch": 8.94,
      "learning_rate": 0.00012134294247688883,
      "loss": 0.6245,
      "step": 1382
    },
    {
      "epoch": 8.95,
      "learning_rate": 0.00012124017857874683,
      "loss": 0.5701,
      "step": 1383
    },
    {
      "epoch": 8.96,
      "learning_rate": 0.00012113739118495704,
      "loss": 0.5107,
      "step": 1384
    },
    {
      "epoch": 8.96,
      "learning_rate": 0.00012103458040922168,
      "loss": 0.5417,
      "step": 1385
    },
    {
      "epoch": 8.97,
      "learning_rate": 0.00012093174636526888,
      "loss": 0.6729,
      "step": 1386
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.00012082888916685254,
      "loss": 0.5786,
      "step": 1387
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.00012072600892775207,
      "loss": 0.626,
      "step": 1388
    },
    {
      "epoch": 8.99,
      "learning_rate": 0.00012062310576177251,
      "loss": 0.5444,
      "step": 1389
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.0001205201797827442,
      "loss": 0.6709,
      "step": 1390
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.00012041723110452263,
      "loss": 0.3982,
      "step": 1391
    },
    {
      "epoch": 9.01,
      "learning_rate": 0.00012031425984098857,
      "loss": 0.47,
      "step": 1392
    },
    {
      "epoch": 9.02,
      "learning_rate": 0.0001202112661060476,
      "loss": 0.4568,
      "step": 1393
    },
    {
      "epoch": 9.02,
      "learning_rate": 0.00012010825001363026,
      "loss": 0.4792,
      "step": 1394
    },
    {
      "epoch": 9.03,
      "learning_rate": 0.00012000521167769181,
      "loss": 0.4819,
      "step": 1395
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.00011990215121221214,
      "loss": 0.5122,
      "step": 1396
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.00011979906873119548,
      "loss": 0.6072,
      "step": 1397
    },
    {
      "epoch": 9.05,
      "learning_rate": 0.00011969596434867063,
      "loss": 0.4561,
      "step": 1398
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.00011959283817869042,
      "loss": 0.4897,
      "step": 1399
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.00011948969033533193,
      "loss": 0.4558,
      "step": 1400
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.00011938652093269611,
      "loss": 0.4709,
      "step": 1401
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.00011928333008490777,
      "loss": 0.488,
      "step": 1402
    },
    {
      "epoch": 9.08,
      "learning_rate": 0.00011918011790611554,
      "loss": 0.4719,
      "step": 1403
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.00011907688451049155,
      "loss": 0.363,
      "step": 1404
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.0001189736300122314,
      "loss": 0.5137,
      "step": 1405
    },
    {
      "epoch": 9.1,
      "learning_rate": 0.00011887035452555409,
      "loss": 0.4832,
      "step": 1406
    },
    {
      "epoch": 9.11,
      "learning_rate": 0.00011876705816470181,
      "loss": 0.4988,
      "step": 1407
    },
    {
      "epoch": 9.11,
      "learning_rate": 0.00011866374104393985,
      "loss": 0.4729,
      "step": 1408
    },
    {
      "epoch": 9.12,
      "learning_rate": 0.0001185604032775564,
      "loss": 0.4231,
      "step": 1409
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.00011845704497986259,
      "loss": 0.5122,
      "step": 1410
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.00011835366626519216,
      "loss": 0.5354,
      "step": 1411
    },
    {
      "epoch": 9.14,
      "learning_rate": 0.00011825026724790153,
      "loss": 0.5337,
      "step": 1412
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.00011814684804236955,
      "loss": 0.4626,
      "step": 1413
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.00011804340876299735,
      "loss": 0.5156,
      "step": 1414
    },
    {
      "epoch": 9.16,
      "learning_rate": 0.00011793994952420831,
      "loss": 0.4883,
      "step": 1415
    },
    {
      "epoch": 9.17,
      "learning_rate": 0.0001178364704404479,
      "loss": 0.4817,
      "step": 1416
    },
    {
      "epoch": 9.17,
      "learning_rate": 0.00011773297162618353,
      "loss": 0.4463,
      "step": 1417
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.0001176294531959044,
      "loss": 0.4556,
      "step": 1418
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.00011752591526412145,
      "loss": 0.4971,
      "step": 1419
    },
    {
      "epoch": 9.19,
      "learning_rate": 0.00011742235794536718,
      "loss": 0.4746,
      "step": 1420
    },
    {
      "epoch": 9.2,
      "learning_rate": 0.00011731878135419551,
      "loss": 0.5142,
      "step": 1421
    },
    {
      "epoch": 9.2,
      "learning_rate": 0.00011721518560518179,
      "loss": 0.4717,
      "step": 1422
    },
    {
      "epoch": 9.21,
      "learning_rate": 0.00011711157081292241,
      "loss": 0.5095,
      "step": 1423
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.0001170079370920349,
      "loss": 0.4822,
      "step": 1424
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.00011690428455715776,
      "loss": 0.5151,
      "step": 1425
    },
    {
      "epoch": 9.23,
      "learning_rate": 0.0001168006133229502,
      "loss": 0.5447,
      "step": 1426
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.00011669692350409223,
      "loss": 0.5796,
      "step": 1427
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.00011659321521528434,
      "loss": 0.457,
      "step": 1428
    },
    {
      "epoch": 9.25,
      "learning_rate": 0.00011648948857124748,
      "loss": 0.667,
      "step": 1429
    },
    {
      "epoch": 9.26,
      "learning_rate": 0.00011638574368672293,
      "loss": 0.5723,
      "step": 1430
    },
    {
      "epoch": 9.26,
      "learning_rate": 0.00011628198067647211,
      "loss": 0.4563,
      "step": 1431
    },
    {
      "epoch": 9.27,
      "learning_rate": 0.0001161781996552765,
      "loss": 0.3975,
      "step": 1432
    },
    {
      "epoch": 9.28,
      "learning_rate": 0.00011607440073793748,
      "loss": 0.4998,
      "step": 1433
    },
    {
      "epoch": 9.28,
      "learning_rate": 0.00011597058403927625,
      "loss": 0.4788,
      "step": 1434
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.00011586674967413372,
      "loss": 0.5032,
      "step": 1435
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.00011576289775737028,
      "loss": 0.4436,
      "step": 1436
    },
    {
      "epoch": 9.3,
      "learning_rate": 0.00011565902840386575,
      "loss": 0.5393,
      "step": 1437
    },
    {
      "epoch": 9.31,
      "learning_rate": 0.00011555514172851925,
      "loss": 0.4473,
      "step": 1438
    },
    {
      "epoch": 9.31,
      "learning_rate": 0.00011545123784624907,
      "loss": 0.5112,
      "step": 1439
    },
    {
      "epoch": 9.32,
      "learning_rate": 0.00011534731687199254,
      "loss": 0.5234,
      "step": 1440
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.0001152433789207058,
      "loss": 0.5811,
      "step": 1441
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.00011513942410736396,
      "loss": 0.4373,
      "step": 1442
    },
    {
      "epoch": 9.34,
      "learning_rate": 0.00011503545254696056,
      "loss": 0.6152,
      "step": 1443
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.00011493146435450784,
      "loss": 0.5244,
      "step": 1444
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.00011482745964503638,
      "loss": 0.5396,
      "step": 1445
    },
    {
      "epoch": 9.36,
      "learning_rate": 0.000114723438533595,
      "loss": 0.4409,
      "step": 1446
    },
    {
      "epoch": 9.37,
      "learning_rate": 0.00011461940113525069,
      "loss": 0.4766,
      "step": 1447
    },
    {
      "epoch": 9.37,
      "learning_rate": 0.00011451534756508848,
      "loss": 0.4509,
      "step": 1448
    },
    {
      "epoch": 9.38,
      "learning_rate": 0.00011441127793821125,
      "loss": 0.5242,
      "step": 1449
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.00011430719236973963,
      "loss": 0.4585,
      "step": 1450
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.00011420309097481198,
      "loss": 0.5083,
      "step": 1451
    },
    {
      "epoch": 9.4,
      "learning_rate": 0.00011409897386858402,
      "loss": 0.4915,
      "step": 1452
    },
    {
      "epoch": 9.4,
      "learning_rate": 0.00011399484116622896,
      "loss": 0.4854,
      "step": 1453
    },
    {
      "epoch": 9.41,
      "learning_rate": 0.00011389069298293726,
      "loss": 0.5569,
      "step": 1454
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.00011378652943391641,
      "loss": 0.5161,
      "step": 1455
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.00011368235063439103,
      "loss": 0.4773,
      "step": 1456
    },
    {
      "epoch": 9.43,
      "learning_rate": 0.0001135781566996025,
      "loss": 0.5195,
      "step": 1457
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.000113473947744809,
      "loss": 0.5981,
      "step": 1458
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.00011336972388528528,
      "loss": 0.5002,
      "step": 1459
    },
    {
      "epoch": 9.45,
      "learning_rate": 0.00011326548523632266,
      "loss": 0.583,
      "step": 1460
    },
    {
      "epoch": 9.46,
      "learning_rate": 0.00011316123191322872,
      "loss": 0.5923,
      "step": 1461
    },
    {
      "epoch": 9.46,
      "learning_rate": 0.00011305696403132732,
      "loss": 0.53,
      "step": 1462
    },
    {
      "epoch": 9.47,
      "learning_rate": 0.00011295268170595847,
      "loss": 0.5493,
      "step": 1463
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.0001128483850524781,
      "loss": 0.5842,
      "step": 1464
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.00011274407418625794,
      "loss": 0.5562,
      "step": 1465
    },
    {
      "epoch": 9.49,
      "learning_rate": 0.00011263974922268559,
      "loss": 0.5173,
      "step": 1466
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.00011253541027716408,
      "loss": 0.4521,
      "step": 1467
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.00011243105746511202,
      "loss": 0.4935,
      "step": 1468
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.00011232669090196332,
      "loss": 0.4717,
      "step": 1469
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.00011222231070316712,
      "loss": 0.5527,
      "step": 1470
    },
    {
      "epoch": 9.52,
      "learning_rate": 0.00011211791698418761,
      "loss": 0.4465,
      "step": 1471
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.00011201350986050397,
      "loss": 0.469,
      "step": 1472
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.00011190908944761014,
      "loss": 0.573,
      "step": 1473
    },
    {
      "epoch": 9.54,
      "learning_rate": 0.00011180465586101487,
      "loss": 0.506,
      "step": 1474
    },
    {
      "epoch": 9.55,
      "learning_rate": 0.00011170020921624138,
      "loss": 0.5249,
      "step": 1475
    },
    {
      "epoch": 9.55,
      "learning_rate": 0.00011159574962882741,
      "loss": 0.5234,
      "step": 1476
    },
    {
      "epoch": 9.56,
      "learning_rate": 0.00011149127721432502,
      "loss": 0.5137,
      "step": 1477
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.00011138679208830034,
      "loss": 0.5554,
      "step": 1478
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.00011128229436633373,
      "loss": 0.4873,
      "step": 1479
    },
    {
      "epoch": 9.58,
      "learning_rate": 0.00011117778416401932,
      "loss": 0.5491,
      "step": 1480
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.00011107326159696521,
      "loss": 0.5503,
      "step": 1481
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.00011096872678079302,
      "loss": 0.4551,
      "step": 1482
    },
    {
      "epoch": 9.6,
      "learning_rate": 0.00011086417983113806,
      "loss": 0.4392,
      "step": 1483
    },
    {
      "epoch": 9.61,
      "learning_rate": 0.00011075962086364893,
      "loss": 0.5015,
      "step": 1484
    },
    {
      "epoch": 9.61,
      "learning_rate": 0.00011065504999398762,
      "loss": 0.4873,
      "step": 1485
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.00011055046733782927,
      "loss": 0.5115,
      "step": 1486
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.00011044587301086205,
      "loss": 0.5361,
      "step": 1487
    },
    {
      "epoch": 9.63,
      "learning_rate": 0.00011034126712878698,
      "loss": 0.5217,
      "step": 1488
    },
    {
      "epoch": 9.64,
      "learning_rate": 0.000110236649807318,
      "loss": 0.4846,
      "step": 1489
    },
    {
      "epoch": 9.64,
      "learning_rate": 0.00011013202116218153,
      "loss": 0.4993,
      "step": 1490
    },
    {
      "epoch": 9.65,
      "learning_rate": 0.00011002738130911666,
      "loss": 0.5696,
      "step": 1491
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.00010992273036387482,
      "loss": 0.5254,
      "step": 1492
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.00010981806844221966,
      "loss": 0.5503,
      "step": 1493
    },
    {
      "epoch": 9.67,
      "learning_rate": 0.00010971339565992712,
      "loss": 0.4585,
      "step": 1494
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.000109608712132785,
      "loss": 0.4727,
      "step": 1495
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.00010950401797659306,
      "loss": 0.4688,
      "step": 1496
    },
    {
      "epoch": 9.69,
      "learning_rate": 0.00010939931330716282,
      "loss": 0.4746,
      "step": 1497
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.00010929459824031738,
      "loss": 0.5371,
      "step": 1498
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.00010918987289189138,
      "loss": 0.5471,
      "step": 1499
    },
    {
      "epoch": 9.71,
      "learning_rate": 0.00010908513737773084,
      "loss": 0.4888,
      "step": 1500
    },
    {
      "epoch": 9.72,
      "learning_rate": 0.00010898039181369304,
      "loss": 0.55,
      "step": 1501
    },
    {
      "epoch": 9.72,
      "learning_rate": 0.0001088756363156463,
      "loss": 0.5598,
      "step": 1502
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.00010877087099947001,
      "loss": 0.4692,
      "step": 1503
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.00010866609598105442,
      "loss": 0.6497,
      "step": 1504
    },
    {
      "epoch": 9.74,
      "learning_rate": 0.00010856131137630039,
      "loss": 0.4983,
      "step": 1505
    },
    {
      "epoch": 9.75,
      "learning_rate": 0.00010845651730111954,
      "loss": 0.603,
      "step": 1506
    },
    {
      "epoch": 9.75,
      "learning_rate": 0.00010835171387143392,
      "loss": 0.4927,
      "step": 1507
    },
    {
      "epoch": 9.76,
      "learning_rate": 0.00010824690120317587,
      "loss": 0.4624,
      "step": 1508
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.000108142079412288,
      "loss": 0.6372,
      "step": 1509
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.00010803724861472304,
      "loss": 0.6899,
      "step": 1510
    },
    {
      "epoch": 9.78,
      "learning_rate": 0.0001079324089264436,
      "loss": 0.5986,
      "step": 1511
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.00010782756046342222,
      "loss": 0.5471,
      "step": 1512
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.000107722703341641,
      "loss": 0.5718,
      "step": 1513
    },
    {
      "epoch": 9.8,
      "learning_rate": 0.00010761783767709182,
      "loss": 0.5073,
      "step": 1514
    },
    {
      "epoch": 9.81,
      "learning_rate": 0.00010751296358577585,
      "loss": 0.5916,
      "step": 1515
    },
    {
      "epoch": 9.81,
      "learning_rate": 0.00010740808118370363,
      "loss": 0.4873,
      "step": 1516
    },
    {
      "epoch": 9.82,
      "learning_rate": 0.00010730319058689491,
      "loss": 0.6118,
      "step": 1517
    },
    {
      "epoch": 9.83,
      "learning_rate": 0.00010719829191137852,
      "loss": 0.4475,
      "step": 1518
    },
    {
      "epoch": 9.83,
      "learning_rate": 0.00010709338527319217,
      "loss": 0.4604,
      "step": 1519
    },
    {
      "epoch": 9.84,
      "learning_rate": 0.00010698847078838244,
      "loss": 0.5073,
      "step": 1520
    },
    {
      "epoch": 9.84,
      "learning_rate": 0.00010688354857300453,
      "loss": 0.4307,
      "step": 1521
    },
    {
      "epoch": 9.85,
      "learning_rate": 0.00010677861874312221,
      "loss": 0.5457,
      "step": 1522
    },
    {
      "epoch": 9.86,
      "learning_rate": 0.00010667368141480769,
      "loss": 0.5186,
      "step": 1523
    },
    {
      "epoch": 9.86,
      "learning_rate": 0.00010656873670414151,
      "loss": 0.5105,
      "step": 1524
    },
    {
      "epoch": 9.87,
      "learning_rate": 0.0001064637847272123,
      "loss": 0.4875,
      "step": 1525
    },
    {
      "epoch": 9.88,
      "learning_rate": 0.00010635882560011677,
      "loss": 0.5464,
      "step": 1526
    },
    {
      "epoch": 9.88,
      "learning_rate": 0.00010625385943895953,
      "loss": 0.5142,
      "step": 1527
    },
    {
      "epoch": 9.89,
      "learning_rate": 0.00010614888635985299,
      "loss": 0.5007,
      "step": 1528
    },
    {
      "epoch": 9.9,
      "learning_rate": 0.00010604390647891713,
      "loss": 0.5801,
      "step": 1529
    },
    {
      "epoch": 9.9,
      "learning_rate": 0.0001059389199122796,
      "loss": 0.502,
      "step": 1530
    },
    {
      "epoch": 9.91,
      "learning_rate": 0.0001058339267760753,
      "loss": 0.4927,
      "step": 1531
    },
    {
      "epoch": 9.92,
      "learning_rate": 0.00010572892718644652,
      "loss": 0.5205,
      "step": 1532
    },
    {
      "epoch": 9.92,
      "learning_rate": 0.00010562392125954261,
      "loss": 0.4705,
      "step": 1533
    },
    {
      "epoch": 9.93,
      "learning_rate": 0.00010551890911151991,
      "loss": 0.4487,
      "step": 1534
    },
    {
      "epoch": 9.94,
      "learning_rate": 0.00010541389085854176,
      "loss": 0.6267,
      "step": 1535
    },
    {
      "epoch": 9.94,
      "learning_rate": 0.00010530886661677812,
      "loss": 0.5106,
      "step": 1536
    },
    {
      "epoch": 9.95,
      "learning_rate": 0.00010520383650240564,
      "loss": 0.4492,
      "step": 1537
    },
    {
      "epoch": 9.95,
      "learning_rate": 0.00010509880063160747,
      "loss": 0.6592,
      "step": 1538
    },
    {
      "epoch": 9.96,
      "learning_rate": 0.0001049937591205731,
      "loss": 0.4683,
      "step": 1539
    },
    {
      "epoch": 9.97,
      "learning_rate": 0.00010488871208549826,
      "loss": 0.5337,
      "step": 1540
    },
    {
      "epoch": 9.97,
      "learning_rate": 0.00010478365964258485,
      "loss": 0.5823,
      "step": 1541
    },
    {
      "epoch": 9.98,
      "learning_rate": 0.00010467860190804069,
      "loss": 0.4856,
      "step": 1542
    },
    {
      "epoch": 9.99,
      "learning_rate": 0.00010457353899807946,
      "loss": 0.5171,
      "step": 1543
    },
    {
      "epoch": 9.99,
      "learning_rate": 0.00010446847102892057,
      "loss": 0.5376,
      "step": 1544
    },
    {
      "epoch": 10.0,
      "learning_rate": 0.00010436339811678904,
      "loss": 0.4397,
      "step": 1545
    },
    {
      "epoch": 10.01,
      "learning_rate": 0.00010425832037791534,
      "loss": 0.4331,
      "step": 1546
    },
    {
      "epoch": 10.01,
      "learning_rate": 0.00010415323792853529,
      "loss": 0.4041,
      "step": 1547
    },
    {
      "epoch": 10.02,
      "learning_rate": 0.00010404815088488995,
      "loss": 0.4487,
      "step": 1548
    },
    {
      "epoch": 10.03,
      "learning_rate": 0.00010394305936322537,
      "loss": 0.4443,
      "step": 1549
    },
    {
      "epoch": 10.03,
      "learning_rate": 0.00010383796347979271,
      "loss": 0.4478,
      "step": 1550
    },
    {
      "epoch": 10.04,
      "learning_rate": 0.00010373286335084775,
      "loss": 0.4729,
      "step": 1551
    },
    {
      "epoch": 10.05,
      "learning_rate": 0.00010362775909265117,
      "loss": 0.4594,
      "step": 1552
    },
    {
      "epoch": 10.05,
      "learning_rate": 0.00010352265082146808,
      "loss": 0.4415,
      "step": 1553
    },
    {
      "epoch": 10.06,
      "learning_rate": 0.00010341753865356804,
      "loss": 0.4658,
      "step": 1554
    },
    {
      "epoch": 10.06,
      "learning_rate": 0.000103312422705225,
      "loss": 0.4731,
      "step": 1555
    },
    {
      "epoch": 10.07,
      "learning_rate": 0.00010320730309271707,
      "loss": 0.5034,
      "step": 1556
    },
    {
      "epoch": 10.08,
      "learning_rate": 0.00010310217993232631,
      "loss": 0.4226,
      "step": 1557
    },
    {
      "epoch": 10.08,
      "learning_rate": 0.00010299705334033885,
      "loss": 0.3848,
      "step": 1558
    },
    {
      "epoch": 10.09,
      "learning_rate": 0.00010289192343304452,
      "loss": 0.4907,
      "step": 1559
    },
    {
      "epoch": 10.1,
      "learning_rate": 0.00010278679032673688,
      "loss": 0.4526,
      "step": 1560
    },
    {
      "epoch": 10.1,
      "learning_rate": 0.00010268165413771295,
      "loss": 0.4919,
      "step": 1561
    },
    {
      "epoch": 10.11,
      "learning_rate": 0.00010257651498227326,
      "loss": 0.4619,
      "step": 1562
    },
    {
      "epoch": 10.12,
      "learning_rate": 0.00010247137297672146,
      "loss": 0.4194,
      "step": 1563
    },
    {
      "epoch": 10.12,
      "learning_rate": 0.00010236622823736457,
      "loss": 0.5503,
      "step": 1564
    },
    {
      "epoch": 10.13,
      "learning_rate": 0.00010226108088051252,
      "loss": 0.4423,
      "step": 1565
    },
    {
      "epoch": 10.14,
      "learning_rate": 0.00010215593102247807,
      "loss": 0.4583,
      "step": 1566
    },
    {
      "epoch": 10.14,
      "learning_rate": 0.00010205077877957685,
      "loss": 0.4021,
      "step": 1567
    },
    {
      "epoch": 10.15,
      "learning_rate": 0.00010194562426812713,
      "loss": 0.4343,
      "step": 1568
    },
    {
      "epoch": 10.16,
      "learning_rate": 0.00010184046760444959,
      "loss": 0.3904,
      "step": 1569
    },
    {
      "epoch": 10.16,
      "learning_rate": 0.00010173530890486737,
      "loss": 0.4331,
      "step": 1570
    },
    {
      "epoch": 10.17,
      "learning_rate": 0.0001016301482857059,
      "loss": 0.5166,
      "step": 1571
    },
    {
      "epoch": 10.17,
      "learning_rate": 0.0001015249858632926,
      "loss": 0.4011,
      "step": 1572
    },
    {
      "epoch": 10.18,
      "learning_rate": 0.00010141982175395701,
      "loss": 0.4419,
      "step": 1573
    },
    {
      "epoch": 10.19,
      "learning_rate": 0.00010131465607403052,
      "loss": 0.3845,
      "step": 1574
    },
    {
      "epoch": 10.19,
      "learning_rate": 0.00010120948893984618,
      "loss": 0.4902,
      "step": 1575
    },
    {
      "epoch": 10.2,
      "learning_rate": 0.00010110432046773872,
      "loss": 0.3975,
      "step": 1576
    },
    {
      "epoch": 10.21,
      "learning_rate": 0.00010099915077404432,
      "loss": 0.4763,
      "step": 1577
    },
    {
      "epoch": 10.21,
      "learning_rate": 0.00010089397997510048,
      "loss": 0.4197,
      "step": 1578
    },
    {
      "epoch": 10.22,
      "learning_rate": 0.000100788808187246,
      "loss": 0.522,
      "step": 1579
    },
    {
      "epoch": 10.23,
      "learning_rate": 0.00010068363552682078,
      "loss": 0.4329,
      "step": 1580
    },
    {
      "epoch": 10.23,
      "learning_rate": 0.00010057846211016555,
      "loss": 0.5305,
      "step": 1581
    },
    {
      "epoch": 10.24,
      "learning_rate": 0.00010047328805362204,
      "loss": 0.4646,
      "step": 1582
    },
    {
      "epoch": 10.25,
      "learning_rate": 0.00010036811347353257,
      "loss": 0.4397,
      "step": 1583
    },
    {
      "epoch": 10.25,
      "learning_rate": 0.00010026293848624009,
      "loss": 0.3929,
      "step": 1584
    },
    {
      "epoch": 10.26,
      "learning_rate": 0.00010015776320808798,
      "loss": 0.426,
      "step": 1585
    },
    {
      "epoch": 10.27,
      "learning_rate": 0.00010005258775542001,
      "loss": 0.3567,
      "step": 1586
    },
    {
      "epoch": 10.27,
      "learning_rate": 9.994741224458004e-05,
      "loss": 0.4465,
      "step": 1587
    },
    {
      "epoch": 10.28,
      "learning_rate": 9.984223679191204e-05,
      "loss": 0.4717,
      "step": 1588
    },
    {
      "epoch": 10.28,
      "learning_rate": 9.973706151375994e-05,
      "loss": 0.4595,
      "step": 1589
    },
    {
      "epoch": 10.29,
      "learning_rate": 9.963188652646745e-05,
      "loss": 0.502,
      "step": 1590
    },
    {
      "epoch": 10.3,
      "learning_rate": 9.952671194637799e-05,
      "loss": 0.426,
      "step": 1591
    },
    {
      "epoch": 10.3,
      "learning_rate": 9.942153788983445e-05,
      "loss": 0.4189,
      "step": 1592
    },
    {
      "epoch": 10.31,
      "learning_rate": 9.931636447317927e-05,
      "loss": 0.4761,
      "step": 1593
    },
    {
      "epoch": 10.32,
      "learning_rate": 9.9211191812754e-05,
      "loss": 0.4197,
      "step": 1594
    },
    {
      "epoch": 10.32,
      "learning_rate": 9.910602002489954e-05,
      "loss": 0.4746,
      "step": 1595
    },
    {
      "epoch": 10.33,
      "learning_rate": 9.90008492259557e-05,
      "loss": 0.4189,
      "step": 1596
    },
    {
      "epoch": 10.34,
      "learning_rate": 9.889567953226129e-05,
      "loss": 0.396,
      "step": 1597
    },
    {
      "epoch": 10.34,
      "learning_rate": 9.879051106015383e-05,
      "loss": 0.4763,
      "step": 1598
    },
    {
      "epoch": 10.35,
      "learning_rate": 9.86853439259695e-05,
      "loss": 0.4575,
      "step": 1599
    },
    {
      "epoch": 10.36,
      "learning_rate": 9.8580178246043e-05,
      "loss": 0.4377,
      "step": 1600
    },
    {
      "epoch": 10.36,
      "learning_rate": 9.847501413670742e-05,
      "loss": 0.5271,
      "step": 1601
    },
    {
      "epoch": 10.37,
      "learning_rate": 9.836985171429412e-05,
      "loss": 0.4602,
      "step": 1602
    },
    {
      "epoch": 10.38,
      "learning_rate": 9.826469109513264e-05,
      "loss": 0.48,
      "step": 1603
    },
    {
      "epoch": 10.38,
      "learning_rate": 9.815953239555041e-05,
      "loss": 0.4318,
      "step": 1604
    },
    {
      "epoch": 10.39,
      "learning_rate": 9.805437573187292e-05,
      "loss": 0.4814,
      "step": 1605
    },
    {
      "epoch": 10.39,
      "learning_rate": 9.794922122042316e-05,
      "loss": 0.4763,
      "step": 1606
    },
    {
      "epoch": 10.4,
      "learning_rate": 9.784406897752196e-05,
      "loss": 0.4604,
      "step": 1607
    },
    {
      "epoch": 10.41,
      "learning_rate": 9.773891911948751e-05,
      "loss": 0.5105,
      "step": 1608
    },
    {
      "epoch": 10.41,
      "learning_rate": 9.763377176263541e-05,
      "loss": 0.51,
      "step": 1609
    },
    {
      "epoch": 10.42,
      "learning_rate": 9.752862702327857e-05,
      "loss": 0.4827,
      "step": 1610
    },
    {
      "epoch": 10.43,
      "learning_rate": 9.74234850177268e-05,
      "loss": 0.447,
      "step": 1611
    },
    {
      "epoch": 10.43,
      "learning_rate": 9.731834586228707e-05,
      "loss": 0.5212,
      "step": 1612
    },
    {
      "epoch": 10.44,
      "learning_rate": 9.721320967326313e-05,
      "loss": 0.4692,
      "step": 1613
    },
    {
      "epoch": 10.45,
      "learning_rate": 9.710807656695549e-05,
      "loss": 0.4443,
      "step": 1614
    },
    {
      "epoch": 10.45,
      "learning_rate": 9.700294665966116e-05,
      "loss": 0.3953,
      "step": 1615
    },
    {
      "epoch": 10.46,
      "learning_rate": 9.689782006767372e-05,
      "loss": 0.4424,
      "step": 1616
    },
    {
      "epoch": 10.47,
      "learning_rate": 9.679269690728299e-05,
      "loss": 0.4888,
      "step": 1617
    },
    {
      "epoch": 10.47,
      "learning_rate": 9.668757729477502e-05,
      "loss": 0.4937,
      "step": 1618
    },
    {
      "epoch": 10.48,
      "learning_rate": 9.658246134643198e-05,
      "loss": 0.4888,
      "step": 1619
    },
    {
      "epoch": 10.49,
      "learning_rate": 9.647734917853195e-05,
      "loss": 0.4885,
      "step": 1620
    },
    {
      "epoch": 10.49,
      "learning_rate": 9.637224090734886e-05,
      "loss": 0.5056,
      "step": 1621
    },
    {
      "epoch": 10.5,
      "learning_rate": 9.626713664915227e-05,
      "loss": 0.45,
      "step": 1622
    },
    {
      "epoch": 10.5,
      "learning_rate": 9.616203652020733e-05,
      "loss": 0.4807,
      "step": 1623
    },
    {
      "epoch": 10.51,
      "learning_rate": 9.605694063677464e-05,
      "loss": 0.5288,
      "step": 1624
    },
    {
      "epoch": 10.52,
      "learning_rate": 9.595184911511008e-05,
      "loss": 0.449,
      "step": 1625
    },
    {
      "epoch": 10.52,
      "learning_rate": 9.58467620714647e-05,
      "loss": 0.4194,
      "step": 1626
    },
    {
      "epoch": 10.53,
      "learning_rate": 9.574167962208467e-05,
      "loss": 0.4519,
      "step": 1627
    },
    {
      "epoch": 10.54,
      "learning_rate": 9.563660188321101e-05,
      "loss": 0.5139,
      "step": 1628
    },
    {
      "epoch": 10.54,
      "learning_rate": 9.553152897107947e-05,
      "loss": 0.4509,
      "step": 1629
    },
    {
      "epoch": 10.55,
      "learning_rate": 9.542646100192056e-05,
      "loss": 0.3665,
      "step": 1630
    },
    {
      "epoch": 10.56,
      "learning_rate": 9.532139809195933e-05,
      "loss": 0.4153,
      "step": 1631
    },
    {
      "epoch": 10.56,
      "learning_rate": 9.521634035741515e-05,
      "loss": 0.4907,
      "step": 1632
    },
    {
      "epoch": 10.57,
      "learning_rate": 9.511128791450172e-05,
      "loss": 0.467,
      "step": 1633
    },
    {
      "epoch": 10.58,
      "learning_rate": 9.500624087942694e-05,
      "loss": 0.5322,
      "step": 1634
    },
    {
      "epoch": 10.58,
      "learning_rate": 9.490119936839257e-05,
      "loss": 0.5122,
      "step": 1635
    },
    {
      "epoch": 10.59,
      "learning_rate": 9.479616349759437e-05,
      "loss": 0.4304,
      "step": 1636
    },
    {
      "epoch": 10.6,
      "learning_rate": 9.469113338322189e-05,
      "loss": 0.5168,
      "step": 1637
    },
    {
      "epoch": 10.6,
      "learning_rate": 9.458610914145826e-05,
      "loss": 0.4058,
      "step": 1638
    },
    {
      "epoch": 10.61,
      "learning_rate": 9.448109088848007e-05,
      "loss": 0.468,
      "step": 1639
    },
    {
      "epoch": 10.61,
      "learning_rate": 9.437607874045744e-05,
      "loss": 0.5374,
      "step": 1640
    },
    {
      "epoch": 10.62,
      "learning_rate": 9.42710728135535e-05,
      "loss": 0.5391,
      "step": 1641
    },
    {
      "epoch": 10.63,
      "learning_rate": 9.41660732239247e-05,
      "loss": 0.3984,
      "step": 1642
    },
    {
      "epoch": 10.63,
      "learning_rate": 9.406108008772042e-05,
      "loss": 0.3984,
      "step": 1643
    },
    {
      "epoch": 10.64,
      "learning_rate": 9.395609352108287e-05,
      "loss": 0.4255,
      "step": 1644
    },
    {
      "epoch": 10.65,
      "learning_rate": 9.385111364014708e-05,
      "loss": 0.4341,
      "step": 1645
    },
    {
      "epoch": 10.65,
      "learning_rate": 9.37461405610405e-05,
      "loss": 0.4158,
      "step": 1646
    },
    {
      "epoch": 10.66,
      "learning_rate": 9.364117439988325e-05,
      "loss": 0.5288,
      "step": 1647
    },
    {
      "epoch": 10.67,
      "learning_rate": 9.353621527278772e-05,
      "loss": 0.4431,
      "step": 1648
    },
    {
      "epoch": 10.67,
      "learning_rate": 9.34312632958585e-05,
      "loss": 0.4712,
      "step": 1649
    },
    {
      "epoch": 10.68,
      "learning_rate": 9.332631858519229e-05,
      "loss": 0.4497,
      "step": 1650
    },
    {
      "epoch": 10.69,
      "learning_rate": 9.322138125687784e-05,
      "loss": 0.4189,
      "step": 1651
    },
    {
      "epoch": 10.69,
      "learning_rate": 9.311645142699552e-05,
      "loss": 0.4814,
      "step": 1652
    },
    {
      "epoch": 10.7,
      "learning_rate": 9.30115292116176e-05,
      "loss": 0.4932,
      "step": 1653
    },
    {
      "epoch": 10.71,
      "learning_rate": 9.290661472680786e-05,
      "loss": 0.4546,
      "step": 1654
    },
    {
      "epoch": 10.71,
      "learning_rate": 9.28017080886215e-05,
      "loss": 0.45,
      "step": 1655
    },
    {
      "epoch": 10.72,
      "learning_rate": 9.26968094131051e-05,
      "loss": 0.4425,
      "step": 1656
    },
    {
      "epoch": 10.72,
      "learning_rate": 9.259191881629643e-05,
      "loss": 0.4995,
      "step": 1657
    },
    {
      "epoch": 10.73,
      "learning_rate": 9.24870364142242e-05,
      "loss": 0.478,
      "step": 1658
    },
    {
      "epoch": 10.74,
      "learning_rate": 9.238216232290822e-05,
      "loss": 0.4976,
      "step": 1659
    },
    {
      "epoch": 10.74,
      "learning_rate": 9.227729665835901e-05,
      "loss": 0.4027,
      "step": 1660
    },
    {
      "epoch": 10.75,
      "learning_rate": 9.21724395365778e-05,
      "loss": 0.4685,
      "step": 1661
    },
    {
      "epoch": 10.76,
      "learning_rate": 9.20675910735564e-05,
      "loss": 0.5107,
      "step": 1662
    },
    {
      "epoch": 10.76,
      "learning_rate": 9.196275138527698e-05,
      "loss": 0.4878,
      "step": 1663
    },
    {
      "epoch": 10.77,
      "learning_rate": 9.185792058771202e-05,
      "loss": 0.408,
      "step": 1664
    },
    {
      "epoch": 10.78,
      "learning_rate": 9.175309879682416e-05,
      "loss": 0.4778,
      "step": 1665
    },
    {
      "epoch": 10.78,
      "learning_rate": 9.16482861285661e-05,
      "loss": 0.4431,
      "step": 1666
    },
    {
      "epoch": 10.79,
      "learning_rate": 9.154348269888046e-05,
      "loss": 0.4724,
      "step": 1667
    },
    {
      "epoch": 10.8,
      "learning_rate": 9.143868862369961e-05,
      "loss": 0.4585,
      "step": 1668
    },
    {
      "epoch": 10.8,
      "learning_rate": 9.133390401894564e-05,
      "loss": 0.478,
      "step": 1669
    },
    {
      "epoch": 10.81,
      "learning_rate": 9.122912900053e-05,
      "loss": 0.436,
      "step": 1670
    },
    {
      "epoch": 10.82,
      "learning_rate": 9.112436368435373e-05,
      "loss": 0.5076,
      "step": 1671
    },
    {
      "epoch": 10.82,
      "learning_rate": 9.101960818630698e-05,
      "loss": 0.4858,
      "step": 1672
    },
    {
      "epoch": 10.83,
      "learning_rate": 9.091486262226917e-05,
      "loss": 0.4624,
      "step": 1673
    },
    {
      "epoch": 10.83,
      "learning_rate": 9.081012710810862e-05,
      "loss": 0.3621,
      "step": 1674
    },
    {
      "epoch": 10.84,
      "learning_rate": 9.070540175968266e-05,
      "loss": 0.4263,
      "step": 1675
    },
    {
      "epoch": 10.85,
      "learning_rate": 9.060068669283722e-05,
      "loss": 0.4617,
      "step": 1676
    },
    {
      "epoch": 10.85,
      "learning_rate": 9.049598202340695e-05,
      "loss": 0.498,
      "step": 1677
    },
    {
      "epoch": 10.86,
      "learning_rate": 9.039128786721502e-05,
      "loss": 0.4736,
      "step": 1678
    },
    {
      "epoch": 10.87,
      "learning_rate": 9.02866043400729e-05,
      "loss": 0.5889,
      "step": 1679
    },
    {
      "epoch": 10.87,
      "learning_rate": 9.018193155778033e-05,
      "loss": 0.4417,
      "step": 1680
    },
    {
      "epoch": 10.88,
      "learning_rate": 9.007726963612523e-05,
      "loss": 0.4734,
      "step": 1681
    },
    {
      "epoch": 10.89,
      "learning_rate": 8.997261869088338e-05,
      "loss": 0.4529,
      "step": 1682
    },
    {
      "epoch": 10.89,
      "learning_rate": 8.98679788378185e-05,
      "loss": 0.4055,
      "step": 1683
    },
    {
      "epoch": 10.9,
      "learning_rate": 8.976335019268202e-05,
      "loss": 0.4956,
      "step": 1684
    },
    {
      "epoch": 10.91,
      "learning_rate": 8.965873287121303e-05,
      "loss": 0.5103,
      "step": 1685
    },
    {
      "epoch": 10.91,
      "learning_rate": 8.955412698913799e-05,
      "loss": 0.4993,
      "step": 1686
    },
    {
      "epoch": 10.92,
      "learning_rate": 8.944953266217076e-05,
      "loss": 0.4797,
      "step": 1687
    },
    {
      "epoch": 10.93,
      "learning_rate": 8.93449500060124e-05,
      "loss": 0.5005,
      "step": 1688
    },
    {
      "epoch": 10.93,
      "learning_rate": 8.92403791363511e-05,
      "loss": 0.5569,
      "step": 1689
    },
    {
      "epoch": 10.94,
      "learning_rate": 8.913582016886198e-05,
      "loss": 0.4326,
      "step": 1690
    },
    {
      "epoch": 10.94,
      "learning_rate": 8.903127321920698e-05,
      "loss": 0.4417,
      "step": 1691
    },
    {
      "epoch": 10.95,
      "learning_rate": 8.892673840303484e-05,
      "loss": 0.3838,
      "step": 1692
    },
    {
      "epoch": 10.96,
      "learning_rate": 8.882221583598069e-05,
      "loss": 0.4495,
      "step": 1693
    },
    {
      "epoch": 10.96,
      "learning_rate": 8.871770563366631e-05,
      "loss": 0.4419,
      "step": 1694
    },
    {
      "epoch": 10.97,
      "learning_rate": 8.861320791169968e-05,
      "loss": 0.4216,
      "step": 1695
    },
    {
      "epoch": 10.98,
      "learning_rate": 8.850872278567501e-05,
      "loss": 0.4465,
      "step": 1696
    },
    {
      "epoch": 10.98,
      "learning_rate": 8.840425037117258e-05,
      "loss": 0.5154,
      "step": 1697
    },
    {
      "epoch": 10.99,
      "learning_rate": 8.829979078375865e-05,
      "loss": 0.489,
      "step": 1698
    },
    {
      "epoch": 11.0,
      "learning_rate": 8.819534413898516e-05,
      "loss": 0.4619,
      "step": 1699
    },
    {
      "epoch": 11.0,
      "learning_rate": 8.809091055238988e-05,
      "loss": 0.4126,
      "step": 1700
    },
    {
      "epoch": 11.01,
      "learning_rate": 8.798649013949605e-05,
      "loss": 0.3711,
      "step": 1701
    },
    {
      "epoch": 11.02,
      "learning_rate": 8.788208301581241e-05,
      "loss": 0.4055,
      "step": 1702
    },
    {
      "epoch": 11.02,
      "learning_rate": 8.777768929683289e-05,
      "loss": 0.4282,
      "step": 1703
    },
    {
      "epoch": 11.03,
      "learning_rate": 8.767330909803672e-05,
      "loss": 0.3984,
      "step": 1704
    },
    {
      "epoch": 11.04,
      "learning_rate": 8.756894253488801e-05,
      "loss": 0.3748,
      "step": 1705
    },
    {
      "epoch": 11.04,
      "learning_rate": 8.746458972283594e-05,
      "loss": 0.4189,
      "step": 1706
    },
    {
      "epoch": 11.05,
      "learning_rate": 8.736025077731443e-05,
      "loss": 0.4138,
      "step": 1707
    },
    {
      "epoch": 11.06,
      "learning_rate": 8.725592581374205e-05,
      "loss": 0.4075,
      "step": 1708
    },
    {
      "epoch": 11.06,
      "learning_rate": 8.715161494752193e-05,
      "loss": 0.3994,
      "step": 1709
    },
    {
      "epoch": 11.07,
      "learning_rate": 8.704731829404155e-05,
      "loss": 0.3682,
      "step": 1710
    },
    {
      "epoch": 11.07,
      "learning_rate": 8.69430359686727e-05,
      "loss": 0.3716,
      "step": 1711
    },
    {
      "epoch": 11.08,
      "learning_rate": 8.683876808677132e-05,
      "loss": 0.3738,
      "step": 1712
    },
    {
      "epoch": 11.09,
      "learning_rate": 8.673451476367736e-05,
      "loss": 0.2982,
      "step": 1713
    },
    {
      "epoch": 11.09,
      "learning_rate": 8.663027611471473e-05,
      "loss": 0.3955,
      "step": 1714
    },
    {
      "epoch": 11.1,
      "learning_rate": 8.652605225519101e-05,
      "loss": 0.4062,
      "step": 1715
    },
    {
      "epoch": 11.11,
      "learning_rate": 8.642184330039754e-05,
      "loss": 0.3484,
      "step": 1716
    },
    {
      "epoch": 11.11,
      "learning_rate": 8.6317649365609e-05,
      "loss": 0.4716,
      "step": 1717
    },
    {
      "epoch": 11.12,
      "learning_rate": 8.621347056608361e-05,
      "loss": 0.3984,
      "step": 1718
    },
    {
      "epoch": 11.13,
      "learning_rate": 8.610930701706277e-05,
      "loss": 0.4241,
      "step": 1719
    },
    {
      "epoch": 11.13,
      "learning_rate": 8.600515883377103e-05,
      "loss": 0.4043,
      "step": 1720
    },
    {
      "epoch": 11.14,
      "learning_rate": 8.590102613141602e-05,
      "loss": 0.4003,
      "step": 1721
    },
    {
      "epoch": 11.15,
      "learning_rate": 8.579690902518807e-05,
      "loss": 0.4143,
      "step": 1722
    },
    {
      "epoch": 11.15,
      "learning_rate": 8.569280763026038e-05,
      "loss": 0.3741,
      "step": 1723
    },
    {
      "epoch": 11.16,
      "learning_rate": 8.558872206178876e-05,
      "loss": 0.3469,
      "step": 1724
    },
    {
      "epoch": 11.17,
      "learning_rate": 8.548465243491154e-05,
      "loss": 0.3804,
      "step": 1725
    },
    {
      "epoch": 11.17,
      "learning_rate": 8.538059886474932e-05,
      "loss": 0.3853,
      "step": 1726
    },
    {
      "epoch": 11.18,
      "learning_rate": 8.527656146640503e-05,
      "loss": 0.3652,
      "step": 1727
    },
    {
      "epoch": 11.18,
      "learning_rate": 8.517254035496366e-05,
      "loss": 0.4087,
      "step": 1728
    },
    {
      "epoch": 11.19,
      "learning_rate": 8.506853564549217e-05,
      "loss": 0.4011,
      "step": 1729
    },
    {
      "epoch": 11.2,
      "learning_rate": 8.496454745303945e-05,
      "loss": 0.418,
      "step": 1730
    },
    {
      "epoch": 11.2,
      "learning_rate": 8.486057589263608e-05,
      "loss": 0.439,
      "step": 1731
    },
    {
      "epoch": 11.21,
      "learning_rate": 8.475662107929418e-05,
      "loss": 0.4121,
      "step": 1732
    },
    {
      "epoch": 11.22,
      "learning_rate": 8.465268312800752e-05,
      "loss": 0.4163,
      "step": 1733
    },
    {
      "epoch": 11.22,
      "learning_rate": 8.454876215375095e-05,
      "loss": 0.3965,
      "step": 1734
    },
    {
      "epoch": 11.23,
      "learning_rate": 8.444485827148078e-05,
      "loss": 0.4534,
      "step": 1735
    },
    {
      "epoch": 11.24,
      "learning_rate": 8.434097159613427e-05,
      "loss": 0.4165,
      "step": 1736
    },
    {
      "epoch": 11.24,
      "learning_rate": 8.423710224262974e-05,
      "loss": 0.4092,
      "step": 1737
    },
    {
      "epoch": 11.25,
      "learning_rate": 8.413325032586628e-05,
      "loss": 0.3489,
      "step": 1738
    },
    {
      "epoch": 11.26,
      "learning_rate": 8.402941596072378e-05,
      "loss": 0.4216,
      "step": 1739
    },
    {
      "epoch": 11.26,
      "learning_rate": 8.392559926206256e-05,
      "loss": 0.489,
      "step": 1740
    },
    {
      "epoch": 11.27,
      "learning_rate": 8.382180034472353e-05,
      "loss": 0.4133,
      "step": 1741
    },
    {
      "epoch": 11.28,
      "learning_rate": 8.371801932352791e-05,
      "loss": 0.4277,
      "step": 1742
    },
    {
      "epoch": 11.28,
      "learning_rate": 8.361425631327707e-05,
      "loss": 0.3987,
      "step": 1743
    },
    {
      "epoch": 11.29,
      "learning_rate": 8.351051142875252e-05,
      "loss": 0.4133,
      "step": 1744
    },
    {
      "epoch": 11.29,
      "learning_rate": 8.340678478471571e-05,
      "loss": 0.4172,
      "step": 1745
    },
    {
      "epoch": 11.3,
      "learning_rate": 8.33030764959078e-05,
      "loss": 0.4619,
      "step": 1746
    },
    {
      "epoch": 11.31,
      "learning_rate": 8.319938667704982e-05,
      "loss": 0.4482,
      "step": 1747
    },
    {
      "epoch": 11.31,
      "learning_rate": 8.309571544284227e-05,
      "loss": 0.4163,
      "step": 1748
    },
    {
      "epoch": 11.32,
      "learning_rate": 8.299206290796511e-05,
      "loss": 0.3499,
      "step": 1749
    },
    {
      "epoch": 11.33,
      "learning_rate": 8.28884291870776e-05,
      "loss": 0.3473,
      "step": 1750
    },
    {
      "epoch": 11.33,
      "learning_rate": 8.278481439481825e-05,
      "loss": 0.4092,
      "step": 1751
    },
    {
      "epoch": 11.34,
      "learning_rate": 8.26812186458045e-05,
      "loss": 0.4282,
      "step": 1752
    },
    {
      "epoch": 11.35,
      "learning_rate": 8.257764205463285e-05,
      "loss": 0.3914,
      "step": 1753
    },
    {
      "epoch": 11.35,
      "learning_rate": 8.247408473587858e-05,
      "loss": 0.3325,
      "step": 1754
    },
    {
      "epoch": 11.36,
      "learning_rate": 8.237054680409561e-05,
      "loss": 0.4292,
      "step": 1755
    },
    {
      "epoch": 11.37,
      "learning_rate": 8.22670283738165e-05,
      "loss": 0.385,
      "step": 1756
    },
    {
      "epoch": 11.37,
      "learning_rate": 8.216352955955213e-05,
      "loss": 0.4043,
      "step": 1757
    },
    {
      "epoch": 11.38,
      "learning_rate": 8.206005047579172e-05,
      "loss": 0.4238,
      "step": 1758
    },
    {
      "epoch": 11.39,
      "learning_rate": 8.195659123700267e-05,
      "loss": 0.4277,
      "step": 1759
    },
    {
      "epoch": 11.39,
      "learning_rate": 8.185315195763047e-05,
      "loss": 0.4138,
      "step": 1760
    },
    {
      "epoch": 11.4,
      "learning_rate": 8.174973275209848e-05,
      "loss": 0.4272,
      "step": 1761
    },
    {
      "epoch": 11.4,
      "learning_rate": 8.164633373480788e-05,
      "loss": 0.4231,
      "step": 1762
    },
    {
      "epoch": 11.41,
      "learning_rate": 8.154295502013746e-05,
      "loss": 0.4358,
      "step": 1763
    },
    {
      "epoch": 11.42,
      "learning_rate": 8.143959672244364e-05,
      "loss": 0.4709,
      "step": 1764
    },
    {
      "epoch": 11.42,
      "learning_rate": 8.133625895606018e-05,
      "loss": 0.3367,
      "step": 1765
    },
    {
      "epoch": 11.43,
      "learning_rate": 8.12329418352982e-05,
      "loss": 0.48,
      "step": 1766
    },
    {
      "epoch": 11.44,
      "learning_rate": 8.112964547444591e-05,
      "loss": 0.375,
      "step": 1767
    },
    {
      "epoch": 11.44,
      "learning_rate": 8.102636998776865e-05,
      "loss": 0.4808,
      "step": 1768
    },
    {
      "epoch": 11.45,
      "learning_rate": 8.09231154895085e-05,
      "loss": 0.4199,
      "step": 1769
    },
    {
      "epoch": 11.46,
      "learning_rate": 8.08198820938845e-05,
      "loss": 0.3865,
      "step": 1770
    },
    {
      "epoch": 11.46,
      "learning_rate": 8.071666991509224e-05,
      "loss": 0.447,
      "step": 1771
    },
    {
      "epoch": 11.47,
      "learning_rate": 8.061347906730392e-05,
      "loss": 0.4072,
      "step": 1772
    },
    {
      "epoch": 11.48,
      "learning_rate": 8.051030966466811e-05,
      "loss": 0.4243,
      "step": 1773
    },
    {
      "epoch": 11.48,
      "learning_rate": 8.040716182130961e-05,
      "loss": 0.3818,
      "step": 1774
    },
    {
      "epoch": 11.49,
      "learning_rate": 8.030403565132942e-05,
      "loss": 0.3347,
      "step": 1775
    },
    {
      "epoch": 11.5,
      "learning_rate": 8.020093126880454e-05,
      "loss": 0.3467,
      "step": 1776
    },
    {
      "epoch": 11.5,
      "learning_rate": 8.00978487877879e-05,
      "loss": 0.417,
      "step": 1777
    },
    {
      "epoch": 11.51,
      "learning_rate": 7.99947883223082e-05,
      "loss": 0.3909,
      "step": 1778
    },
    {
      "epoch": 11.51,
      "learning_rate": 7.989174998636973e-05,
      "loss": 0.47,
      "step": 1779
    },
    {
      "epoch": 11.52,
      "learning_rate": 7.978873389395243e-05,
      "loss": 0.4783,
      "step": 1780
    },
    {
      "epoch": 11.53,
      "learning_rate": 7.968574015901147e-05,
      "loss": 0.4143,
      "step": 1781
    },
    {
      "epoch": 11.53,
      "learning_rate": 7.958276889547738e-05,
      "loss": 0.4005,
      "step": 1782
    },
    {
      "epoch": 11.54,
      "learning_rate": 7.947982021725583e-05,
      "loss": 0.3674,
      "step": 1783
    },
    {
      "epoch": 11.55,
      "learning_rate": 7.93768942382275e-05,
      "loss": 0.4636,
      "step": 1784
    },
    {
      "epoch": 11.55,
      "learning_rate": 7.927399107224794e-05,
      "loss": 0.4565,
      "step": 1785
    },
    {
      "epoch": 11.56,
      "learning_rate": 7.917111083314751e-05,
      "loss": 0.3433,
      "step": 1786
    },
    {
      "epoch": 11.57,
      "learning_rate": 7.906825363473115e-05,
      "loss": 0.3843,
      "step": 1787
    },
    {
      "epoch": 11.57,
      "learning_rate": 7.896541959077834e-05,
      "loss": 0.411,
      "step": 1788
    },
    {
      "epoch": 11.58,
      "learning_rate": 7.8862608815043e-05,
      "loss": 0.4182,
      "step": 1789
    },
    {
      "epoch": 11.59,
      "learning_rate": 7.875982142125318e-05,
      "loss": 0.4084,
      "step": 1790
    },
    {
      "epoch": 11.59,
      "learning_rate": 7.865705752311118e-05,
      "loss": 0.5076,
      "step": 1791
    },
    {
      "epoch": 11.6,
      "learning_rate": 7.855431723429333e-05,
      "loss": 0.4072,
      "step": 1792
    },
    {
      "epoch": 11.61,
      "learning_rate": 7.845160066844968e-05,
      "loss": 0.4066,
      "step": 1793
    },
    {
      "epoch": 11.61,
      "learning_rate": 7.834890793920422e-05,
      "loss": 0.5337,
      "step": 1794
    },
    {
      "epoch": 11.62,
      "learning_rate": 7.824623916015447e-05,
      "loss": 0.4116,
      "step": 1795
    },
    {
      "epoch": 11.62,
      "learning_rate": 7.814359444487153e-05,
      "loss": 0.4417,
      "step": 1796
    },
    {
      "epoch": 11.63,
      "learning_rate": 7.804097390689979e-05,
      "loss": 0.4248,
      "step": 1797
    },
    {
      "epoch": 11.64,
      "learning_rate": 7.793837765975693e-05,
      "loss": 0.4167,
      "step": 1798
    },
    {
      "epoch": 11.64,
      "learning_rate": 7.78358058169338e-05,
      "loss": 0.4458,
      "step": 1799
    },
    {
      "epoch": 11.65,
      "learning_rate": 7.773325849189421e-05,
      "loss": 0.4099,
      "step": 1800
    },
    {
      "epoch": 11.66,
      "learning_rate": 7.763073579807485e-05,
      "loss": 0.4512,
      "step": 1801
    },
    {
      "epoch": 11.66,
      "learning_rate": 7.752823784888521e-05,
      "loss": 0.4724,
      "step": 1802
    },
    {
      "epoch": 11.67,
      "learning_rate": 7.74257647577074e-05,
      "loss": 0.4338,
      "step": 1803
    },
    {
      "epoch": 11.68,
      "learning_rate": 7.732331663789592e-05,
      "loss": 0.4285,
      "step": 1804
    },
    {
      "epoch": 11.68,
      "learning_rate": 7.722089360277778e-05,
      "loss": 0.4521,
      "step": 1805
    },
    {
      "epoch": 11.69,
      "learning_rate": 7.71184957656522e-05,
      "loss": 0.4497,
      "step": 1806
    },
    {
      "epoch": 11.7,
      "learning_rate": 7.70161232397905e-05,
      "loss": 0.4797,
      "step": 1807
    },
    {
      "epoch": 11.7,
      "learning_rate": 7.691377613843606e-05,
      "loss": 0.4028,
      "step": 1808
    },
    {
      "epoch": 11.71,
      "learning_rate": 7.68114545748041e-05,
      "loss": 0.408,
      "step": 1809
    },
    {
      "epoch": 11.72,
      "learning_rate": 7.670915866208152e-05,
      "loss": 0.3549,
      "step": 1810
    },
    {
      "epoch": 11.72,
      "learning_rate": 7.660688851342695e-05,
      "loss": 0.4468,
      "step": 1811
    },
    {
      "epoch": 11.73,
      "learning_rate": 7.650464424197052e-05,
      "loss": 0.3745,
      "step": 1812
    },
    {
      "epoch": 11.73,
      "learning_rate": 7.640242596081366e-05,
      "loss": 0.4421,
      "step": 1813
    },
    {
      "epoch": 11.74,
      "learning_rate": 7.630023378302911e-05,
      "loss": 0.4277,
      "step": 1814
    },
    {
      "epoch": 11.75,
      "learning_rate": 7.619806782166073e-05,
      "loss": 0.4573,
      "step": 1815
    },
    {
      "epoch": 11.75,
      "learning_rate": 7.609592818972328e-05,
      "loss": 0.4351,
      "step": 1816
    },
    {
      "epoch": 11.76,
      "learning_rate": 7.599381500020254e-05,
      "loss": 0.4348,
      "step": 1817
    },
    {
      "epoch": 11.77,
      "learning_rate": 7.589172836605498e-05,
      "loss": 0.4285,
      "step": 1818
    },
    {
      "epoch": 11.77,
      "learning_rate": 7.578966840020767e-05,
      "loss": 0.3867,
      "step": 1819
    },
    {
      "epoch": 11.78,
      "learning_rate": 7.568763521555825e-05,
      "loss": 0.4363,
      "step": 1820
    },
    {
      "epoch": 11.79,
      "learning_rate": 7.558562892497466e-05,
      "loss": 0.3823,
      "step": 1821
    },
    {
      "epoch": 11.79,
      "learning_rate": 7.548364964129509e-05,
      "loss": 0.4622,
      "step": 1822
    },
    {
      "epoch": 11.8,
      "learning_rate": 7.538169747732788e-05,
      "loss": 0.4321,
      "step": 1823
    },
    {
      "epoch": 11.81,
      "learning_rate": 7.52797725458514e-05,
      "loss": 0.396,
      "step": 1824
    },
    {
      "epoch": 11.81,
      "learning_rate": 7.517787495961385e-05,
      "loss": 0.4026,
      "step": 1825
    },
    {
      "epoch": 11.82,
      "learning_rate": 7.507600483133321e-05,
      "loss": 0.3997,
      "step": 1826
    },
    {
      "epoch": 11.83,
      "learning_rate": 7.497416227369713e-05,
      "loss": 0.522,
      "step": 1827
    },
    {
      "epoch": 11.83,
      "learning_rate": 7.487234739936262e-05,
      "loss": 0.4619,
      "step": 1828
    },
    {
      "epoch": 11.84,
      "learning_rate": 7.477056032095621e-05,
      "loss": 0.4275,
      "step": 1829
    },
    {
      "epoch": 11.84,
      "learning_rate": 7.466880115107358e-05,
      "loss": 0.5508,
      "step": 1830
    },
    {
      "epoch": 11.85,
      "learning_rate": 7.45670700022796e-05,
      "loss": 0.448,
      "step": 1831
    },
    {
      "epoch": 11.86,
      "learning_rate": 7.446536698710816e-05,
      "loss": 0.379,
      "step": 1832
    },
    {
      "epoch": 11.86,
      "learning_rate": 7.436369221806201e-05,
      "loss": 0.4749,
      "step": 1833
    },
    {
      "epoch": 11.87,
      "learning_rate": 7.426204580761256e-05,
      "loss": 0.4519,
      "step": 1834
    },
    {
      "epoch": 11.88,
      "learning_rate": 7.416042786819997e-05,
      "loss": 0.4069,
      "step": 1835
    },
    {
      "epoch": 11.88,
      "learning_rate": 7.40588385122329e-05,
      "loss": 0.4167,
      "step": 1836
    },
    {
      "epoch": 11.89,
      "learning_rate": 7.395727785208835e-05,
      "loss": 0.4336,
      "step": 1837
    },
    {
      "epoch": 11.9,
      "learning_rate": 7.385574600011156e-05,
      "loss": 0.4436,
      "step": 1838
    },
    {
      "epoch": 11.9,
      "learning_rate": 7.375424306861595e-05,
      "loss": 0.5203,
      "step": 1839
    },
    {
      "epoch": 11.91,
      "learning_rate": 7.365276916988289e-05,
      "loss": 0.3975,
      "step": 1840
    },
    {
      "epoch": 11.92,
      "learning_rate": 7.35513244161617e-05,
      "loss": 0.4238,
      "step": 1841
    },
    {
      "epoch": 11.92,
      "learning_rate": 7.344990891966944e-05,
      "loss": 0.4336,
      "step": 1842
    },
    {
      "epoch": 11.93,
      "learning_rate": 7.334852279259076e-05,
      "loss": 0.38,
      "step": 1843
    },
    {
      "epoch": 11.94,
      "learning_rate": 7.324716614707793e-05,
      "loss": 0.3569,
      "step": 1844
    },
    {
      "epoch": 11.94,
      "learning_rate": 7.314583909525047e-05,
      "loss": 0.3867,
      "step": 1845
    },
    {
      "epoch": 11.95,
      "learning_rate": 7.304454174919526e-05,
      "loss": 0.4827,
      "step": 1846
    },
    {
      "epoch": 11.95,
      "learning_rate": 7.294327422096626e-05,
      "loss": 0.3892,
      "step": 1847
    },
    {
      "epoch": 11.96,
      "learning_rate": 7.284203662258451e-05,
      "loss": 0.5027,
      "step": 1848
    },
    {
      "epoch": 11.97,
      "learning_rate": 7.274082906603789e-05,
      "loss": 0.4131,
      "step": 1849
    },
    {
      "epoch": 11.97,
      "learning_rate": 7.263965166328111e-05,
      "loss": 0.4937,
      "step": 1850
    },
    {
      "epoch": 11.98,
      "learning_rate": 7.253850452623541e-05,
      "loss": 0.4902,
      "step": 1851
    },
    {
      "epoch": 11.99,
      "learning_rate": 7.243738776678866e-05,
      "loss": 0.4551,
      "step": 1852
    },
    {
      "epoch": 11.99,
      "learning_rate": 7.233630149679509e-05,
      "loss": 0.3921,
      "step": 1853
    },
    {
      "epoch": 12.0,
      "learning_rate": 7.22352458280752e-05,
      "loss": 0.4302,
      "step": 1854
    },
    {
      "epoch": 12.01,
      "learning_rate": 7.21342208724156e-05,
      "loss": 0.4097,
      "step": 1855
    },
    {
      "epoch": 12.01,
      "learning_rate": 7.203322674156905e-05,
      "loss": 0.4435,
      "step": 1856
    },
    {
      "epoch": 12.02,
      "learning_rate": 7.193226354725404e-05,
      "loss": 0.355,
      "step": 1857
    },
    {
      "epoch": 12.03,
      "learning_rate": 7.183133140115497e-05,
      "loss": 0.4055,
      "step": 1858
    },
    {
      "epoch": 12.03,
      "learning_rate": 7.173043041492182e-05,
      "loss": 0.4351,
      "step": 1859
    },
    {
      "epoch": 12.04,
      "learning_rate": 7.162956070017019e-05,
      "loss": 0.4001,
      "step": 1860
    },
    {
      "epoch": 12.05,
      "learning_rate": 7.152872236848096e-05,
      "loss": 0.3853,
      "step": 1861
    },
    {
      "epoch": 12.05,
      "learning_rate": 7.142791553140045e-05,
      "loss": 0.3757,
      "step": 1862
    },
    {
      "epoch": 12.06,
      "learning_rate": 7.132714030043996e-05,
      "loss": 0.4153,
      "step": 1863
    },
    {
      "epoch": 12.06,
      "learning_rate": 7.122639678707595e-05,
      "loss": 0.3802,
      "step": 1864
    },
    {
      "epoch": 12.07,
      "learning_rate": 7.11256851027498e-05,
      "loss": 0.3174,
      "step": 1865
    },
    {
      "epoch": 12.08,
      "learning_rate": 7.102500535886761e-05,
      "loss": 0.33,
      "step": 1866
    },
    {
      "epoch": 12.08,
      "learning_rate": 7.092435766680023e-05,
      "loss": 0.4014,
      "step": 1867
    },
    {
      "epoch": 12.09,
      "learning_rate": 7.0823742137883e-05,
      "loss": 0.3806,
      "step": 1868
    },
    {
      "epoch": 12.1,
      "learning_rate": 7.072315888341565e-05,
      "loss": 0.343,
      "step": 1869
    },
    {
      "epoch": 12.1,
      "learning_rate": 7.062260801466229e-05,
      "loss": 0.4197,
      "step": 1870
    },
    {
      "epoch": 12.11,
      "learning_rate": 7.052208964285114e-05,
      "loss": 0.3424,
      "step": 1871
    },
    {
      "epoch": 12.12,
      "learning_rate": 7.042160387917451e-05,
      "loss": 0.3035,
      "step": 1872
    },
    {
      "epoch": 12.12,
      "learning_rate": 7.032115083478868e-05,
      "loss": 0.375,
      "step": 1873
    },
    {
      "epoch": 12.13,
      "learning_rate": 7.022073062081358e-05,
      "loss": 0.3612,
      "step": 1874
    },
    {
      "epoch": 12.14,
      "learning_rate": 7.012034334833299e-05,
      "loss": 0.3645,
      "step": 1875
    },
    {
      "epoch": 12.14,
      "learning_rate": 7.001998912839416e-05,
      "loss": 0.3228,
      "step": 1876
    },
    {
      "epoch": 12.15,
      "learning_rate": 6.991966807200785e-05,
      "loss": 0.4045,
      "step": 1877
    },
    {
      "epoch": 12.16,
      "learning_rate": 6.981938029014804e-05,
      "loss": 0.3787,
      "step": 1878
    },
    {
      "epoch": 12.16,
      "learning_rate": 6.971912589375202e-05,
      "loss": 0.3579,
      "step": 1879
    },
    {
      "epoch": 12.17,
      "learning_rate": 6.961890499371999e-05,
      "loss": 0.3274,
      "step": 1880
    },
    {
      "epoch": 12.17,
      "learning_rate": 6.951871770091523e-05,
      "loss": 0.4019,
      "step": 1881
    },
    {
      "epoch": 12.18,
      "learning_rate": 6.941856412616378e-05,
      "loss": 0.3416,
      "step": 1882
    },
    {
      "epoch": 12.19,
      "learning_rate": 6.931844438025447e-05,
      "loss": 0.3506,
      "step": 1883
    },
    {
      "epoch": 12.19,
      "learning_rate": 6.92183585739386e-05,
      "loss": 0.4487,
      "step": 1884
    },
    {
      "epoch": 12.2,
      "learning_rate": 6.911830681792997e-05,
      "loss": 0.394,
      "step": 1885
    },
    {
      "epoch": 12.21,
      "learning_rate": 6.901828922290472e-05,
      "loss": 0.4199,
      "step": 1886
    },
    {
      "epoch": 12.21,
      "learning_rate": 6.891830589950114e-05,
      "loss": 0.3206,
      "step": 1887
    },
    {
      "epoch": 12.22,
      "learning_rate": 6.881835695831973e-05,
      "loss": 0.3828,
      "step": 1888
    },
    {
      "epoch": 12.23,
      "learning_rate": 6.871844250992287e-05,
      "loss": 0.3423,
      "step": 1889
    },
    {
      "epoch": 12.23,
      "learning_rate": 6.86185626648348e-05,
      "loss": 0.3524,
      "step": 1890
    },
    {
      "epoch": 12.24,
      "learning_rate": 6.851871753354153e-05,
      "loss": 0.3721,
      "step": 1891
    },
    {
      "epoch": 12.25,
      "learning_rate": 6.841890722649054e-05,
      "loss": 0.3967,
      "step": 1892
    },
    {
      "epoch": 12.25,
      "learning_rate": 6.831913185409095e-05,
      "loss": 0.3804,
      "step": 1893
    },
    {
      "epoch": 12.26,
      "learning_rate": 6.821939152671311e-05,
      "loss": 0.3783,
      "step": 1894
    },
    {
      "epoch": 12.27,
      "learning_rate": 6.811968635468869e-05,
      "loss": 0.3567,
      "step": 1895
    },
    {
      "epoch": 12.27,
      "learning_rate": 6.80200164483104e-05,
      "loss": 0.3308,
      "step": 1896
    },
    {
      "epoch": 12.28,
      "learning_rate": 6.792038191783205e-05,
      "loss": 0.3772,
      "step": 1897
    },
    {
      "epoch": 12.28,
      "learning_rate": 6.782078287346813e-05,
      "loss": 0.4473,
      "step": 1898
    },
    {
      "epoch": 12.29,
      "learning_rate": 6.772121942539407e-05,
      "loss": 0.3607,
      "step": 1899
    },
    {
      "epoch": 12.3,
      "learning_rate": 6.762169168374584e-05,
      "loss": 0.4115,
      "step": 1900
    },
    {
      "epoch": 12.3,
      "learning_rate": 6.752219975861985e-05,
      "loss": 0.3367,
      "step": 1901
    },
    {
      "epoch": 12.31,
      "learning_rate": 6.742274376007301e-05,
      "loss": 0.3767,
      "step": 1902
    },
    {
      "epoch": 12.32,
      "learning_rate": 6.732332379812246e-05,
      "loss": 0.2892,
      "step": 1903
    },
    {
      "epoch": 12.32,
      "learning_rate": 6.722393998274537e-05,
      "loss": 0.3123,
      "step": 1904
    },
    {
      "epoch": 12.33,
      "learning_rate": 6.712459242387904e-05,
      "loss": 0.37,
      "step": 1905
    },
    {
      "epoch": 12.34,
      "learning_rate": 6.702528123142063e-05,
      "loss": 0.354,
      "step": 1906
    },
    {
      "epoch": 12.34,
      "learning_rate": 6.692600651522707e-05,
      "loss": 0.3182,
      "step": 1907
    },
    {
      "epoch": 12.35,
      "learning_rate": 6.682676838511499e-05,
      "loss": 0.3369,
      "step": 1908
    },
    {
      "epoch": 12.36,
      "learning_rate": 6.67275669508604e-05,
      "loss": 0.3707,
      "step": 1909
    },
    {
      "epoch": 12.36,
      "learning_rate": 6.662840232219888e-05,
      "loss": 0.3655,
      "step": 1910
    },
    {
      "epoch": 12.37,
      "learning_rate": 6.652927460882521e-05,
      "loss": 0.4209,
      "step": 1911
    },
    {
      "epoch": 12.38,
      "learning_rate": 6.643018392039335e-05,
      "loss": 0.4507,
      "step": 1912
    },
    {
      "epoch": 12.38,
      "learning_rate": 6.633113036651634e-05,
      "loss": 0.4316,
      "step": 1913
    },
    {
      "epoch": 12.39,
      "learning_rate": 6.623211405676612e-05,
      "loss": 0.373,
      "step": 1914
    },
    {
      "epoch": 12.39,
      "learning_rate": 6.613313510067338e-05,
      "loss": 0.4363,
      "step": 1915
    },
    {
      "epoch": 12.4,
      "learning_rate": 6.603419360772756e-05,
      "loss": 0.3577,
      "step": 1916
    },
    {
      "epoch": 12.41,
      "learning_rate": 6.593528968737664e-05,
      "loss": 0.3618,
      "step": 1917
    },
    {
      "epoch": 12.41,
      "learning_rate": 6.583642344902698e-05,
      "loss": 0.3491,
      "step": 1918
    },
    {
      "epoch": 12.42,
      "learning_rate": 6.573759500204336e-05,
      "loss": 0.3748,
      "step": 1919
    },
    {
      "epoch": 12.43,
      "learning_rate": 6.563880445574873e-05,
      "loss": 0.3867,
      "step": 1920
    },
    {
      "epoch": 12.43,
      "learning_rate": 6.554005191942399e-05,
      "loss": 0.3912,
      "step": 1921
    },
    {
      "epoch": 12.44,
      "learning_rate": 6.544133750230814e-05,
      "loss": 0.3352,
      "step": 1922
    },
    {
      "epoch": 12.45,
      "learning_rate": 6.534266131359798e-05,
      "loss": 0.3368,
      "step": 1923
    },
    {
      "epoch": 12.45,
      "learning_rate": 6.5244023462448e-05,
      "loss": 0.3561,
      "step": 1924
    },
    {
      "epoch": 12.46,
      "learning_rate": 6.514542405797026e-05,
      "loss": 0.4429,
      "step": 1925
    },
    {
      "epoch": 12.47,
      "learning_rate": 6.504686320923438e-05,
      "loss": 0.4275,
      "step": 1926
    },
    {
      "epoch": 12.47,
      "learning_rate": 6.494834102526719e-05,
      "loss": 0.3613,
      "step": 1927
    },
    {
      "epoch": 12.48,
      "learning_rate": 6.484985761505286e-05,
      "loss": 0.3191,
      "step": 1928
    },
    {
      "epoch": 12.49,
      "learning_rate": 6.475141308753264e-05,
      "loss": 0.3918,
      "step": 1929
    },
    {
      "epoch": 12.49,
      "learning_rate": 6.465300755160473e-05,
      "loss": 0.3462,
      "step": 1930
    },
    {
      "epoch": 12.5,
      "learning_rate": 6.455464111612432e-05,
      "loss": 0.3396,
      "step": 1931
    },
    {
      "epoch": 12.5,
      "learning_rate": 6.445631388990319e-05,
      "loss": 0.3779,
      "step": 1932
    },
    {
      "epoch": 12.51,
      "learning_rate": 6.435802598170984e-05,
      "loss": 0.3787,
      "step": 1933
    },
    {
      "epoch": 12.52,
      "learning_rate": 6.425977750026921e-05,
      "loss": 0.3907,
      "step": 1934
    },
    {
      "epoch": 12.52,
      "learning_rate": 6.416156855426272e-05,
      "loss": 0.3918,
      "step": 1935
    },
    {
      "epoch": 12.53,
      "learning_rate": 6.406339925232799e-05,
      "loss": 0.4312,
      "step": 1936
    },
    {
      "epoch": 12.54,
      "learning_rate": 6.396526970305881e-05,
      "loss": 0.3763,
      "step": 1937
    },
    {
      "epoch": 12.54,
      "learning_rate": 6.386718001500502e-05,
      "loss": 0.3799,
      "step": 1938
    },
    {
      "epoch": 12.55,
      "learning_rate": 6.376913029667226e-05,
      "loss": 0.3655,
      "step": 1939
    },
    {
      "epoch": 12.56,
      "learning_rate": 6.367112065652208e-05,
      "loss": 0.3597,
      "step": 1940
    },
    {
      "epoch": 12.56,
      "learning_rate": 6.357315120297165e-05,
      "loss": 0.3718,
      "step": 1941
    },
    {
      "epoch": 12.57,
      "learning_rate": 6.347522204439364e-05,
      "loss": 0.3953,
      "step": 1942
    },
    {
      "epoch": 12.58,
      "learning_rate": 6.337733328911624e-05,
      "loss": 0.3633,
      "step": 1943
    },
    {
      "epoch": 12.58,
      "learning_rate": 6.327948504542291e-05,
      "loss": 0.5002,
      "step": 1944
    },
    {
      "epoch": 12.59,
      "learning_rate": 6.31816774215522e-05,
      "loss": 0.3682,
      "step": 1945
    },
    {
      "epoch": 12.6,
      "learning_rate": 6.308391052569788e-05,
      "loss": 0.3777,
      "step": 1946
    },
    {
      "epoch": 12.6,
      "learning_rate": 6.298618446600856e-05,
      "loss": 0.3613,
      "step": 1947
    },
    {
      "epoch": 12.61,
      "learning_rate": 6.288849935058775e-05,
      "loss": 0.3835,
      "step": 1948
    },
    {
      "epoch": 12.61,
      "learning_rate": 6.279085528749359e-05,
      "loss": 0.4001,
      "step": 1949
    },
    {
      "epoch": 12.62,
      "learning_rate": 6.269325238473889e-05,
      "loss": 0.3816,
      "step": 1950
    },
    {
      "epoch": 12.63,
      "learning_rate": 6.259569075029083e-05,
      "loss": 0.4019,
      "step": 1951
    },
    {
      "epoch": 12.63,
      "learning_rate": 6.249817049207105e-05,
      "loss": 0.3613,
      "step": 1952
    },
    {
      "epoch": 12.64,
      "learning_rate": 6.240069171795535e-05,
      "loss": 0.4116,
      "step": 1953
    },
    {
      "epoch": 12.65,
      "learning_rate": 6.230325453577366e-05,
      "loss": 0.4434,
      "step": 1954
    },
    {
      "epoch": 12.65,
      "learning_rate": 6.220585905330994e-05,
      "loss": 0.3549,
      "step": 1955
    },
    {
      "epoch": 12.66,
      "learning_rate": 6.21085053783019e-05,
      "loss": 0.3885,
      "step": 1956
    },
    {
      "epoch": 12.67,
      "learning_rate": 6.201119361844117e-05,
      "loss": 0.4919,
      "step": 1957
    },
    {
      "epoch": 12.67,
      "learning_rate": 6.191392388137284e-05,
      "loss": 0.3804,
      "step": 1958
    },
    {
      "epoch": 12.68,
      "learning_rate": 6.181669627469569e-05,
      "loss": 0.3923,
      "step": 1959
    },
    {
      "epoch": 12.69,
      "learning_rate": 6.171951090596175e-05,
      "loss": 0.3579,
      "step": 1960
    },
    {
      "epoch": 12.69,
      "learning_rate": 6.162236788267646e-05,
      "loss": 0.3806,
      "step": 1961
    },
    {
      "epoch": 12.7,
      "learning_rate": 6.152526731229825e-05,
      "loss": 0.3372,
      "step": 1962
    },
    {
      "epoch": 12.71,
      "learning_rate": 6.142820930223874e-05,
      "loss": 0.4482,
      "step": 1963
    },
    {
      "epoch": 12.71,
      "learning_rate": 6.133119395986243e-05,
      "loss": 0.3547,
      "step": 1964
    },
    {
      "epoch": 12.72,
      "learning_rate": 6.123422139248654e-05,
      "loss": 0.3097,
      "step": 1965
    },
    {
      "epoch": 12.72,
      "learning_rate": 6.11372917073811e-05,
      "loss": 0.4216,
      "step": 1966
    },
    {
      "epoch": 12.73,
      "learning_rate": 6.104040501176867e-05,
      "loss": 0.4042,
      "step": 1967
    },
    {
      "epoch": 12.74,
      "learning_rate": 6.0943561412824137e-05,
      "loss": 0.3291,
      "step": 1968
    },
    {
      "epoch": 12.74,
      "learning_rate": 6.0846761017674835e-05,
      "loss": 0.3323,
      "step": 1969
    },
    {
      "epoch": 12.75,
      "learning_rate": 6.075000393340031e-05,
      "loss": 0.3752,
      "step": 1970
    },
    {
      "epoch": 12.76,
      "learning_rate": 6.065329026703217e-05,
      "loss": 0.3672,
      "step": 1971
    },
    {
      "epoch": 12.76,
      "learning_rate": 6.055662012555394e-05,
      "loss": 0.3801,
      "step": 1972
    },
    {
      "epoch": 12.77,
      "learning_rate": 6.04599936159011e-05,
      "loss": 0.4097,
      "step": 1973
    },
    {
      "epoch": 12.78,
      "learning_rate": 6.03634108449608e-05,
      "loss": 0.3995,
      "step": 1974
    },
    {
      "epoch": 12.78,
      "learning_rate": 6.026687191957179e-05,
      "loss": 0.373,
      "step": 1975
    },
    {
      "epoch": 12.79,
      "learning_rate": 6.017037694652436e-05,
      "loss": 0.3862,
      "step": 1976
    },
    {
      "epoch": 12.8,
      "learning_rate": 6.007392603256019e-05,
      "loss": 0.4976,
      "step": 1977
    },
    {
      "epoch": 12.8,
      "learning_rate": 5.9977519284372194e-05,
      "loss": 0.3792,
      "step": 1978
    },
    {
      "epoch": 12.81,
      "learning_rate": 5.988115680860448e-05,
      "loss": 0.4052,
      "step": 1979
    },
    {
      "epoch": 12.82,
      "learning_rate": 5.978483871185204e-05,
      "loss": 0.4385,
      "step": 1980
    },
    {
      "epoch": 12.82,
      "learning_rate": 5.968856510066096e-05,
      "loss": 0.4016,
      "step": 1981
    },
    {
      "epoch": 12.83,
      "learning_rate": 5.9592336081527986e-05,
      "loss": 0.3184,
      "step": 1982
    },
    {
      "epoch": 12.83,
      "learning_rate": 5.94961517609006e-05,
      "loss": 0.3228,
      "step": 1983
    },
    {
      "epoch": 12.84,
      "learning_rate": 5.940001224517685e-05,
      "loss": 0.3242,
      "step": 1984
    },
    {
      "epoch": 12.85,
      "learning_rate": 5.930391764070512e-05,
      "loss": 0.3848,
      "step": 1985
    },
    {
      "epoch": 12.85,
      "learning_rate": 5.920786805378423e-05,
      "loss": 0.3936,
      "step": 1986
    },
    {
      "epoch": 12.86,
      "learning_rate": 5.911186359066316e-05,
      "loss": 0.4314,
      "step": 1987
    },
    {
      "epoch": 12.87,
      "learning_rate": 5.9015904357540985e-05,
      "loss": 0.4077,
      "step": 1988
    },
    {
      "epoch": 12.87,
      "learning_rate": 5.891999046056669e-05,
      "loss": 0.3612,
      "step": 1989
    },
    {
      "epoch": 12.88,
      "learning_rate": 5.882412200583923e-05,
      "loss": 0.4592,
      "step": 1990
    },
    {
      "epoch": 12.89,
      "learning_rate": 5.872829909940714e-05,
      "loss": 0.3755,
      "step": 1991
    },
    {
      "epoch": 12.89,
      "learning_rate": 5.863252184726865e-05,
      "loss": 0.3818,
      "step": 1992
    },
    {
      "epoch": 12.9,
      "learning_rate": 5.853679035537152e-05,
      "loss": 0.4277,
      "step": 1993
    },
    {
      "epoch": 12.91,
      "learning_rate": 5.844110472961285e-05,
      "loss": 0.3342,
      "step": 1994
    },
    {
      "epoch": 12.91,
      "learning_rate": 5.8345465075839024e-05,
      "loss": 0.4075,
      "step": 1995
    },
    {
      "epoch": 12.92,
      "learning_rate": 5.8249871499845554e-05,
      "loss": 0.3644,
      "step": 1996
    },
    {
      "epoch": 12.93,
      "learning_rate": 5.815432410737691e-05,
      "loss": 0.4087,
      "step": 1997
    },
    {
      "epoch": 12.93,
      "learning_rate": 5.805882300412662e-05,
      "loss": 0.4041,
      "step": 1998
    },
    {
      "epoch": 12.94,
      "learning_rate": 5.796336829573692e-05,
      "loss": 0.3845,
      "step": 1999
    },
    {
      "epoch": 12.94,
      "learning_rate": 5.786796008779873e-05,
      "loss": 0.3826,
      "step": 2000
    },
    {
      "epoch": 12.95,
      "learning_rate": 5.777259848585155e-05,
      "loss": 0.4513,
      "step": 2001
    },
    {
      "epoch": 12.96,
      "learning_rate": 5.767728359538336e-05,
      "loss": 0.3792,
      "step": 2002
    },
    {
      "epoch": 12.96,
      "learning_rate": 5.7582015521830354e-05,
      "loss": 0.3718,
      "step": 2003
    },
    {
      "epoch": 12.97,
      "learning_rate": 5.7486794370577026e-05,
      "loss": 0.3723,
      "step": 2004
    },
    {
      "epoch": 12.98,
      "learning_rate": 5.739162024695597e-05,
      "loss": 0.3365,
      "step": 2005
    },
    {
      "epoch": 12.98,
      "learning_rate": 5.7296493256247704e-05,
      "loss": 0.3616,
      "step": 2006
    },
    {
      "epoch": 12.99,
      "learning_rate": 5.720141350368072e-05,
      "loss": 0.3694,
      "step": 2007
    },
    {
      "epoch": 13.0,
      "learning_rate": 5.710638109443107e-05,
      "loss": 0.3801,
      "step": 2008
    },
    {
      "epoch": 13.0,
      "learning_rate": 5.701139613362264e-05,
      "loss": 0.3296,
      "step": 2009
    },
    {
      "epoch": 13.01,
      "learning_rate": 5.691645872632663e-05,
      "loss": 0.3673,
      "step": 2010
    },
    {
      "epoch": 13.02,
      "learning_rate": 5.682156897756179e-05,
      "loss": 0.3126,
      "step": 2011
    },
    {
      "epoch": 13.02,
      "learning_rate": 5.672672699229411e-05,
      "loss": 0.3804,
      "step": 2012
    },
    {
      "epoch": 13.03,
      "learning_rate": 5.663193287543671e-05,
      "loss": 0.286,
      "step": 2013
    },
    {
      "epoch": 13.04,
      "learning_rate": 5.653718673184982e-05,
      "loss": 0.3018,
      "step": 2014
    },
    {
      "epoch": 13.04,
      "learning_rate": 5.64424886663405e-05,
      "loss": 0.3799,
      "step": 2015
    },
    {
      "epoch": 13.05,
      "learning_rate": 5.634783878366272e-05,
      "loss": 0.3342,
      "step": 2016
    },
    {
      "epoch": 13.06,
      "learning_rate": 5.625323718851714e-05,
      "loss": 0.3584,
      "step": 2017
    },
    {
      "epoch": 13.06,
      "learning_rate": 5.615868398555095e-05,
      "loss": 0.3708,
      "step": 2018
    },
    {
      "epoch": 13.07,
      "learning_rate": 5.6064179279357877e-05,
      "loss": 0.3372,
      "step": 2019
    },
    {
      "epoch": 13.07,
      "learning_rate": 5.5969723174477996e-05,
      "loss": 0.3525,
      "step": 2020
    },
    {
      "epoch": 13.08,
      "learning_rate": 5.5875315775397505e-05,
      "loss": 0.3348,
      "step": 2021
    },
    {
      "epoch": 13.09,
      "learning_rate": 5.578095718654888e-05,
      "loss": 0.3745,
      "step": 2022
    },
    {
      "epoch": 13.09,
      "learning_rate": 5.568664751231054e-05,
      "loss": 0.3359,
      "step": 2023
    },
    {
      "epoch": 13.1,
      "learning_rate": 5.559238685700674e-05,
      "loss": 0.3113,
      "step": 2024
    },
    {
      "epoch": 13.11,
      "learning_rate": 5.549817532490759e-05,
      "loss": 0.2878,
      "step": 2025
    },
    {
      "epoch": 13.11,
      "learning_rate": 5.5404013020228864e-05,
      "loss": 0.3104,
      "step": 2026
    },
    {
      "epoch": 13.12,
      "learning_rate": 5.530990004713179e-05,
      "loss": 0.3247,
      "step": 2027
    },
    {
      "epoch": 13.13,
      "learning_rate": 5.521583650972311e-05,
      "loss": 0.3446,
      "step": 2028
    },
    {
      "epoch": 13.13,
      "learning_rate": 5.5121822512054854e-05,
      "loss": 0.2841,
      "step": 2029
    },
    {
      "epoch": 13.14,
      "learning_rate": 5.502785815812427e-05,
      "loss": 0.3004,
      "step": 2030
    },
    {
      "epoch": 13.15,
      "learning_rate": 5.49339435518737e-05,
      "loss": 0.3638,
      "step": 2031
    },
    {
      "epoch": 13.15,
      "learning_rate": 5.4840078797190354e-05,
      "loss": 0.3322,
      "step": 2032
    },
    {
      "epoch": 13.16,
      "learning_rate": 5.474626399790642e-05,
      "loss": 0.2948,
      "step": 2033
    },
    {
      "epoch": 13.17,
      "learning_rate": 5.465249925779877e-05,
      "loss": 0.3604,
      "step": 2034
    },
    {
      "epoch": 13.17,
      "learning_rate": 5.4558784680588926e-05,
      "loss": 0.3345,
      "step": 2035
    },
    {
      "epoch": 13.18,
      "learning_rate": 5.446512036994287e-05,
      "loss": 0.3359,
      "step": 2036
    },
    {
      "epoch": 13.18,
      "learning_rate": 5.4371506429471084e-05,
      "loss": 0.3657,
      "step": 2037
    },
    {
      "epoch": 13.19,
      "learning_rate": 5.427794296272819e-05,
      "loss": 0.3154,
      "step": 2038
    },
    {
      "epoch": 13.2,
      "learning_rate": 5.4184430073213053e-05,
      "loss": 0.3679,
      "step": 2039
    },
    {
      "epoch": 13.2,
      "learning_rate": 5.4090967864368646e-05,
      "loss": 0.3704,
      "step": 2040
    },
    {
      "epoch": 13.21,
      "learning_rate": 5.399755643958174e-05,
      "loss": 0.3234,
      "step": 2041
    },
    {
      "epoch": 13.22,
      "learning_rate": 5.390419590218305e-05,
      "loss": 0.2993,
      "step": 2042
    },
    {
      "epoch": 13.22,
      "learning_rate": 5.381088635544698e-05,
      "loss": 0.3623,
      "step": 2043
    },
    {
      "epoch": 13.23,
      "learning_rate": 5.371762790259144e-05,
      "loss": 0.3298,
      "step": 2044
    },
    {
      "epoch": 13.24,
      "learning_rate": 5.362442064677793e-05,
      "loss": 0.3613,
      "step": 2045
    },
    {
      "epoch": 13.24,
      "learning_rate": 5.3531264691111246e-05,
      "loss": 0.3068,
      "step": 2046
    },
    {
      "epoch": 13.25,
      "learning_rate": 5.343816013863952e-05,
      "loss": 0.3236,
      "step": 2047
    },
    {
      "epoch": 13.26,
      "learning_rate": 5.33451070923539e-05,
      "loss": 0.3972,
      "step": 2048
    },
    {
      "epoch": 13.26,
      "learning_rate": 5.32521056551887e-05,
      "loss": 0.3203,
      "step": 2049
    },
    {
      "epoch": 13.27,
      "learning_rate": 5.3159155930021e-05,
      "loss": 0.3489,
      "step": 2050
    },
    {
      "epoch": 13.28,
      "learning_rate": 5.306625801967076e-05,
      "loss": 0.364,
      "step": 2051
    },
    {
      "epoch": 13.28,
      "learning_rate": 5.297341202690062e-05,
      "loss": 0.3623,
      "step": 2052
    },
    {
      "epoch": 13.29,
      "learning_rate": 5.2880618054415774e-05,
      "loss": 0.3289,
      "step": 2053
    },
    {
      "epoch": 13.29,
      "learning_rate": 5.2787876204863915e-05,
      "loss": 0.3154,
      "step": 2054
    },
    {
      "epoch": 13.3,
      "learning_rate": 5.2695186580834966e-05,
      "loss": 0.3235,
      "step": 2055
    },
    {
      "epoch": 13.31,
      "learning_rate": 5.260254928486122e-05,
      "loss": 0.3199,
      "step": 2056
    },
    {
      "epoch": 13.31,
      "learning_rate": 5.2509964419416955e-05,
      "loss": 0.3291,
      "step": 2057
    },
    {
      "epoch": 13.32,
      "learning_rate": 5.241743208691854e-05,
      "loss": 0.3088,
      "step": 2058
    },
    {
      "epoch": 13.33,
      "learning_rate": 5.2324952389724216e-05,
      "loss": 0.4086,
      "step": 2059
    },
    {
      "epoch": 13.33,
      "learning_rate": 5.2232525430134e-05,
      "loss": 0.3848,
      "step": 2060
    },
    {
      "epoch": 13.34,
      "learning_rate": 5.2140151310389516e-05,
      "loss": 0.298,
      "step": 2061
    },
    {
      "epoch": 13.35,
      "learning_rate": 5.204783013267401e-05,
      "loss": 0.3389,
      "step": 2062
    },
    {
      "epoch": 13.35,
      "learning_rate": 5.195556199911214e-05,
      "loss": 0.3431,
      "step": 2063
    },
    {
      "epoch": 13.36,
      "learning_rate": 5.186334701176986e-05,
      "loss": 0.4155,
      "step": 2064
    },
    {
      "epoch": 13.37,
      "learning_rate": 5.177118527265438e-05,
      "loss": 0.3778,
      "step": 2065
    },
    {
      "epoch": 13.37,
      "learning_rate": 5.1679076883714006e-05,
      "loss": 0.3328,
      "step": 2066
    },
    {
      "epoch": 13.38,
      "learning_rate": 5.1587021946837955e-05,
      "loss": 0.334,
      "step": 2067
    },
    {
      "epoch": 13.39,
      "learning_rate": 5.149502056385638e-05,
      "loss": 0.2888,
      "step": 2068
    },
    {
      "epoch": 13.39,
      "learning_rate": 5.140307283654019e-05,
      "loss": 0.4131,
      "step": 2069
    },
    {
      "epoch": 13.4,
      "learning_rate": 5.131117886660094e-05,
      "loss": 0.3782,
      "step": 2070
    },
    {
      "epoch": 13.4,
      "learning_rate": 5.121933875569074e-05,
      "loss": 0.3289,
      "step": 2071
    },
    {
      "epoch": 13.41,
      "learning_rate": 5.112755260540202e-05,
      "loss": 0.3816,
      "step": 2072
    },
    {
      "epoch": 13.42,
      "learning_rate": 5.1035820517267664e-05,
      "loss": 0.3201,
      "step": 2073
    },
    {
      "epoch": 13.42,
      "learning_rate": 5.0944142592760616e-05,
      "loss": 0.3423,
      "step": 2074
    },
    {
      "epoch": 13.43,
      "learning_rate": 5.0852518933293994e-05,
      "loss": 0.3599,
      "step": 2075
    },
    {
      "epoch": 13.44,
      "learning_rate": 5.0760949640220866e-05,
      "loss": 0.343,
      "step": 2076
    },
    {
      "epoch": 13.44,
      "learning_rate": 5.066943481483418e-05,
      "loss": 0.2974,
      "step": 2077
    },
    {
      "epoch": 13.45,
      "learning_rate": 5.0577974558366624e-05,
      "loss": 0.3715,
      "step": 2078
    },
    {
      "epoch": 13.46,
      "learning_rate": 5.0486568971990444e-05,
      "loss": 0.3569,
      "step": 2079
    },
    {
      "epoch": 13.46,
      "learning_rate": 5.039521815681752e-05,
      "loss": 0.3046,
      "step": 2080
    },
    {
      "epoch": 13.47,
      "learning_rate": 5.03039222138991e-05,
      "loss": 0.3602,
      "step": 2081
    },
    {
      "epoch": 13.48,
      "learning_rate": 5.021268124422574e-05,
      "loss": 0.3597,
      "step": 2082
    },
    {
      "epoch": 13.48,
      "learning_rate": 5.012149534872714e-05,
      "loss": 0.3455,
      "step": 2083
    },
    {
      "epoch": 13.49,
      "learning_rate": 5.003036462827221e-05,
      "loss": 0.3668,
      "step": 2084
    },
    {
      "epoch": 13.5,
      "learning_rate": 4.993928918366861e-05,
      "loss": 0.3535,
      "step": 2085
    },
    {
      "epoch": 13.5,
      "learning_rate": 4.984826911566306e-05,
      "loss": 0.3323,
      "step": 2086
    },
    {
      "epoch": 13.51,
      "learning_rate": 4.9757304524940954e-05,
      "loss": 0.3523,
      "step": 2087
    },
    {
      "epoch": 13.51,
      "learning_rate": 4.966639551212622e-05,
      "loss": 0.3723,
      "step": 2088
    },
    {
      "epoch": 13.52,
      "learning_rate": 4.957554217778144e-05,
      "loss": 0.351,
      "step": 2089
    },
    {
      "epoch": 13.53,
      "learning_rate": 4.94847446224076e-05,
      "loss": 0.3457,
      "step": 2090
    },
    {
      "epoch": 13.53,
      "learning_rate": 4.939400294644385e-05,
      "loss": 0.3821,
      "step": 2091
    },
    {
      "epoch": 13.54,
      "learning_rate": 4.930331725026767e-05,
      "loss": 0.3772,
      "step": 2092
    },
    {
      "epoch": 13.55,
      "learning_rate": 4.921268763419454e-05,
      "loss": 0.3525,
      "step": 2093
    },
    {
      "epoch": 13.55,
      "learning_rate": 4.912211419847794e-05,
      "loss": 0.3862,
      "step": 2094
    },
    {
      "epoch": 13.56,
      "learning_rate": 4.9031597043309206e-05,
      "loss": 0.3691,
      "step": 2095
    },
    {
      "epoch": 13.57,
      "learning_rate": 4.894113626881741e-05,
      "loss": 0.3877,
      "step": 2096
    },
    {
      "epoch": 13.57,
      "learning_rate": 4.8850731975069195e-05,
      "loss": 0.316,
      "step": 2097
    },
    {
      "epoch": 13.58,
      "learning_rate": 4.87603842620688e-05,
      "loss": 0.3381,
      "step": 2098
    },
    {
      "epoch": 13.59,
      "learning_rate": 4.867009322975787e-05,
      "loss": 0.332,
      "step": 2099
    },
    {
      "epoch": 13.59,
      "learning_rate": 4.857985897801533e-05,
      "loss": 0.3235,
      "step": 2100
    },
    {
      "epoch": 13.6,
      "learning_rate": 4.8489681606657335e-05,
      "loss": 0.3485,
      "step": 2101
    },
    {
      "epoch": 13.61,
      "learning_rate": 4.839956121543699e-05,
      "loss": 0.3008,
      "step": 2102
    },
    {
      "epoch": 13.61,
      "learning_rate": 4.830949790404454e-05,
      "loss": 0.3192,
      "step": 2103
    },
    {
      "epoch": 13.62,
      "learning_rate": 4.821949177210702e-05,
      "loss": 0.391,
      "step": 2104
    },
    {
      "epoch": 13.62,
      "learning_rate": 4.812954291918814e-05,
      "loss": 0.3184,
      "step": 2105
    },
    {
      "epoch": 13.63,
      "learning_rate": 4.803965144478835e-05,
      "loss": 0.342,
      "step": 2106
    },
    {
      "epoch": 13.64,
      "learning_rate": 4.794981744834463e-05,
      "loss": 0.3123,
      "step": 2107
    },
    {
      "epoch": 13.64,
      "learning_rate": 4.7860041029230265e-05,
      "loss": 0.3223,
      "step": 2108
    },
    {
      "epoch": 13.65,
      "learning_rate": 4.7770322286754974e-05,
      "loss": 0.3232,
      "step": 2109
    },
    {
      "epoch": 13.66,
      "learning_rate": 4.768066132016461e-05,
      "loss": 0.3335,
      "step": 2110
    },
    {
      "epoch": 13.66,
      "learning_rate": 4.759105822864114e-05,
      "loss": 0.3687,
      "step": 2111
    },
    {
      "epoch": 13.67,
      "learning_rate": 4.7501513111302496e-05,
      "loss": 0.3342,
      "step": 2112
    },
    {
      "epoch": 13.68,
      "learning_rate": 4.7412026067202534e-05,
      "loss": 0.3171,
      "step": 2113
    },
    {
      "epoch": 13.68,
      "learning_rate": 4.732259719533073e-05,
      "loss": 0.3123,
      "step": 2114
    },
    {
      "epoch": 13.69,
      "learning_rate": 4.723322659461234e-05,
      "loss": 0.3099,
      "step": 2115
    },
    {
      "epoch": 13.7,
      "learning_rate": 4.7143914363908135e-05,
      "loss": 0.3707,
      "step": 2116
    },
    {
      "epoch": 13.7,
      "learning_rate": 4.70546606020143e-05,
      "loss": 0.3044,
      "step": 2117
    },
    {
      "epoch": 13.71,
      "learning_rate": 4.696546540766238e-05,
      "loss": 0.343,
      "step": 2118
    },
    {
      "epoch": 13.72,
      "learning_rate": 4.6876328879519026e-05,
      "loss": 0.3472,
      "step": 2119
    },
    {
      "epoch": 13.72,
      "learning_rate": 4.678725111618616e-05,
      "loss": 0.3867,
      "step": 2120
    },
    {
      "epoch": 13.73,
      "learning_rate": 4.66982322162005e-05,
      "loss": 0.3931,
      "step": 2121
    },
    {
      "epoch": 13.73,
      "learning_rate": 4.660927227803381e-05,
      "loss": 0.2876,
      "step": 2122
    },
    {
      "epoch": 13.74,
      "learning_rate": 4.652037140009259e-05,
      "loss": 0.3745,
      "step": 2123
    },
    {
      "epoch": 13.75,
      "learning_rate": 4.643152968071798e-05,
      "loss": 0.4106,
      "step": 2124
    },
    {
      "epoch": 13.75,
      "learning_rate": 4.634274721818575e-05,
      "loss": 0.3801,
      "step": 2125
    },
    {
      "epoch": 13.76,
      "learning_rate": 4.625402411070596e-05,
      "loss": 0.334,
      "step": 2126
    },
    {
      "epoch": 13.77,
      "learning_rate": 4.616536045642319e-05,
      "loss": 0.3323,
      "step": 2127
    },
    {
      "epoch": 13.77,
      "learning_rate": 4.607675635341616e-05,
      "loss": 0.3682,
      "step": 2128
    },
    {
      "epoch": 13.78,
      "learning_rate": 4.5988211899697754e-05,
      "loss": 0.312,
      "step": 2129
    },
    {
      "epoch": 13.79,
      "learning_rate": 4.589972719321483e-05,
      "loss": 0.3324,
      "step": 2130
    },
    {
      "epoch": 13.79,
      "learning_rate": 4.581130233184826e-05,
      "loss": 0.3909,
      "step": 2131
    },
    {
      "epoch": 13.8,
      "learning_rate": 4.5722937413412526e-05,
      "loss": 0.3685,
      "step": 2132
    },
    {
      "epoch": 13.81,
      "learning_rate": 4.563463253565595e-05,
      "loss": 0.3828,
      "step": 2133
    },
    {
      "epoch": 13.81,
      "learning_rate": 4.5546387796260416e-05,
      "loss": 0.3306,
      "step": 2134
    },
    {
      "epoch": 13.82,
      "learning_rate": 4.545820329284129e-05,
      "loss": 0.3551,
      "step": 2135
    },
    {
      "epoch": 13.83,
      "learning_rate": 4.537007912294723e-05,
      "loss": 0.3629,
      "step": 2136
    },
    {
      "epoch": 13.83,
      "learning_rate": 4.528201538406025e-05,
      "loss": 0.3882,
      "step": 2137
    },
    {
      "epoch": 13.84,
      "learning_rate": 4.5194012173595426e-05,
      "loss": 0.3136,
      "step": 2138
    },
    {
      "epoch": 13.84,
      "learning_rate": 4.510606958890094e-05,
      "loss": 0.4578,
      "step": 2139
    },
    {
      "epoch": 13.85,
      "learning_rate": 4.50181877272579e-05,
      "loss": 0.3972,
      "step": 2140
    },
    {
      "epoch": 13.86,
      "learning_rate": 4.493036668588025e-05,
      "loss": 0.3743,
      "step": 2141
    },
    {
      "epoch": 13.86,
      "learning_rate": 4.484260656191467e-05,
      "loss": 0.2885,
      "step": 2142
    },
    {
      "epoch": 13.87,
      "learning_rate": 4.4754907452440365e-05,
      "loss": 0.3002,
      "step": 2143
    },
    {
      "epoch": 13.88,
      "learning_rate": 4.466726945446912e-05,
      "loss": 0.342,
      "step": 2144
    },
    {
      "epoch": 13.88,
      "learning_rate": 4.457969266494513e-05,
      "loss": 0.3826,
      "step": 2145
    },
    {
      "epoch": 13.89,
      "learning_rate": 4.4492177180744855e-05,
      "loss": 0.3325,
      "step": 2146
    },
    {
      "epoch": 13.9,
      "learning_rate": 4.4404723098676934e-05,
      "loss": 0.3701,
      "step": 2147
    },
    {
      "epoch": 13.9,
      "learning_rate": 4.4317330515482146e-05,
      "loss": 0.4097,
      "step": 2148
    },
    {
      "epoch": 13.91,
      "learning_rate": 4.422999952783311e-05,
      "loss": 0.3518,
      "step": 2149
    },
    {
      "epoch": 13.92,
      "learning_rate": 4.414273023233442e-05,
      "loss": 0.3657,
      "step": 2150
    },
    {
      "epoch": 13.92,
      "learning_rate": 4.405552272552241e-05,
      "loss": 0.4136,
      "step": 2151
    },
    {
      "epoch": 13.93,
      "learning_rate": 4.3968377103865024e-05,
      "loss": 0.3093,
      "step": 2152
    },
    {
      "epoch": 13.94,
      "learning_rate": 4.388129346376178e-05,
      "loss": 0.3882,
      "step": 2153
    },
    {
      "epoch": 13.94,
      "learning_rate": 4.379427190154365e-05,
      "loss": 0.3376,
      "step": 2154
    },
    {
      "epoch": 13.95,
      "learning_rate": 4.370731251347285e-05,
      "loss": 0.4253,
      "step": 2155
    },
    {
      "epoch": 13.95,
      "learning_rate": 4.362041539574294e-05,
      "loss": 0.4419,
      "step": 2156
    },
    {
      "epoch": 13.96,
      "learning_rate": 4.3533580644478514e-05,
      "loss": 0.4175,
      "step": 2157
    },
    {
      "epoch": 13.97,
      "learning_rate": 4.344680835573521e-05,
      "loss": 0.3909,
      "step": 2158
    },
    {
      "epoch": 13.97,
      "learning_rate": 4.336009862549956e-05,
      "loss": 0.4578,
      "step": 2159
    },
    {
      "epoch": 13.98,
      "learning_rate": 4.3273451549688934e-05,
      "loss": 0.3132,
      "step": 2160
    },
    {
      "epoch": 13.99,
      "learning_rate": 4.318686722415129e-05,
      "loss": 0.3879,
      "step": 2161
    },
    {
      "epoch": 13.99,
      "learning_rate": 4.310034574466528e-05,
      "loss": 0.4131,
      "step": 2162
    },
    {
      "epoch": 14.0,
      "learning_rate": 4.3013887206939984e-05,
      "loss": 0.3235,
      "step": 2163
    },
    {
      "epoch": 14.01,
      "learning_rate": 4.292749170661486e-05,
      "loss": 0.2832,
      "step": 2164
    },
    {
      "epoch": 14.01,
      "learning_rate": 4.284115933925971e-05,
      "loss": 0.3484,
      "step": 2165
    },
    {
      "epoch": 14.02,
      "learning_rate": 4.275489020037434e-05,
      "loss": 0.2834,
      "step": 2166
    },
    {
      "epoch": 14.03,
      "learning_rate": 4.2668684385388746e-05,
      "loss": 0.3062,
      "step": 2167
    },
    {
      "epoch": 14.03,
      "learning_rate": 4.258254198966286e-05,
      "loss": 0.2915,
      "step": 2168
    },
    {
      "epoch": 14.04,
      "learning_rate": 4.2496463108486395e-05,
      "loss": 0.2644,
      "step": 2169
    },
    {
      "epoch": 14.05,
      "learning_rate": 4.241044783707885e-05,
      "loss": 0.3315,
      "step": 2170
    },
    {
      "epoch": 14.05,
      "learning_rate": 4.2324496270589364e-05,
      "loss": 0.3138,
      "step": 2171
    },
    {
      "epoch": 14.06,
      "learning_rate": 4.223860850409664e-05,
      "loss": 0.2728,
      "step": 2172
    },
    {
      "epoch": 14.06,
      "learning_rate": 4.21527846326087e-05,
      "loss": 0.3435,
      "step": 2173
    },
    {
      "epoch": 14.07,
      "learning_rate": 4.2067024751062964e-05,
      "loss": 0.3198,
      "step": 2174
    },
    {
      "epoch": 14.08,
      "learning_rate": 4.198132895432607e-05,
      "loss": 0.3699,
      "step": 2175
    },
    {
      "epoch": 14.08,
      "learning_rate": 4.189569733719375e-05,
      "loss": 0.3046,
      "step": 2176
    },
    {
      "epoch": 14.09,
      "learning_rate": 4.181012999439077e-05,
      "loss": 0.2736,
      "step": 2177
    },
    {
      "epoch": 14.1,
      "learning_rate": 4.1724627020570694e-05,
      "loss": 0.3423,
      "step": 2178
    },
    {
      "epoch": 14.1,
      "learning_rate": 4.1639188510315985e-05,
      "loss": 0.2719,
      "step": 2179
    },
    {
      "epoch": 14.11,
      "learning_rate": 4.1553814558137785e-05,
      "loss": 0.2598,
      "step": 2180
    },
    {
      "epoch": 14.12,
      "learning_rate": 4.146850525847579e-05,
      "loss": 0.3214,
      "step": 2181
    },
    {
      "epoch": 14.12,
      "learning_rate": 4.138326070569822e-05,
      "loss": 0.272,
      "step": 2182
    },
    {
      "epoch": 14.13,
      "learning_rate": 4.129808099410159e-05,
      "loss": 0.3169,
      "step": 2183
    },
    {
      "epoch": 14.14,
      "learning_rate": 4.1212966217910834e-05,
      "loss": 0.2848,
      "step": 2184
    },
    {
      "epoch": 14.14,
      "learning_rate": 4.1127916471278846e-05,
      "loss": 0.241,
      "step": 2185
    },
    {
      "epoch": 14.15,
      "learning_rate": 4.104293184828677e-05,
      "loss": 0.3047,
      "step": 2186
    },
    {
      "epoch": 14.16,
      "learning_rate": 4.095801244294364e-05,
      "loss": 0.3291,
      "step": 2187
    },
    {
      "epoch": 14.16,
      "learning_rate": 4.087315834918635e-05,
      "loss": 0.3103,
      "step": 2188
    },
    {
      "epoch": 14.17,
      "learning_rate": 4.078836966087959e-05,
      "loss": 0.27,
      "step": 2189
    },
    {
      "epoch": 14.17,
      "learning_rate": 4.070364647181557e-05,
      "loss": 0.3452,
      "step": 2190
    },
    {
      "epoch": 14.18,
      "learning_rate": 4.061898887571418e-05,
      "loss": 0.2574,
      "step": 2191
    },
    {
      "epoch": 14.19,
      "learning_rate": 4.053439696622272e-05,
      "loss": 0.2987,
      "step": 2192
    },
    {
      "epoch": 14.19,
      "learning_rate": 4.0449870836915794e-05,
      "loss": 0.4028,
      "step": 2193
    },
    {
      "epoch": 14.2,
      "learning_rate": 4.036541058129526e-05,
      "loss": 0.2786,
      "step": 2194
    },
    {
      "epoch": 14.21,
      "learning_rate": 4.028101629279015e-05,
      "loss": 0.306,
      "step": 2195
    },
    {
      "epoch": 14.21,
      "learning_rate": 4.0196688064756416e-05,
      "loss": 0.3239,
      "step": 2196
    },
    {
      "epoch": 14.22,
      "learning_rate": 4.011242599047702e-05,
      "loss": 0.3152,
      "step": 2197
    },
    {
      "epoch": 14.23,
      "learning_rate": 4.002823016316178e-05,
      "loss": 0.2831,
      "step": 2198
    },
    {
      "epoch": 14.23,
      "learning_rate": 3.9944100675947094e-05,
      "loss": 0.307,
      "step": 2199
    },
    {
      "epoch": 14.24,
      "learning_rate": 3.98600376218961e-05,
      "loss": 0.2634,
      "step": 2200
    },
    {
      "epoch": 14.25,
      "learning_rate": 3.977604109399843e-05,
      "loss": 0.2759,
      "step": 2201
    },
    {
      "epoch": 14.25,
      "learning_rate": 3.9692111185170044e-05,
      "loss": 0.3152,
      "step": 2202
    },
    {
      "epoch": 14.26,
      "learning_rate": 3.960824798825331e-05,
      "loss": 0.3678,
      "step": 2203
    },
    {
      "epoch": 14.27,
      "learning_rate": 3.952445159601674e-05,
      "loss": 0.2909,
      "step": 2204
    },
    {
      "epoch": 14.27,
      "learning_rate": 3.944072210115498e-05,
      "loss": 0.2699,
      "step": 2205
    },
    {
      "epoch": 14.28,
      "learning_rate": 3.935705959628866e-05,
      "loss": 0.3049,
      "step": 2206
    },
    {
      "epoch": 14.28,
      "learning_rate": 3.9273464173964336e-05,
      "loss": 0.3149,
      "step": 2207
    },
    {
      "epoch": 14.29,
      "learning_rate": 3.918993592665425e-05,
      "loss": 0.3145,
      "step": 2208
    },
    {
      "epoch": 14.3,
      "learning_rate": 3.9106474946756465e-05,
      "loss": 0.3318,
      "step": 2209
    },
    {
      "epoch": 14.3,
      "learning_rate": 3.902308132659457e-05,
      "loss": 0.3372,
      "step": 2210
    },
    {
      "epoch": 14.31,
      "learning_rate": 3.8939755158417665e-05,
      "loss": 0.3115,
      "step": 2211
    },
    {
      "epoch": 14.32,
      "learning_rate": 3.8856496534400254e-05,
      "loss": 0.324,
      "step": 2212
    },
    {
      "epoch": 14.32,
      "learning_rate": 3.8773305546642034e-05,
      "loss": 0.3562,
      "step": 2213
    },
    {
      "epoch": 14.33,
      "learning_rate": 3.869018228716796e-05,
      "loss": 0.3602,
      "step": 2214
    },
    {
      "epoch": 14.34,
      "learning_rate": 3.860712684792811e-05,
      "loss": 0.319,
      "step": 2215
    },
    {
      "epoch": 14.34,
      "learning_rate": 3.852413932079739e-05,
      "loss": 0.3293,
      "step": 2216
    },
    {
      "epoch": 14.35,
      "learning_rate": 3.844121979757572e-05,
      "loss": 0.3403,
      "step": 2217
    },
    {
      "epoch": 14.36,
      "learning_rate": 3.835836836998779e-05,
      "loss": 0.287,
      "step": 2218
    },
    {
      "epoch": 14.36,
      "learning_rate": 3.8275585129682834e-05,
      "loss": 0.3179,
      "step": 2219
    },
    {
      "epoch": 14.37,
      "learning_rate": 3.819287016823479e-05,
      "loss": 0.2887,
      "step": 2220
    },
    {
      "epoch": 14.38,
      "learning_rate": 3.8110223577142036e-05,
      "loss": 0.375,
      "step": 2221
    },
    {
      "epoch": 14.38,
      "learning_rate": 3.802764544782727e-05,
      "loss": 0.3484,
      "step": 2222
    },
    {
      "epoch": 14.39,
      "learning_rate": 3.794513587163754e-05,
      "loss": 0.4126,
      "step": 2223
    },
    {
      "epoch": 14.39,
      "learning_rate": 3.7862694939844024e-05,
      "loss": 0.3335,
      "step": 2224
    },
    {
      "epoch": 14.4,
      "learning_rate": 3.7780322743641885e-05,
      "loss": 0.3369,
      "step": 2225
    },
    {
      "epoch": 14.41,
      "learning_rate": 3.769801937415037e-05,
      "loss": 0.3063,
      "step": 2226
    },
    {
      "epoch": 14.41,
      "learning_rate": 3.7615784922412547e-05,
      "loss": 0.3669,
      "step": 2227
    },
    {
      "epoch": 14.42,
      "learning_rate": 3.7533619479395245e-05,
      "loss": 0.3296,
      "step": 2228
    },
    {
      "epoch": 14.43,
      "learning_rate": 3.745152313598897e-05,
      "loss": 0.3796,
      "step": 2229
    },
    {
      "epoch": 14.43,
      "learning_rate": 3.736949598300774e-05,
      "loss": 0.2866,
      "step": 2230
    },
    {
      "epoch": 14.44,
      "learning_rate": 3.7287538111189136e-05,
      "loss": 0.2844,
      "step": 2231
    },
    {
      "epoch": 14.45,
      "learning_rate": 3.720564961119396e-05,
      "loss": 0.2798,
      "step": 2232
    },
    {
      "epoch": 14.45,
      "learning_rate": 3.712383057360639e-05,
      "loss": 0.4103,
      "step": 2233
    },
    {
      "epoch": 14.46,
      "learning_rate": 3.704208108893373e-05,
      "loss": 0.3184,
      "step": 2234
    },
    {
      "epoch": 14.47,
      "learning_rate": 3.696040124760635e-05,
      "loss": 0.3914,
      "step": 2235
    },
    {
      "epoch": 14.47,
      "learning_rate": 3.687879113997761e-05,
      "loss": 0.3225,
      "step": 2236
    },
    {
      "epoch": 14.48,
      "learning_rate": 3.6797250856323615e-05,
      "loss": 0.3411,
      "step": 2237
    },
    {
      "epoch": 14.49,
      "learning_rate": 3.671578048684338e-05,
      "loss": 0.2867,
      "step": 2238
    },
    {
      "epoch": 14.49,
      "learning_rate": 3.663438012165848e-05,
      "loss": 0.3368,
      "step": 2239
    },
    {
      "epoch": 14.5,
      "learning_rate": 3.6553049850813114e-05,
      "loss": 0.2974,
      "step": 2240
    },
    {
      "epoch": 14.5,
      "learning_rate": 3.64717897642739e-05,
      "loss": 0.307,
      "step": 2241
    },
    {
      "epoch": 14.51,
      "learning_rate": 3.639059995192988e-05,
      "loss": 0.3022,
      "step": 2242
    },
    {
      "epoch": 14.52,
      "learning_rate": 3.630948050359225e-05,
      "loss": 0.3623,
      "step": 2243
    },
    {
      "epoch": 14.52,
      "learning_rate": 3.622843150899446e-05,
      "loss": 0.3037,
      "step": 2244
    },
    {
      "epoch": 14.53,
      "learning_rate": 3.614745305779201e-05,
      "loss": 0.2977,
      "step": 2245
    },
    {
      "epoch": 14.54,
      "learning_rate": 3.60665452395624e-05,
      "loss": 0.2457,
      "step": 2246
    },
    {
      "epoch": 14.54,
      "learning_rate": 3.5985708143804855e-05,
      "loss": 0.3938,
      "step": 2247
    },
    {
      "epoch": 14.55,
      "learning_rate": 3.590494185994056e-05,
      "loss": 0.3199,
      "step": 2248
    },
    {
      "epoch": 14.56,
      "learning_rate": 3.582424647731219e-05,
      "loss": 0.3003,
      "step": 2249
    },
    {
      "epoch": 14.56,
      "learning_rate": 3.5743622085184106e-05,
      "loss": 0.3364,
      "step": 2250
    },
    {
      "epoch": 14.57,
      "learning_rate": 3.566306877274213e-05,
      "loss": 0.3419,
      "step": 2251
    },
    {
      "epoch": 14.58,
      "learning_rate": 3.5582586629093405e-05,
      "loss": 0.269,
      "step": 2252
    },
    {
      "epoch": 14.58,
      "learning_rate": 3.550217574326644e-05,
      "loss": 0.3269,
      "step": 2253
    },
    {
      "epoch": 14.59,
      "learning_rate": 3.542183620421078e-05,
      "loss": 0.2925,
      "step": 2254
    },
    {
      "epoch": 14.6,
      "learning_rate": 3.534156810079715e-05,
      "loss": 0.3116,
      "step": 2255
    },
    {
      "epoch": 14.6,
      "learning_rate": 3.5261371521817244e-05,
      "loss": 0.3623,
      "step": 2256
    },
    {
      "epoch": 14.61,
      "learning_rate": 3.518124655598359e-05,
      "loss": 0.3239,
      "step": 2257
    },
    {
      "epoch": 14.61,
      "learning_rate": 3.5101193291929556e-05,
      "loss": 0.2765,
      "step": 2258
    },
    {
      "epoch": 14.62,
      "learning_rate": 3.50212118182092e-05,
      "loss": 0.3234,
      "step": 2259
    },
    {
      "epoch": 14.63,
      "learning_rate": 3.4941302223297044e-05,
      "loss": 0.3754,
      "step": 2260
    },
    {
      "epoch": 14.63,
      "learning_rate": 3.486146459558824e-05,
      "loss": 0.3276,
      "step": 2261
    },
    {
      "epoch": 14.64,
      "learning_rate": 3.4781699023398306e-05,
      "loss": 0.2959,
      "step": 2262
    },
    {
      "epoch": 14.65,
      "learning_rate": 3.4702005594962964e-05,
      "loss": 0.3007,
      "step": 2263
    },
    {
      "epoch": 14.65,
      "learning_rate": 3.462238439843821e-05,
      "loss": 0.314,
      "step": 2264
    },
    {
      "epoch": 14.66,
      "learning_rate": 3.454283552190018e-05,
      "loss": 0.3054,
      "step": 2265
    },
    {
      "epoch": 14.67,
      "learning_rate": 3.4463359053344855e-05,
      "loss": 0.283,
      "step": 2266
    },
    {
      "epoch": 14.67,
      "learning_rate": 3.4383955080688255e-05,
      "loss": 0.3264,
      "step": 2267
    },
    {
      "epoch": 14.68,
      "learning_rate": 3.430462369176619e-05,
      "loss": 0.3494,
      "step": 2268
    },
    {
      "epoch": 14.69,
      "learning_rate": 3.4225364974334126e-05,
      "loss": 0.3149,
      "step": 2269
    },
    {
      "epoch": 14.69,
      "learning_rate": 3.414617901606719e-05,
      "loss": 0.3037,
      "step": 2270
    },
    {
      "epoch": 14.7,
      "learning_rate": 3.4067065904560025e-05,
      "loss": 0.3342,
      "step": 2271
    },
    {
      "epoch": 14.71,
      "learning_rate": 3.39880257273266e-05,
      "loss": 0.3425,
      "step": 2272
    },
    {
      "epoch": 14.71,
      "learning_rate": 3.390905857180031e-05,
      "loss": 0.3652,
      "step": 2273
    },
    {
      "epoch": 14.72,
      "learning_rate": 3.383016452533375e-05,
      "loss": 0.3097,
      "step": 2274
    },
    {
      "epoch": 14.72,
      "learning_rate": 3.375134367519862e-05,
      "loss": 0.3701,
      "step": 2275
    },
    {
      "epoch": 14.73,
      "learning_rate": 3.367259610858566e-05,
      "loss": 0.3812,
      "step": 2276
    },
    {
      "epoch": 14.74,
      "learning_rate": 3.359392191260459e-05,
      "loss": 0.3083,
      "step": 2277
    },
    {
      "epoch": 14.74,
      "learning_rate": 3.351532117428385e-05,
      "loss": 0.3477,
      "step": 2278
    },
    {
      "epoch": 14.75,
      "learning_rate": 3.343679398057077e-05,
      "loss": 0.3936,
      "step": 2279
    },
    {
      "epoch": 14.76,
      "learning_rate": 3.335834041833118e-05,
      "loss": 0.3062,
      "step": 2280
    },
    {
      "epoch": 14.76,
      "learning_rate": 3.3279960574349566e-05,
      "loss": 0.3368,
      "step": 2281
    },
    {
      "epoch": 14.77,
      "learning_rate": 3.320165453532884e-05,
      "loss": 0.3333,
      "step": 2282
    },
    {
      "epoch": 14.78,
      "learning_rate": 3.312342238789029e-05,
      "loss": 0.2677,
      "step": 2283
    },
    {
      "epoch": 14.78,
      "learning_rate": 3.304526421857339e-05,
      "loss": 0.3191,
      "step": 2284
    },
    {
      "epoch": 14.79,
      "learning_rate": 3.296718011383586e-05,
      "loss": 0.3076,
      "step": 2285
    },
    {
      "epoch": 14.8,
      "learning_rate": 3.288917016005345e-05,
      "loss": 0.3,
      "step": 2286
    },
    {
      "epoch": 14.8,
      "learning_rate": 3.281123444351991e-05,
      "loss": 0.3405,
      "step": 2287
    },
    {
      "epoch": 14.81,
      "learning_rate": 3.273337305044689e-05,
      "loss": 0.34,
      "step": 2288
    },
    {
      "epoch": 14.82,
      "learning_rate": 3.265558606696373e-05,
      "loss": 0.3416,
      "step": 2289
    },
    {
      "epoch": 14.82,
      "learning_rate": 3.2577873579117535e-05,
      "loss": 0.2917,
      "step": 2290
    },
    {
      "epoch": 14.83,
      "learning_rate": 3.2500235672873005e-05,
      "loss": 0.387,
      "step": 2291
    },
    {
      "epoch": 14.83,
      "learning_rate": 3.242267243411232e-05,
      "loss": 0.327,
      "step": 2292
    },
    {
      "epoch": 14.84,
      "learning_rate": 3.23451839486351e-05,
      "loss": 0.3494,
      "step": 2293
    },
    {
      "epoch": 14.85,
      "learning_rate": 3.2267770302158175e-05,
      "loss": 0.2596,
      "step": 2294
    },
    {
      "epoch": 14.85,
      "learning_rate": 3.219043158031572e-05,
      "loss": 0.3109,
      "step": 2295
    },
    {
      "epoch": 14.86,
      "learning_rate": 3.211316786865889e-05,
      "loss": 0.3569,
      "step": 2296
    },
    {
      "epoch": 14.87,
      "learning_rate": 3.203597925265598e-05,
      "loss": 0.3833,
      "step": 2297
    },
    {
      "epoch": 14.87,
      "learning_rate": 3.1958865817692174e-05,
      "loss": 0.3357,
      "step": 2298
    },
    {
      "epoch": 14.88,
      "learning_rate": 3.1881827649069486e-05,
      "loss": 0.3521,
      "step": 2299
    },
    {
      "epoch": 14.89,
      "learning_rate": 3.180486483200671e-05,
      "loss": 0.3372,
      "step": 2300
    },
    {
      "epoch": 14.89,
      "learning_rate": 3.1727977451639176e-05,
      "loss": 0.3132,
      "step": 2301
    },
    {
      "epoch": 14.9,
      "learning_rate": 3.1651165593018904e-05,
      "loss": 0.3812,
      "step": 2302
    },
    {
      "epoch": 14.91,
      "learning_rate": 3.157442934111429e-05,
      "loss": 0.3381,
      "step": 2303
    },
    {
      "epoch": 14.91,
      "learning_rate": 3.1497768780810124e-05,
      "loss": 0.4084,
      "step": 2304
    },
    {
      "epoch": 14.92,
      "learning_rate": 3.1421183996907455e-05,
      "loss": 0.2861,
      "step": 2305
    },
    {
      "epoch": 14.93,
      "learning_rate": 3.134467507412355e-05,
      "loss": 0.3273,
      "step": 2306
    },
    {
      "epoch": 14.93,
      "learning_rate": 3.126824209709166e-05,
      "loss": 0.3651,
      "step": 2307
    },
    {
      "epoch": 14.94,
      "learning_rate": 3.119188515036112e-05,
      "loss": 0.338,
      "step": 2308
    },
    {
      "epoch": 14.94,
      "learning_rate": 3.111560431839713e-05,
      "loss": 0.2926,
      "step": 2309
    },
    {
      "epoch": 14.95,
      "learning_rate": 3.103939968558072e-05,
      "loss": 0.3171,
      "step": 2310
    },
    {
      "epoch": 14.96,
      "learning_rate": 3.096327133620854e-05,
      "loss": 0.3053,
      "step": 2311
    },
    {
      "epoch": 14.96,
      "learning_rate": 3.088721935449299e-05,
      "loss": 0.3154,
      "step": 2312
    },
    {
      "epoch": 14.97,
      "learning_rate": 3.081124382456185e-05,
      "loss": 0.3217,
      "step": 2313
    },
    {
      "epoch": 14.98,
      "learning_rate": 3.0735344830458434e-05,
      "loss": 0.2863,
      "step": 2314
    },
    {
      "epoch": 14.98,
      "learning_rate": 3.065952245614136e-05,
      "loss": 0.3053,
      "step": 2315
    },
    {
      "epoch": 14.99,
      "learning_rate": 3.058377678548449e-05,
      "loss": 0.3195,
      "step": 2316
    },
    {
      "epoch": 15.0,
      "learning_rate": 3.050810790227683e-05,
      "loss": 0.3219,
      "step": 2317
    },
    {
      "epoch": 15.0,
      "learning_rate": 3.04325158902225e-05,
      "loss": 0.2935,
      "step": 2318
    },
    {
      "epoch": 15.01,
      "learning_rate": 3.0357000832940465e-05,
      "loss": 0.3264,
      "step": 2319
    },
    {
      "epoch": 15.02,
      "learning_rate": 3.028156281396466e-05,
      "loss": 0.2257,
      "step": 2320
    },
    {
      "epoch": 15.02,
      "learning_rate": 3.020620191674378e-05,
      "loss": 0.2784,
      "step": 2321
    },
    {
      "epoch": 15.03,
      "learning_rate": 3.0130918224641214e-05,
      "loss": 0.2905,
      "step": 2322
    },
    {
      "epoch": 15.04,
      "learning_rate": 3.0055711820934938e-05,
      "loss": 0.2939,
      "step": 2323
    },
    {
      "epoch": 15.04,
      "learning_rate": 2.9980582788817467e-05,
      "loss": 0.2686,
      "step": 2324
    },
    {
      "epoch": 15.05,
      "learning_rate": 2.990553121139563e-05,
      "loss": 0.2572,
      "step": 2325
    },
    {
      "epoch": 15.06,
      "learning_rate": 2.98305571716907e-05,
      "loss": 0.2915,
      "step": 2326
    },
    {
      "epoch": 15.06,
      "learning_rate": 2.9755660752638072e-05,
      "loss": 0.28,
      "step": 2327
    },
    {
      "epoch": 15.07,
      "learning_rate": 2.968084203708733e-05,
      "loss": 0.3142,
      "step": 2328
    },
    {
      "epoch": 15.07,
      "learning_rate": 2.9606101107802165e-05,
      "loss": 0.2823,
      "step": 2329
    },
    {
      "epoch": 15.08,
      "learning_rate": 2.953143804746007e-05,
      "loss": 0.2664,
      "step": 2330
    },
    {
      "epoch": 15.09,
      "learning_rate": 2.9456852938652525e-05,
      "loss": 0.2751,
      "step": 2331
    },
    {
      "epoch": 15.09,
      "learning_rate": 2.9382345863884754e-05,
      "loss": 0.3354,
      "step": 2332
    },
    {
      "epoch": 15.1,
      "learning_rate": 2.9307916905575628e-05,
      "loss": 0.2754,
      "step": 2333
    },
    {
      "epoch": 15.11,
      "learning_rate": 2.9233566146057633e-05,
      "loss": 0.3579,
      "step": 2334
    },
    {
      "epoch": 15.11,
      "learning_rate": 2.9159293667576805e-05,
      "loss": 0.2659,
      "step": 2335
    },
    {
      "epoch": 15.12,
      "learning_rate": 2.908509955229243e-05,
      "loss": 0.2928,
      "step": 2336
    },
    {
      "epoch": 15.13,
      "learning_rate": 2.9010983882277254e-05,
      "loss": 0.3109,
      "step": 2337
    },
    {
      "epoch": 15.13,
      "learning_rate": 2.8936946739517202e-05,
      "loss": 0.2679,
      "step": 2338
    },
    {
      "epoch": 15.14,
      "learning_rate": 2.8862988205911333e-05,
      "loss": 0.3007,
      "step": 2339
    },
    {
      "epoch": 15.15,
      "learning_rate": 2.8789108363271777e-05,
      "loss": 0.3027,
      "step": 2340
    },
    {
      "epoch": 15.15,
      "learning_rate": 2.8715307293323534e-05,
      "loss": 0.2793,
      "step": 2341
    },
    {
      "epoch": 15.16,
      "learning_rate": 2.8641585077704545e-05,
      "loss": 0.2512,
      "step": 2342
    },
    {
      "epoch": 15.17,
      "learning_rate": 2.8567941797965535e-05,
      "loss": 0.2999,
      "step": 2343
    },
    {
      "epoch": 15.17,
      "learning_rate": 2.8494377535569815e-05,
      "loss": 0.2886,
      "step": 2344
    },
    {
      "epoch": 15.18,
      "learning_rate": 2.8420892371893372e-05,
      "loss": 0.3149,
      "step": 2345
    },
    {
      "epoch": 15.18,
      "learning_rate": 2.8347486388224688e-05,
      "loss": 0.3005,
      "step": 2346
    },
    {
      "epoch": 15.19,
      "learning_rate": 2.8274159665764643e-05,
      "loss": 0.2944,
      "step": 2347
    },
    {
      "epoch": 15.2,
      "learning_rate": 2.8200912285626414e-05,
      "loss": 0.2928,
      "step": 2348
    },
    {
      "epoch": 15.2,
      "learning_rate": 2.8127744328835424e-05,
      "loss": 0.2682,
      "step": 2349
    },
    {
      "epoch": 15.21,
      "learning_rate": 2.8054655876329274e-05,
      "loss": 0.2544,
      "step": 2350
    },
    {
      "epoch": 15.22,
      "learning_rate": 2.798164700895758e-05,
      "loss": 0.2777,
      "step": 2351
    },
    {
      "epoch": 15.22,
      "learning_rate": 2.7908717807481942e-05,
      "loss": 0.2993,
      "step": 2352
    },
    {
      "epoch": 15.23,
      "learning_rate": 2.7835868352575856e-05,
      "loss": 0.3141,
      "step": 2353
    },
    {
      "epoch": 15.24,
      "learning_rate": 2.776309872482451e-05,
      "loss": 0.278,
      "step": 2354
    },
    {
      "epoch": 15.24,
      "learning_rate": 2.769040900472488e-05,
      "loss": 0.337,
      "step": 2355
    },
    {
      "epoch": 15.25,
      "learning_rate": 2.761779927268553e-05,
      "loss": 0.3395,
      "step": 2356
    },
    {
      "epoch": 15.26,
      "learning_rate": 2.7545269609026546e-05,
      "loss": 0.2548,
      "step": 2357
    },
    {
      "epoch": 15.26,
      "learning_rate": 2.747282009397938e-05,
      "loss": 0.3218,
      "step": 2358
    },
    {
      "epoch": 15.27,
      "learning_rate": 2.7400450807686938e-05,
      "loss": 0.2939,
      "step": 2359
    },
    {
      "epoch": 15.28,
      "learning_rate": 2.7328161830203257e-05,
      "loss": 0.3501,
      "step": 2360
    },
    {
      "epoch": 15.28,
      "learning_rate": 2.7255953241493616e-05,
      "loss": 0.3286,
      "step": 2361
    },
    {
      "epoch": 15.29,
      "learning_rate": 2.7183825121434347e-05,
      "loss": 0.2953,
      "step": 2362
    },
    {
      "epoch": 15.29,
      "learning_rate": 2.7111777549812776e-05,
      "loss": 0.278,
      "step": 2363
    },
    {
      "epoch": 15.3,
      "learning_rate": 2.7039810606327155e-05,
      "loss": 0.2714,
      "step": 2364
    },
    {
      "epoch": 15.31,
      "learning_rate": 2.696792437058644e-05,
      "loss": 0.2578,
      "step": 2365
    },
    {
      "epoch": 15.31,
      "learning_rate": 2.6896118922110415e-05,
      "loss": 0.2498,
      "step": 2366
    },
    {
      "epoch": 15.32,
      "learning_rate": 2.6824394340329463e-05,
      "loss": 0.2971,
      "step": 2367
    },
    {
      "epoch": 15.33,
      "learning_rate": 2.6752750704584505e-05,
      "loss": 0.282,
      "step": 2368
    },
    {
      "epoch": 15.33,
      "learning_rate": 2.668118809412694e-05,
      "loss": 0.3076,
      "step": 2369
    },
    {
      "epoch": 15.34,
      "learning_rate": 2.6609706588118567e-05,
      "loss": 0.2472,
      "step": 2370
    },
    {
      "epoch": 15.35,
      "learning_rate": 2.653830626563134e-05,
      "loss": 0.3928,
      "step": 2371
    },
    {
      "epoch": 15.35,
      "learning_rate": 2.6466987205647552e-05,
      "loss": 0.3345,
      "step": 2372
    },
    {
      "epoch": 15.36,
      "learning_rate": 2.6395749487059563e-05,
      "loss": 0.2782,
      "step": 2373
    },
    {
      "epoch": 15.37,
      "learning_rate": 2.6324593188669687e-05,
      "loss": 0.2273,
      "step": 2374
    },
    {
      "epoch": 15.37,
      "learning_rate": 2.6253518389190258e-05,
      "loss": 0.2898,
      "step": 2375
    },
    {
      "epoch": 15.38,
      "learning_rate": 2.6182525167243444e-05,
      "loss": 0.3037,
      "step": 2376
    },
    {
      "epoch": 15.39,
      "learning_rate": 2.611161360136111e-05,
      "loss": 0.2412,
      "step": 2377
    },
    {
      "epoch": 15.39,
      "learning_rate": 2.6040783769984845e-05,
      "loss": 0.2756,
      "step": 2378
    },
    {
      "epoch": 15.4,
      "learning_rate": 2.5970035751465828e-05,
      "loss": 0.2776,
      "step": 2379
    },
    {
      "epoch": 15.4,
      "learning_rate": 2.5899369624064717e-05,
      "loss": 0.3376,
      "step": 2380
    },
    {
      "epoch": 15.41,
      "learning_rate": 2.58287854659516e-05,
      "loss": 0.3247,
      "step": 2381
    },
    {
      "epoch": 15.42,
      "learning_rate": 2.5758283355205913e-05,
      "loss": 0.3154,
      "step": 2382
    },
    {
      "epoch": 15.42,
      "learning_rate": 2.5687863369816235e-05,
      "loss": 0.2687,
      "step": 2383
    },
    {
      "epoch": 15.43,
      "learning_rate": 2.5617525587680402e-05,
      "loss": 0.2529,
      "step": 2384
    },
    {
      "epoch": 15.44,
      "learning_rate": 2.554727008660527e-05,
      "loss": 0.3081,
      "step": 2385
    },
    {
      "epoch": 15.44,
      "learning_rate": 2.5477096944306712e-05,
      "loss": 0.3003,
      "step": 2386
    },
    {
      "epoch": 15.45,
      "learning_rate": 2.5407006238409446e-05,
      "loss": 0.293,
      "step": 2387
    },
    {
      "epoch": 15.46,
      "learning_rate": 2.5336998046447068e-05,
      "loss": 0.2397,
      "step": 2388
    },
    {
      "epoch": 15.46,
      "learning_rate": 2.5267072445861793e-05,
      "loss": 0.3159,
      "step": 2389
    },
    {
      "epoch": 15.47,
      "learning_rate": 2.5197229514004606e-05,
      "loss": 0.2566,
      "step": 2390
    },
    {
      "epoch": 15.48,
      "learning_rate": 2.512746932813491e-05,
      "loss": 0.3108,
      "step": 2391
    },
    {
      "epoch": 15.48,
      "learning_rate": 2.5057791965420684e-05,
      "loss": 0.2373,
      "step": 2392
    },
    {
      "epoch": 15.49,
      "learning_rate": 2.4988197502938227e-05,
      "loss": 0.301,
      "step": 2393
    },
    {
      "epoch": 15.5,
      "learning_rate": 2.4918686017672188e-05,
      "loss": 0.2764,
      "step": 2394
    },
    {
      "epoch": 15.5,
      "learning_rate": 2.4849257586515352e-05,
      "loss": 0.327,
      "step": 2395
    },
    {
      "epoch": 15.51,
      "learning_rate": 2.4779912286268682e-05,
      "loss": 0.2726,
      "step": 2396
    },
    {
      "epoch": 15.51,
      "learning_rate": 2.471065019364118e-05,
      "loss": 0.3442,
      "step": 2397
    },
    {
      "epoch": 15.52,
      "learning_rate": 2.4641471385249804e-05,
      "loss": 0.2847,
      "step": 2398
    },
    {
      "epoch": 15.53,
      "learning_rate": 2.4572375937619364e-05,
      "loss": 0.2974,
      "step": 2399
    },
    {
      "epoch": 15.53,
      "learning_rate": 2.450336392718251e-05,
      "loss": 0.282,
      "step": 2400
    },
    {
      "epoch": 15.54,
      "learning_rate": 2.44344354302795e-05,
      "loss": 0.3081,
      "step": 2401
    },
    {
      "epoch": 15.55,
      "learning_rate": 2.43655905231583e-05,
      "loss": 0.2628,
      "step": 2402
    },
    {
      "epoch": 15.55,
      "learning_rate": 2.4296829281974355e-05,
      "loss": 0.3499,
      "step": 2403
    },
    {
      "epoch": 15.56,
      "learning_rate": 2.422815178279062e-05,
      "loss": 0.3511,
      "step": 2404
    },
    {
      "epoch": 15.57,
      "learning_rate": 2.4159558101577316e-05,
      "loss": 0.3186,
      "step": 2405
    },
    {
      "epoch": 15.57,
      "learning_rate": 2.409104831421207e-05,
      "loss": 0.3574,
      "step": 2406
    },
    {
      "epoch": 15.58,
      "learning_rate": 2.4022622496479586e-05,
      "loss": 0.3176,
      "step": 2407
    },
    {
      "epoch": 15.59,
      "learning_rate": 2.3954280724071753e-05,
      "loss": 0.3427,
      "step": 2408
    },
    {
      "epoch": 15.59,
      "learning_rate": 2.3886023072587492e-05,
      "loss": 0.3284,
      "step": 2409
    },
    {
      "epoch": 15.6,
      "learning_rate": 2.3817849617532638e-05,
      "loss": 0.3301,
      "step": 2410
    },
    {
      "epoch": 15.61,
      "learning_rate": 2.3749760434319933e-05,
      "loss": 0.2836,
      "step": 2411
    },
    {
      "epoch": 15.61,
      "learning_rate": 2.3681755598268828e-05,
      "loss": 0.3464,
      "step": 2412
    },
    {
      "epoch": 15.62,
      "learning_rate": 2.3613835184605525e-05,
      "loss": 0.2654,
      "step": 2413
    },
    {
      "epoch": 15.62,
      "learning_rate": 2.3545999268462825e-05,
      "loss": 0.3601,
      "step": 2414
    },
    {
      "epoch": 15.63,
      "learning_rate": 2.347824792488007e-05,
      "loss": 0.3285,
      "step": 2415
    },
    {
      "epoch": 15.64,
      "learning_rate": 2.341058122880304e-05,
      "loss": 0.2803,
      "step": 2416
    },
    {
      "epoch": 15.64,
      "learning_rate": 2.3342999255083885e-05,
      "loss": 0.3647,
      "step": 2417
    },
    {
      "epoch": 15.65,
      "learning_rate": 2.3275502078480994e-05,
      "loss": 0.3088,
      "step": 2418
    },
    {
      "epoch": 15.66,
      "learning_rate": 2.3208089773659e-05,
      "loss": 0.3445,
      "step": 2419
    },
    {
      "epoch": 15.66,
      "learning_rate": 2.3140762415188645e-05,
      "loss": 0.3228,
      "step": 2420
    },
    {
      "epoch": 15.67,
      "learning_rate": 2.307352007754674e-05,
      "loss": 0.3069,
      "step": 2421
    },
    {
      "epoch": 15.68,
      "learning_rate": 2.3006362835115925e-05,
      "loss": 0.3208,
      "step": 2422
    },
    {
      "epoch": 15.68,
      "learning_rate": 2.2939290762184873e-05,
      "loss": 0.2683,
      "step": 2423
    },
    {
      "epoch": 15.69,
      "learning_rate": 2.287230393294788e-05,
      "loss": 0.29,
      "step": 2424
    },
    {
      "epoch": 15.7,
      "learning_rate": 2.2805402421505074e-05,
      "loss": 0.3005,
      "step": 2425
    },
    {
      "epoch": 15.7,
      "learning_rate": 2.2738586301862154e-05,
      "loss": 0.2847,
      "step": 2426
    },
    {
      "epoch": 15.71,
      "learning_rate": 2.267185564793036e-05,
      "loss": 0.3188,
      "step": 2427
    },
    {
      "epoch": 15.72,
      "learning_rate": 2.2605210533526388e-05,
      "loss": 0.3015,
      "step": 2428
    },
    {
      "epoch": 15.72,
      "learning_rate": 2.253865103237236e-05,
      "loss": 0.2955,
      "step": 2429
    },
    {
      "epoch": 15.73,
      "learning_rate": 2.247217721809558e-05,
      "loss": 0.3163,
      "step": 2430
    },
    {
      "epoch": 15.73,
      "learning_rate": 2.2405789164228685e-05,
      "loss": 0.2915,
      "step": 2431
    },
    {
      "epoch": 15.74,
      "learning_rate": 2.2339486944209375e-05,
      "loss": 0.3928,
      "step": 2432
    },
    {
      "epoch": 15.75,
      "learning_rate": 2.227327063138044e-05,
      "loss": 0.3243,
      "step": 2433
    },
    {
      "epoch": 15.75,
      "learning_rate": 2.2207140298989616e-05,
      "loss": 0.2898,
      "step": 2434
    },
    {
      "epoch": 15.76,
      "learning_rate": 2.2141096020189565e-05,
      "loss": 0.278,
      "step": 2435
    },
    {
      "epoch": 15.77,
      "learning_rate": 2.2075137868037676e-05,
      "loss": 0.3188,
      "step": 2436
    },
    {
      "epoch": 15.77,
      "learning_rate": 2.200926591549618e-05,
      "loss": 0.3076,
      "step": 2437
    },
    {
      "epoch": 15.78,
      "learning_rate": 2.1943480235431825e-05,
      "loss": 0.2942,
      "step": 2438
    },
    {
      "epoch": 15.79,
      "learning_rate": 2.187778090061603e-05,
      "loss": 0.3079,
      "step": 2439
    },
    {
      "epoch": 15.79,
      "learning_rate": 2.1812167983724696e-05,
      "loss": 0.3452,
      "step": 2440
    },
    {
      "epoch": 15.8,
      "learning_rate": 2.174664155733803e-05,
      "loss": 0.2697,
      "step": 2441
    },
    {
      "epoch": 15.81,
      "learning_rate": 2.1681201693940668e-05,
      "loss": 0.3029,
      "step": 2442
    },
    {
      "epoch": 15.81,
      "learning_rate": 2.1615848465921452e-05,
      "loss": 0.2869,
      "step": 2443
    },
    {
      "epoch": 15.82,
      "learning_rate": 2.1550581945573377e-05,
      "loss": 0.2998,
      "step": 2444
    },
    {
      "epoch": 15.83,
      "learning_rate": 2.1485402205093573e-05,
      "loss": 0.306,
      "step": 2445
    },
    {
      "epoch": 15.83,
      "learning_rate": 2.142030931658313e-05,
      "loss": 0.2673,
      "step": 2446
    },
    {
      "epoch": 15.84,
      "learning_rate": 2.135530335204704e-05,
      "loss": 0.3777,
      "step": 2447
    },
    {
      "epoch": 15.84,
      "learning_rate": 2.129038438339419e-05,
      "loss": 0.3129,
      "step": 2448
    },
    {
      "epoch": 15.85,
      "learning_rate": 2.1225552482437228e-05,
      "loss": 0.322,
      "step": 2449
    },
    {
      "epoch": 15.86,
      "learning_rate": 2.116080772089246e-05,
      "loss": 0.3097,
      "step": 2450
    },
    {
      "epoch": 15.86,
      "learning_rate": 2.1096150170379837e-05,
      "loss": 0.3551,
      "step": 2451
    },
    {
      "epoch": 15.87,
      "learning_rate": 2.1031579902422827e-05,
      "loss": 0.2832,
      "step": 2452
    },
    {
      "epoch": 15.88,
      "learning_rate": 2.09670969884483e-05,
      "loss": 0.3062,
      "step": 2453
    },
    {
      "epoch": 15.88,
      "learning_rate": 2.090270149978657e-05,
      "loss": 0.3271,
      "step": 2454
    },
    {
      "epoch": 15.89,
      "learning_rate": 2.083839350767116e-05,
      "loss": 0.3293,
      "step": 2455
    },
    {
      "epoch": 15.9,
      "learning_rate": 2.0774173083238878e-05,
      "loss": 0.3137,
      "step": 2456
    },
    {
      "epoch": 15.9,
      "learning_rate": 2.0710040297529643e-05,
      "loss": 0.2883,
      "step": 2457
    },
    {
      "epoch": 15.91,
      "learning_rate": 2.0645995221486446e-05,
      "loss": 0.2583,
      "step": 2458
    },
    {
      "epoch": 15.92,
      "learning_rate": 2.058203792595519e-05,
      "loss": 0.3188,
      "step": 2459
    },
    {
      "epoch": 15.92,
      "learning_rate": 2.051816848168473e-05,
      "loss": 0.311,
      "step": 2460
    },
    {
      "epoch": 15.93,
      "learning_rate": 2.045438695932673e-05,
      "loss": 0.3682,
      "step": 2461
    },
    {
      "epoch": 15.94,
      "learning_rate": 2.0390693429435627e-05,
      "loss": 0.2764,
      "step": 2462
    },
    {
      "epoch": 15.94,
      "learning_rate": 2.032708796246845e-05,
      "loss": 0.2993,
      "step": 2463
    },
    {
      "epoch": 15.95,
      "learning_rate": 2.026357062878491e-05,
      "loss": 0.2439,
      "step": 2464
    },
    {
      "epoch": 15.95,
      "learning_rate": 2.0200141498647095e-05,
      "loss": 0.3497,
      "step": 2465
    },
    {
      "epoch": 15.96,
      "learning_rate": 2.0136800642219656e-05,
      "loss": 0.3127,
      "step": 2466
    },
    {
      "epoch": 15.97,
      "learning_rate": 2.0073548129569507e-05,
      "loss": 0.2965,
      "step": 2467
    },
    {
      "epoch": 15.97,
      "learning_rate": 2.001038403066592e-05,
      "loss": 0.243,
      "step": 2468
    },
    {
      "epoch": 15.98,
      "learning_rate": 1.994730841538024e-05,
      "loss": 0.2766,
      "step": 2469
    },
    {
      "epoch": 15.99,
      "learning_rate": 1.988432135348608e-05,
      "loss": 0.2306,
      "step": 2470
    },
    {
      "epoch": 15.99,
      "learning_rate": 1.982142291465896e-05,
      "loss": 0.3475,
      "step": 2471
    },
    {
      "epoch": 16.0,
      "learning_rate": 1.975861316847645e-05,
      "loss": 0.2723,
      "step": 2472
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.9695892184417996e-05,
      "loss": 0.3147,
      "step": 2473
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.9633260031864852e-05,
      "loss": 0.281,
      "step": 2474
    },
    {
      "epoch": 16.02,
      "learning_rate": 1.9570716780100007e-05,
      "loss": 0.3092,
      "step": 2475
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.9508262498308118e-05,
      "loss": 0.3081,
      "step": 2476
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.944589725557536e-05,
      "loss": 0.3007,
      "step": 2477
    },
    {
      "epoch": 16.04,
      "learning_rate": 1.9383621120889517e-05,
      "loss": 0.2765,
      "step": 2478
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.932143416313972e-05,
      "loss": 0.2825,
      "step": 2479
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.9259336451116506e-05,
      "loss": 0.2786,
      "step": 2480
    },
    {
      "epoch": 16.06,
      "learning_rate": 1.9197328053511686e-05,
      "loss": 0.2681,
      "step": 2481
    },
    {
      "epoch": 16.06,
      "learning_rate": 1.9135409038918217e-05,
      "loss": 0.2791,
      "step": 2482
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.9073579475830227e-05,
      "loss": 0.2694,
      "step": 2483
    },
    {
      "epoch": 16.08,
      "learning_rate": 1.9011839432642908e-05,
      "loss": 0.2942,
      "step": 2484
    },
    {
      "epoch": 16.08,
      "learning_rate": 1.8950188977652405e-05,
      "loss": 0.2533,
      "step": 2485
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.8888628179055735e-05,
      "loss": 0.3069,
      "step": 2486
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.8827157104950798e-05,
      "loss": 0.2825,
      "step": 2487
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.876577582333616e-05,
      "loss": 0.2982,
      "step": 2488
    },
    {
      "epoch": 16.11,
      "learning_rate": 1.8704484402111134e-05,
      "loss": 0.2607,
      "step": 2489
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.8643282909075623e-05,
      "loss": 0.2963,
      "step": 2490
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.8582171411929995e-05,
      "loss": 0.2454,
      "step": 2491
    },
    {
      "epoch": 16.13,
      "learning_rate": 1.8521149978275145e-05,
      "loss": 0.2974,
      "step": 2492
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.8460218675612296e-05,
      "loss": 0.2817,
      "step": 2493
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.8399377571342936e-05,
      "loss": 0.3074,
      "step": 2494
    },
    {
      "epoch": 16.15,
      "learning_rate": 1.8338626732768838e-05,
      "loss": 0.266,
      "step": 2495
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.8277966227091892e-05,
      "loss": 0.2991,
      "step": 2496
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.8217396121414087e-05,
      "loss": 0.2833,
      "step": 2497
    },
    {
      "epoch": 16.17,
      "learning_rate": 1.8156916482737374e-05,
      "loss": 0.2769,
      "step": 2498
    },
    {
      "epoch": 16.17,
      "learning_rate": 1.80965273779637e-05,
      "loss": 0.2477,
      "step": 2499
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.8036228873894746e-05,
      "loss": 0.2546,
      "step": 2500
    },
    {
      "epoch": 16.19,
      "learning_rate": 1.7976021037232105e-05,
      "loss": 0.2908,
      "step": 2501
    },
    {
      "epoch": 16.19,
      "learning_rate": 1.791590393457696e-05,
      "loss": 0.267,
      "step": 2502
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.7855877632430194e-05,
      "loss": 0.2583,
      "step": 2503
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.779594219719225e-05,
      "loss": 0.2821,
      "step": 2504
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.7736097695163046e-05,
      "loss": 0.2167,
      "step": 2505
    },
    {
      "epoch": 16.22,
      "learning_rate": 1.767634419254185e-05,
      "loss": 0.2347,
      "step": 2506
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.7616681755427345e-05,
      "loss": 0.3098,
      "step": 2507
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.755711044981745e-05,
      "loss": 0.2317,
      "step": 2508
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.7497630341609274e-05,
      "loss": 0.255,
      "step": 2509
    },
    {
      "epoch": 16.25,
      "learning_rate": 1.7438241496599063e-05,
      "loss": 0.2997,
      "step": 2510
    },
    {
      "epoch": 16.25,
      "learning_rate": 1.73789439804821e-05,
      "loss": 0.2764,
      "step": 2511
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.731973785885258e-05,
      "loss": 0.2383,
      "step": 2512
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.7260623197203695e-05,
      "loss": 0.2689,
      "step": 2513
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.7201600060927427e-05,
      "loss": 0.2831,
      "step": 2514
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.7142668515314496e-05,
      "loss": 0.2794,
      "step": 2515
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.708382862555431e-05,
      "loss": 0.2777,
      "step": 2516
    },
    {
      "epoch": 16.29,
      "learning_rate": 1.70250804567349e-05,
      "loss": 0.4182,
      "step": 2517
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.696642407384287e-05,
      "loss": 0.2986,
      "step": 2518
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.6907859541763183e-05,
      "loss": 0.3098,
      "step": 2519
    },
    {
      "epoch": 16.31,
      "learning_rate": 1.684938692527931e-05,
      "loss": 0.2965,
      "step": 2520
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.6791006289073008e-05,
      "loss": 0.3104,
      "step": 2521
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.6732717697724298e-05,
      "loss": 0.2952,
      "step": 2522
    },
    {
      "epoch": 16.33,
      "learning_rate": 1.6674521215711324e-05,
      "loss": 0.2335,
      "step": 2523
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.6616416907410404e-05,
      "loss": 0.2676,
      "step": 2524
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.6558404837095877e-05,
      "loss": 0.3071,
      "step": 2525
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.6500485068940054e-05,
      "loss": 0.3124,
      "step": 2526
    },
    {
      "epoch": 16.36,
      "learning_rate": 1.6442657667013116e-05,
      "loss": 0.2388,
      "step": 2527
    },
    {
      "epoch": 16.36,
      "learning_rate": 1.6384922695283122e-05,
      "loss": 0.324,
      "step": 2528
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.6327280217615792e-05,
      "loss": 0.3391,
      "step": 2529
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.6269730297774642e-05,
      "loss": 0.2744,
      "step": 2530
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.6212272999420718e-05,
      "loss": 0.3687,
      "step": 2531
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.6154908386112687e-05,
      "loss": 0.2599,
      "step": 2532
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.609763652130658e-05,
      "loss": 0.2284,
      "step": 2533
    },
    {
      "epoch": 16.4,
      "learning_rate": 1.604045746835594e-05,
      "loss": 0.3015,
      "step": 2534
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.5983371290511563e-05,
      "loss": 0.3027,
      "step": 2535
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.5926378050921554e-05,
      "loss": 0.2491,
      "step": 2536
    },
    {
      "epoch": 16.42,
      "learning_rate": 1.5869477812631183e-05,
      "loss": 0.2944,
      "step": 2537
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.5812670638582882e-05,
      "loss": 0.3022,
      "step": 2538
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.5755956591616104e-05,
      "loss": 0.285,
      "step": 2539
    },
    {
      "epoch": 16.44,
      "learning_rate": 1.5699335734467303e-05,
      "loss": 0.2559,
      "step": 2540
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.5642808129769805e-05,
      "loss": 0.288,
      "step": 2541
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.5586373840053836e-05,
      "loss": 0.3073,
      "step": 2542
    },
    {
      "epoch": 16.46,
      "learning_rate": 1.5530032927746364e-05,
      "loss": 0.2452,
      "step": 2543
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.5473785455171087e-05,
      "loss": 0.3118,
      "step": 2544
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.5417631484548322e-05,
      "loss": 0.2943,
      "step": 2545
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.5361571077994986e-05,
      "loss": 0.3025,
      "step": 2546
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.5305604297524445e-05,
      "loss": 0.2803,
      "step": 2547
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.5249731205046546e-05,
      "loss": 0.3184,
      "step": 2548
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.519395186236745e-05,
      "loss": 0.281,
      "step": 2549
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.5138266331189654e-05,
      "loss": 0.2834,
      "step": 2550
    },
    {
      "epoch": 16.51,
      "learning_rate": 1.5082674673111897e-05,
      "loss": 0.2847,
      "step": 2551
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.5027176949629007e-05,
      "loss": 0.2637,
      "step": 2552
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.4971773222131957e-05,
      "loss": 0.2758,
      "step": 2553
    },
    {
      "epoch": 16.53,
      "learning_rate": 1.4916463551907745e-05,
      "loss": 0.2915,
      "step": 2554
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.4861248000139293e-05,
      "loss": 0.2898,
      "step": 2555
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.4806126627905437e-05,
      "loss": 0.3025,
      "step": 2556
    },
    {
      "epoch": 16.55,
      "learning_rate": 1.4751099496180842e-05,
      "loss": 0.313,
      "step": 2557
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.4696166665835853e-05,
      "loss": 0.2869,
      "step": 2558
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.4641328197636595e-05,
      "loss": 0.2894,
      "step": 2559
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.4586584152244743e-05,
      "loss": 0.233,
      "step": 2560
    },
    {
      "epoch": 16.58,
      "learning_rate": 1.4531934590217566e-05,
      "loss": 0.2727,
      "step": 2561
    },
    {
      "epoch": 16.58,
      "learning_rate": 1.4477379572007787e-05,
      "loss": 0.235,
      "step": 2562
    },
    {
      "epoch": 16.59,
      "learning_rate": 1.4422919157963577e-05,
      "loss": 0.2814,
      "step": 2563
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.4368553408328412e-05,
      "loss": 0.2692,
      "step": 2564
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.4314282383241096e-05,
      "loss": 0.2596,
      "step": 2565
    },
    {
      "epoch": 16.61,
      "learning_rate": 1.4260106142735596e-05,
      "loss": 0.3169,
      "step": 2566
    },
    {
      "epoch": 16.61,
      "learning_rate": 1.4206024746741075e-05,
      "loss": 0.2358,
      "step": 2567
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.415203825508179e-05,
      "loss": 0.2511,
      "step": 2568
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.4098146727477002e-05,
      "loss": 0.2468,
      "step": 2569
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.404435022354087e-05,
      "loss": 0.2681,
      "step": 2570
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.3990648802782502e-05,
      "loss": 0.322,
      "step": 2571
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.3937042524605826e-05,
      "loss": 0.3223,
      "step": 2572
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.3883531448309484e-05,
      "loss": 0.2389,
      "step": 2573
    },
    {
      "epoch": 16.66,
      "learning_rate": 1.383011563308686e-05,
      "loss": 0.303,
      "step": 2574
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.3776795138025922e-05,
      "loss": 0.2394,
      "step": 2575
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.3723570022109178e-05,
      "loss": 0.2721,
      "step": 2576
    },
    {
      "epoch": 16.68,
      "learning_rate": 1.3670440344213674e-05,
      "loss": 0.2539,
      "step": 2577
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.3617406163110857e-05,
      "loss": 0.2574,
      "step": 2578
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.3564467537466584e-05,
      "loss": 0.2584,
      "step": 2579
    },
    {
      "epoch": 16.7,
      "learning_rate": 1.3511624525840915e-05,
      "loss": 0.321,
      "step": 2580
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.3458877186688235e-05,
      "loss": 0.2673,
      "step": 2581
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.3406225578357034e-05,
      "loss": 0.249,
      "step": 2582
    },
    {
      "epoch": 16.72,
      "learning_rate": 1.3353669759089949e-05,
      "loss": 0.3169,
      "step": 2583
    },
    {
      "epoch": 16.72,
      "learning_rate": 1.3301209787023628e-05,
      "loss": 0.3008,
      "step": 2584
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.3248845720188707e-05,
      "loss": 0.2474,
      "step": 2585
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.3196577616509743e-05,
      "loss": 0.2759,
      "step": 2586
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.3144405533805138e-05,
      "loss": 0.3153,
      "step": 2587
    },
    {
      "epoch": 16.75,
      "learning_rate": 1.3092329529787017e-05,
      "loss": 0.307,
      "step": 2588
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.3040349662061313e-05,
      "loss": 0.295,
      "step": 2589
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.2988465988127562e-05,
      "loss": 0.2329,
      "step": 2590
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.2936678565378913e-05,
      "loss": 0.3254,
      "step": 2591
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.2884987451102037e-05,
      "loss": 0.3054,
      "step": 2592
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.2833392702477044e-05,
      "loss": 0.2935,
      "step": 2593
    },
    {
      "epoch": 16.79,
      "learning_rate": 1.2781894376577474e-05,
      "loss": 0.2697,
      "step": 2594
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.2730492530370208e-05,
      "loss": 0.3029,
      "step": 2595
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.2679187220715405e-05,
      "loss": 0.3306,
      "step": 2596
    },
    {
      "epoch": 16.81,
      "learning_rate": 1.2627978504366378e-05,
      "loss": 0.2997,
      "step": 2597
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.2576866437969704e-05,
      "loss": 0.2722,
      "step": 2598
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.252585107806491e-05,
      "loss": 0.3468,
      "step": 2599
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.2474932481084645e-05,
      "loss": 0.2506,
      "step": 2600
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.2424110703354497e-05,
      "loss": 0.2659,
      "step": 2601
    },
    {
      "epoch": 16.84,
      "learning_rate": 1.2373385801092941e-05,
      "loss": 0.2476,
      "step": 2602
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.2322757830411302e-05,
      "loss": 0.3105,
      "step": 2603
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.2272226847313683e-05,
      "loss": 0.2996,
      "step": 2604
    },
    {
      "epoch": 16.86,
      "learning_rate": 1.2221792907696871e-05,
      "loss": 0.3505,
      "step": 2605
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.2171456067350329e-05,
      "loss": 0.272,
      "step": 2606
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.2121216381956113e-05,
      "loss": 0.2633,
      "step": 2607
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.2071073907088803e-05,
      "loss": 0.2485,
      "step": 2608
    },
    {
      "epoch": 16.89,
      "learning_rate": 1.202102869821543e-05,
      "loss": 0.3237,
      "step": 2609
    },
    {
      "epoch": 16.89,
      "learning_rate": 1.197108081069549e-05,
      "loss": 0.2582,
      "step": 2610
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.1921230299780705e-05,
      "loss": 0.2581,
      "step": 2611
    },
    {
      "epoch": 16.91,
      "learning_rate": 1.1871477220615223e-05,
      "loss": 0.3154,
      "step": 2612
    },
    {
      "epoch": 16.91,
      "learning_rate": 1.1821821628235286e-05,
      "loss": 0.2844,
      "step": 2613
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.1772263577569376e-05,
      "loss": 0.2993,
      "step": 2614
    },
    {
      "epoch": 16.93,
      "learning_rate": 1.1722803123438053e-05,
      "loss": 0.2483,
      "step": 2615
    },
    {
      "epoch": 16.93,
      "learning_rate": 1.167344032055394e-05,
      "loss": 0.2942,
      "step": 2616
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.1624175223521583e-05,
      "loss": 0.301,
      "step": 2617
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.1575007886837486e-05,
      "loss": 0.2513,
      "step": 2618
    },
    {
      "epoch": 16.95,
      "learning_rate": 1.1525938364890021e-05,
      "loss": 0.2728,
      "step": 2619
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.1476966711959347e-05,
      "loss": 0.2634,
      "step": 2620
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.1428092982217342e-05,
      "loss": 0.2822,
      "step": 2621
    },
    {
      "epoch": 16.97,
      "learning_rate": 1.1379317229727605e-05,
      "loss": 0.2819,
      "step": 2622
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.1330639508445296e-05,
      "loss": 0.3241,
      "step": 2623
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.1282059872217176e-05,
      "loss": 0.2419,
      "step": 2624
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.1233578374781495e-05,
      "loss": 0.3307,
      "step": 2625
    },
    {
      "epoch": 17.0,
      "learning_rate": 1.1185195069767951e-05,
      "loss": 0.3503,
      "step": 2626
    },
    {
      "epoch": 17.0,
      "learning_rate": 1.1136910010697632e-05,
      "loss": 0.2434,
      "step": 2627
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.108872325098288e-05,
      "loss": 0.2535,
      "step": 2628
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.10406348439274e-05,
      "loss": 0.2651,
      "step": 2629
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.0992644842726007e-05,
      "loss": 0.2705,
      "step": 2630
    },
    {
      "epoch": 17.03,
      "learning_rate": 1.0944753300464727e-05,
      "loss": 0.2341,
      "step": 2631
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.0896960270120638e-05,
      "loss": 0.265,
      "step": 2632
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.084926580456188e-05,
      "loss": 0.3032,
      "step": 2633
    },
    {
      "epoch": 17.05,
      "learning_rate": 1.08016699565475e-05,
      "loss": 0.2698,
      "step": 2634
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.0754172778727512e-05,
      "loss": 0.2263,
      "step": 2635
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.0706774323642766e-05,
      "loss": 0.2644,
      "step": 2636
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.0659474643724886e-05,
      "loss": 0.2487,
      "step": 2637
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.0612273791296268e-05,
      "loss": 0.261,
      "step": 2638
    },
    {
      "epoch": 17.08,
      "learning_rate": 1.0565171818569975e-05,
      "loss": 0.256,
      "step": 2639
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.0518168777649639e-05,
      "loss": 0.3027,
      "step": 2640
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.047126472052954e-05,
      "loss": 0.2719,
      "step": 2641
    },
    {
      "epoch": 17.1,
      "learning_rate": 1.042445969909439e-05,
      "loss": 0.2469,
      "step": 2642
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.0377753765119424e-05,
      "loss": 0.2585,
      "step": 2643
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.0331146970270177e-05,
      "loss": 0.2439,
      "step": 2644
    },
    {
      "epoch": 17.12,
      "learning_rate": 1.02846393661026e-05,
      "loss": 0.2772,
      "step": 2645
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.0238231004062849e-05,
      "loss": 0.3232,
      "step": 2646
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.0191921935487347e-05,
      "loss": 0.2891,
      "step": 2647
    },
    {
      "epoch": 17.14,
      "learning_rate": 1.0145712211602665e-05,
      "loss": 0.2125,
      "step": 2648
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.00996018835255e-05,
      "loss": 0.2971,
      "step": 2649
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.0053591002262564e-05,
      "loss": 0.2764,
      "step": 2650
    },
    {
      "epoch": 17.16,
      "learning_rate": 1.0007679618710608e-05,
      "loss": 0.2931,
      "step": 2651
    },
    {
      "epoch": 17.17,
      "learning_rate": 9.961867783656253e-06,
      "loss": 0.2455,
      "step": 2652
    },
    {
      "epoch": 17.17,
      "learning_rate": 9.91615554777604e-06,
      "loss": 0.2219,
      "step": 2653
    },
    {
      "epoch": 17.18,
      "learning_rate": 9.870542961636354e-06,
      "loss": 0.3218,
      "step": 2654
    },
    {
      "epoch": 17.18,
      "learning_rate": 9.82503007569332e-06,
      "loss": 0.2283,
      "step": 2655
    },
    {
      "epoch": 17.19,
      "learning_rate": 9.77961694029278e-06,
      "loss": 0.2563,
      "step": 2656
    },
    {
      "epoch": 17.2,
      "learning_rate": 9.73430360567026e-06,
      "loss": 0.2695,
      "step": 2657
    },
    {
      "epoch": 17.2,
      "learning_rate": 9.689090121950818e-06,
      "loss": 0.2382,
      "step": 2658
    },
    {
      "epoch": 17.21,
      "learning_rate": 9.643976539149136e-06,
      "loss": 0.248,
      "step": 2659
    },
    {
      "epoch": 17.22,
      "learning_rate": 9.59896290716935e-06,
      "loss": 0.2676,
      "step": 2660
    },
    {
      "epoch": 17.22,
      "learning_rate": 9.554049275805033e-06,
      "loss": 0.3192,
      "step": 2661
    },
    {
      "epoch": 17.23,
      "learning_rate": 9.509235694739126e-06,
      "loss": 0.2969,
      "step": 2662
    },
    {
      "epoch": 17.24,
      "learning_rate": 9.464522213543947e-06,
      "loss": 0.3484,
      "step": 2663
    },
    {
      "epoch": 17.24,
      "learning_rate": 9.419908881681006e-06,
      "loss": 0.2637,
      "step": 2664
    },
    {
      "epoch": 17.25,
      "learning_rate": 9.375395748501092e-06,
      "loss": 0.2407,
      "step": 2665
    },
    {
      "epoch": 17.26,
      "learning_rate": 9.330982863244131e-06,
      "loss": 0.3262,
      "step": 2666
    },
    {
      "epoch": 17.26,
      "learning_rate": 9.286670275039167e-06,
      "loss": 0.2848,
      "step": 2667
    },
    {
      "epoch": 17.27,
      "learning_rate": 9.242458032904311e-06,
      "loss": 0.2463,
      "step": 2668
    },
    {
      "epoch": 17.28,
      "learning_rate": 9.198346185746598e-06,
      "loss": 0.2817,
      "step": 2669
    },
    {
      "epoch": 17.28,
      "learning_rate": 9.154334782362084e-06,
      "loss": 0.3107,
      "step": 2670
    },
    {
      "epoch": 17.29,
      "learning_rate": 9.110423871435702e-06,
      "loss": 0.256,
      "step": 2671
    },
    {
      "epoch": 17.29,
      "learning_rate": 9.066613501541187e-06,
      "loss": 0.2717,
      "step": 2672
    },
    {
      "epoch": 17.3,
      "learning_rate": 9.022903721141108e-06,
      "loss": 0.2544,
      "step": 2673
    },
    {
      "epoch": 17.31,
      "learning_rate": 8.979294578586738e-06,
      "loss": 0.2463,
      "step": 2674
    },
    {
      "epoch": 17.31,
      "learning_rate": 8.935786122117984e-06,
      "loss": 0.2543,
      "step": 2675
    },
    {
      "epoch": 17.32,
      "learning_rate": 8.892378399863454e-06,
      "loss": 0.2577,
      "step": 2676
    },
    {
      "epoch": 17.33,
      "learning_rate": 8.849071459840242e-06,
      "loss": 0.2672,
      "step": 2677
    },
    {
      "epoch": 17.33,
      "learning_rate": 8.805865349954035e-06,
      "loss": 0.2947,
      "step": 2678
    },
    {
      "epoch": 17.34,
      "learning_rate": 8.762760117998926e-06,
      "loss": 0.3224,
      "step": 2679
    },
    {
      "epoch": 17.35,
      "learning_rate": 8.719755811657471e-06,
      "loss": 0.2728,
      "step": 2680
    },
    {
      "epoch": 17.35,
      "learning_rate": 8.676852478500518e-06,
      "loss": 0.3204,
      "step": 2681
    },
    {
      "epoch": 17.36,
      "learning_rate": 8.634050165987272e-06,
      "loss": 0.2432,
      "step": 2682
    },
    {
      "epoch": 17.37,
      "learning_rate": 8.59134892146517e-06,
      "loss": 0.2556,
      "step": 2683
    },
    {
      "epoch": 17.37,
      "learning_rate": 8.548748792169847e-06,
      "loss": 0.3088,
      "step": 2684
    },
    {
      "epoch": 17.38,
      "learning_rate": 8.506249825225087e-06,
      "loss": 0.3029,
      "step": 2685
    },
    {
      "epoch": 17.39,
      "learning_rate": 8.463852067642797e-06,
      "loss": 0.2473,
      "step": 2686
    },
    {
      "epoch": 17.39,
      "learning_rate": 8.42155556632287e-06,
      "loss": 0.2666,
      "step": 2687
    },
    {
      "epoch": 17.4,
      "learning_rate": 8.379360368053224e-06,
      "loss": 0.3103,
      "step": 2688
    },
    {
      "epoch": 17.4,
      "learning_rate": 8.337266519509724e-06,
      "loss": 0.2729,
      "step": 2689
    },
    {
      "epoch": 17.41,
      "learning_rate": 8.295274067256143e-06,
      "loss": 0.2517,
      "step": 2690
    },
    {
      "epoch": 17.42,
      "learning_rate": 8.253383057744003e-06,
      "loss": 0.259,
      "step": 2691
    },
    {
      "epoch": 17.42,
      "learning_rate": 8.211593537312711e-06,
      "loss": 0.2993,
      "step": 2692
    },
    {
      "epoch": 17.43,
      "learning_rate": 8.169905552189372e-06,
      "loss": 0.238,
      "step": 2693
    },
    {
      "epoch": 17.44,
      "learning_rate": 8.128319148488728e-06,
      "loss": 0.2568,
      "step": 2694
    },
    {
      "epoch": 17.44,
      "learning_rate": 8.086834372213226e-06,
      "loss": 0.3303,
      "step": 2695
    },
    {
      "epoch": 17.45,
      "learning_rate": 8.045451269252857e-06,
      "loss": 0.2808,
      "step": 2696
    },
    {
      "epoch": 17.46,
      "learning_rate": 8.004169885385138e-06,
      "loss": 0.3303,
      "step": 2697
    },
    {
      "epoch": 17.46,
      "learning_rate": 7.962990266275094e-06,
      "loss": 0.2004,
      "step": 2698
    },
    {
      "epoch": 17.47,
      "learning_rate": 7.921912457475123e-06,
      "loss": 0.2368,
      "step": 2699
    },
    {
      "epoch": 17.48,
      "learning_rate": 7.880936504425062e-06,
      "loss": 0.2656,
      "step": 2700
    },
    {
      "epoch": 17.48,
      "learning_rate": 7.840062452452047e-06,
      "loss": 0.2932,
      "step": 2701
    },
    {
      "epoch": 17.49,
      "learning_rate": 7.799290346770506e-06,
      "loss": 0.2611,
      "step": 2702
    },
    {
      "epoch": 17.5,
      "learning_rate": 7.758620232482084e-06,
      "loss": 0.2787,
      "step": 2703
    },
    {
      "epoch": 17.5,
      "learning_rate": 7.718052154575583e-06,
      "loss": 0.345,
      "step": 2704
    },
    {
      "epoch": 17.51,
      "learning_rate": 7.677586157926975e-06,
      "loss": 0.2538,
      "step": 2705
    },
    {
      "epoch": 17.51,
      "learning_rate": 7.637222287299283e-06,
      "loss": 0.2504,
      "step": 2706
    },
    {
      "epoch": 17.52,
      "learning_rate": 7.596960587342605e-06,
      "loss": 0.2603,
      "step": 2707
    },
    {
      "epoch": 17.53,
      "learning_rate": 7.5568011025939355e-06,
      "loss": 0.2295,
      "step": 2708
    },
    {
      "epoch": 17.53,
      "learning_rate": 7.516743877477284e-06,
      "loss": 0.3042,
      "step": 2709
    },
    {
      "epoch": 17.54,
      "learning_rate": 7.47678895630346e-06,
      "loss": 0.3301,
      "step": 2710
    },
    {
      "epoch": 17.55,
      "learning_rate": 7.436936383270188e-06,
      "loss": 0.2892,
      "step": 2711
    },
    {
      "epoch": 17.55,
      "learning_rate": 7.397186202461925e-06,
      "loss": 0.2567,
      "step": 2712
    },
    {
      "epoch": 17.56,
      "learning_rate": 7.357538457849878e-06,
      "loss": 0.2657,
      "step": 2713
    },
    {
      "epoch": 17.57,
      "learning_rate": 7.317993193291928e-06,
      "loss": 0.2565,
      "step": 2714
    },
    {
      "epoch": 17.57,
      "learning_rate": 7.27855045253264e-06,
      "loss": 0.2471,
      "step": 2715
    },
    {
      "epoch": 17.58,
      "learning_rate": 7.23921027920309e-06,
      "loss": 0.2573,
      "step": 2716
    },
    {
      "epoch": 17.59,
      "learning_rate": 7.199972716820957e-06,
      "loss": 0.2621,
      "step": 2717
    },
    {
      "epoch": 17.59,
      "learning_rate": 7.1608378087903885e-06,
      "loss": 0.2828,
      "step": 2718
    },
    {
      "epoch": 17.6,
      "learning_rate": 7.121805598401987e-06,
      "loss": 0.276,
      "step": 2719
    },
    {
      "epoch": 17.61,
      "learning_rate": 7.08287612883275e-06,
      "loss": 0.281,
      "step": 2720
    },
    {
      "epoch": 17.61,
      "learning_rate": 7.044049443146017e-06,
      "loss": 0.2786,
      "step": 2721
    },
    {
      "epoch": 17.62,
      "learning_rate": 7.005325584291423e-06,
      "loss": 0.2828,
      "step": 2722
    },
    {
      "epoch": 17.62,
      "learning_rate": 6.966704595104889e-06,
      "loss": 0.2458,
      "step": 2723
    },
    {
      "epoch": 17.63,
      "learning_rate": 6.928186518308499e-06,
      "loss": 0.2737,
      "step": 2724
    },
    {
      "epoch": 17.64,
      "learning_rate": 6.889771396510514e-06,
      "loss": 0.2747,
      "step": 2725
    },
    {
      "epoch": 17.64,
      "learning_rate": 6.851459272205329e-06,
      "loss": 0.262,
      "step": 2726
    },
    {
      "epoch": 17.65,
      "learning_rate": 6.8132501877734015e-06,
      "loss": 0.2842,
      "step": 2727
    },
    {
      "epoch": 17.66,
      "learning_rate": 6.775144185481164e-06,
      "loss": 0.2734,
      "step": 2728
    },
    {
      "epoch": 17.66,
      "learning_rate": 6.737141307481054e-06,
      "loss": 0.2921,
      "step": 2729
    },
    {
      "epoch": 17.67,
      "learning_rate": 6.699241595811434e-06,
      "loss": 0.2961,
      "step": 2730
    },
    {
      "epoch": 17.68,
      "learning_rate": 6.661445092396545e-06,
      "loss": 0.217,
      "step": 2731
    },
    {
      "epoch": 17.68,
      "learning_rate": 6.623751839046455e-06,
      "loss": 0.2278,
      "step": 2732
    },
    {
      "epoch": 17.69,
      "learning_rate": 6.586161877457031e-06,
      "loss": 0.292,
      "step": 2733
    },
    {
      "epoch": 17.7,
      "learning_rate": 6.548675249209835e-06,
      "loss": 0.2583,
      "step": 2734
    },
    {
      "epoch": 17.7,
      "learning_rate": 6.511291995772184e-06,
      "loss": 0.282,
      "step": 2735
    },
    {
      "epoch": 17.71,
      "learning_rate": 6.4740121584970004e-06,
      "loss": 0.2849,
      "step": 2736
    },
    {
      "epoch": 17.72,
      "learning_rate": 6.436835778622829e-06,
      "loss": 0.3137,
      "step": 2737
    },
    {
      "epoch": 17.72,
      "learning_rate": 6.399762897273787e-06,
      "loss": 0.2206,
      "step": 2738
    },
    {
      "epoch": 17.73,
      "learning_rate": 6.362793555459434e-06,
      "loss": 0.2979,
      "step": 2739
    },
    {
      "epoch": 17.73,
      "learning_rate": 6.325927794074882e-06,
      "loss": 0.2568,
      "step": 2740
    },
    {
      "epoch": 17.74,
      "learning_rate": 6.289165653900597e-06,
      "loss": 0.2634,
      "step": 2741
    },
    {
      "epoch": 17.75,
      "learning_rate": 6.2525071756024625e-06,
      "loss": 0.304,
      "step": 2742
    },
    {
      "epoch": 17.75,
      "learning_rate": 6.215952399731661e-06,
      "loss": 0.2893,
      "step": 2743
    },
    {
      "epoch": 17.76,
      "learning_rate": 6.179501366724716e-06,
      "loss": 0.3123,
      "step": 2744
    },
    {
      "epoch": 17.77,
      "learning_rate": 6.143154116903294e-06,
      "loss": 0.2653,
      "step": 2745
    },
    {
      "epoch": 17.77,
      "learning_rate": 6.106910690474354e-06,
      "loss": 0.2639,
      "step": 2746
    },
    {
      "epoch": 17.78,
      "learning_rate": 6.070771127529962e-06,
      "loss": 0.2545,
      "step": 2747
    },
    {
      "epoch": 17.79,
      "learning_rate": 6.034735468047304e-06,
      "loss": 0.2273,
      "step": 2748
    },
    {
      "epoch": 17.79,
      "learning_rate": 5.998803751888604e-06,
      "loss": 0.2783,
      "step": 2749
    },
    {
      "epoch": 17.8,
      "learning_rate": 5.962976018801159e-06,
      "loss": 0.26,
      "step": 2750
    },
    {
      "epoch": 17.81,
      "learning_rate": 5.927252308417164e-06,
      "loss": 0.2786,
      "step": 2751
    },
    {
      "epoch": 17.81,
      "learning_rate": 5.89163266025381e-06,
      "loss": 0.2433,
      "step": 2752
    },
    {
      "epoch": 17.82,
      "learning_rate": 5.8561171137131554e-06,
      "loss": 0.2541,
      "step": 2753
    },
    {
      "epoch": 17.83,
      "learning_rate": 5.820705708082108e-06,
      "loss": 0.2538,
      "step": 2754
    },
    {
      "epoch": 17.83,
      "learning_rate": 5.785398482532334e-06,
      "loss": 0.2872,
      "step": 2755
    },
    {
      "epoch": 17.84,
      "learning_rate": 5.750195476120335e-06,
      "loss": 0.2806,
      "step": 2756
    },
    {
      "epoch": 17.84,
      "learning_rate": 5.7150967277872394e-06,
      "loss": 0.2642,
      "step": 2757
    },
    {
      "epoch": 17.85,
      "learning_rate": 5.680102276358923e-06,
      "loss": 0.2905,
      "step": 2758
    },
    {
      "epoch": 17.86,
      "learning_rate": 5.645212160545832e-06,
      "loss": 0.2725,
      "step": 2759
    },
    {
      "epoch": 17.86,
      "learning_rate": 5.610426418943038e-06,
      "loss": 0.2737,
      "step": 2760
    },
    {
      "epoch": 17.87,
      "learning_rate": 5.575745090030138e-06,
      "loss": 0.2227,
      "step": 2761
    },
    {
      "epoch": 17.88,
      "learning_rate": 5.541168212171244e-06,
      "loss": 0.2548,
      "step": 2762
    },
    {
      "epoch": 17.88,
      "learning_rate": 5.506695823614871e-06,
      "loss": 0.2539,
      "step": 2763
    },
    {
      "epoch": 17.89,
      "learning_rate": 5.472327962494017e-06,
      "loss": 0.2181,
      "step": 2764
    },
    {
      "epoch": 17.9,
      "learning_rate": 5.438064666826026e-06,
      "loss": 0.2516,
      "step": 2765
    },
    {
      "epoch": 17.9,
      "learning_rate": 5.403905974512568e-06,
      "loss": 0.2469,
      "step": 2766
    },
    {
      "epoch": 17.91,
      "learning_rate": 5.369851923339597e-06,
      "loss": 0.2244,
      "step": 2767
    },
    {
      "epoch": 17.92,
      "learning_rate": 5.335902550977357e-06,
      "loss": 0.2919,
      "step": 2768
    },
    {
      "epoch": 17.92,
      "learning_rate": 5.3020578949802166e-06,
      "loss": 0.304,
      "step": 2769
    },
    {
      "epoch": 17.93,
      "learning_rate": 5.268317992786786e-06,
      "loss": 0.2994,
      "step": 2770
    },
    {
      "epoch": 17.94,
      "learning_rate": 5.2346828817197655e-06,
      "loss": 0.2601,
      "step": 2771
    },
    {
      "epoch": 17.94,
      "learning_rate": 5.201152598985915e-06,
      "loss": 0.2671,
      "step": 2772
    },
    {
      "epoch": 17.95,
      "learning_rate": 5.1677271816760855e-06,
      "loss": 0.2628,
      "step": 2773
    },
    {
      "epoch": 17.95,
      "learning_rate": 5.134406666765102e-06,
      "loss": 0.2404,
      "step": 2774
    },
    {
      "epoch": 17.96,
      "learning_rate": 5.101191091111712e-06,
      "loss": 0.2559,
      "step": 2775
    },
    {
      "epoch": 17.97,
      "learning_rate": 5.068080491458638e-06,
      "loss": 0.2427,
      "step": 2776
    },
    {
      "epoch": 17.97,
      "learning_rate": 5.035074904432447e-06,
      "loss": 0.2731,
      "step": 2777
    },
    {
      "epoch": 17.98,
      "learning_rate": 5.002174366543555e-06,
      "loss": 0.2684,
      "step": 2778
    },
    {
      "epoch": 17.99,
      "learning_rate": 4.969378914186173e-06,
      "loss": 0.2582,
      "step": 2779
    },
    {
      "epoch": 17.99,
      "learning_rate": 4.93668858363826e-06,
      "loss": 0.3143,
      "step": 2780
    },
    {
      "epoch": 18.0,
      "learning_rate": 4.9041034110614785e-06,
      "loss": 0.2688,
      "step": 2781
    },
    {
      "epoch": 18.01,
      "learning_rate": 4.871623432501204e-06,
      "loss": 0.2515,
      "step": 2782
    },
    {
      "epoch": 18.01,
      "learning_rate": 4.8392486838864105e-06,
      "loss": 0.2705,
      "step": 2783
    },
    {
      "epoch": 18.02,
      "learning_rate": 4.806979201029693e-06,
      "loss": 0.2345,
      "step": 2784
    },
    {
      "epoch": 18.03,
      "learning_rate": 4.774815019627188e-06,
      "loss": 0.2461,
      "step": 2785
    },
    {
      "epoch": 18.03,
      "learning_rate": 4.742756175258556e-06,
      "loss": 0.319,
      "step": 2786
    },
    {
      "epoch": 18.04,
      "learning_rate": 4.71080270338693e-06,
      "loss": 0.2643,
      "step": 2787
    },
    {
      "epoch": 18.05,
      "learning_rate": 4.678954639358868e-06,
      "loss": 0.2991,
      "step": 2788
    },
    {
      "epoch": 18.05,
      "learning_rate": 4.647212018404357e-06,
      "loss": 0.21,
      "step": 2789
    },
    {
      "epoch": 18.06,
      "learning_rate": 4.61557487563673e-06,
      "loss": 0.2764,
      "step": 2790
    },
    {
      "epoch": 18.06,
      "learning_rate": 4.584043246052649e-06,
      "loss": 0.2317,
      "step": 2791
    },
    {
      "epoch": 18.07,
      "learning_rate": 4.552617164532025e-06,
      "loss": 0.2457,
      "step": 2792
    },
    {
      "epoch": 18.08,
      "learning_rate": 4.52129666583806e-06,
      "loss": 0.2479,
      "step": 2793
    },
    {
      "epoch": 18.08,
      "learning_rate": 4.490081784617117e-06,
      "loss": 0.2698,
      "step": 2794
    },
    {
      "epoch": 18.09,
      "learning_rate": 4.458972555398766e-06,
      "loss": 0.276,
      "step": 2795
    },
    {
      "epoch": 18.1,
      "learning_rate": 4.427969012595679e-06,
      "loss": 0.2266,
      "step": 2796
    },
    {
      "epoch": 18.1,
      "learning_rate": 4.397071190503654e-06,
      "loss": 0.2646,
      "step": 2797
    },
    {
      "epoch": 18.11,
      "learning_rate": 4.366279123301476e-06,
      "loss": 0.2675,
      "step": 2798
    },
    {
      "epoch": 18.12,
      "learning_rate": 4.3355928450509954e-06,
      "loss": 0.2452,
      "step": 2799
    },
    {
      "epoch": 18.12,
      "learning_rate": 4.30501238969705e-06,
      "loss": 0.2781,
      "step": 2800
    },
    {
      "epoch": 18.13,
      "learning_rate": 4.274537791067379e-06,
      "loss": 0.2717,
      "step": 2801
    },
    {
      "epoch": 18.14,
      "learning_rate": 4.244169082872662e-06,
      "loss": 0.2655,
      "step": 2802
    },
    {
      "epoch": 18.14,
      "learning_rate": 4.2139062987064045e-06,
      "loss": 0.2991,
      "step": 2803
    },
    {
      "epoch": 18.15,
      "learning_rate": 4.183749472044973e-06,
      "loss": 0.2852,
      "step": 2804
    },
    {
      "epoch": 18.16,
      "learning_rate": 4.153698636247494e-06,
      "loss": 0.2618,
      "step": 2805
    },
    {
      "epoch": 18.16,
      "learning_rate": 4.123753824555865e-06,
      "loss": 0.2468,
      "step": 2806
    },
    {
      "epoch": 18.17,
      "learning_rate": 4.0939150700947226e-06,
      "loss": 0.2988,
      "step": 2807
    },
    {
      "epoch": 18.17,
      "learning_rate": 4.0641824058713525e-06,
      "loss": 0.255,
      "step": 2808
    },
    {
      "epoch": 18.18,
      "learning_rate": 4.034555864775713e-06,
      "loss": 0.2637,
      "step": 2809
    },
    {
      "epoch": 18.19,
      "learning_rate": 4.00503547958031e-06,
      "loss": 0.3033,
      "step": 2810
    },
    {
      "epoch": 18.19,
      "learning_rate": 3.9756212829403e-06,
      "loss": 0.2745,
      "step": 2811
    },
    {
      "epoch": 18.2,
      "learning_rate": 3.946313307393323e-06,
      "loss": 0.2593,
      "step": 2812
    },
    {
      "epoch": 18.21,
      "learning_rate": 3.917111585359534e-06,
      "loss": 0.2412,
      "step": 2813
    },
    {
      "epoch": 18.21,
      "learning_rate": 3.888016149141549e-06,
      "loss": 0.2274,
      "step": 2814
    },
    {
      "epoch": 18.22,
      "learning_rate": 3.859027030924445e-06,
      "loss": 0.3152,
      "step": 2815
    },
    {
      "epoch": 18.23,
      "learning_rate": 3.830144262775614e-06,
      "loss": 0.2284,
      "step": 2816
    },
    {
      "epoch": 18.23,
      "learning_rate": 3.801367876644868e-06,
      "loss": 0.2612,
      "step": 2817
    },
    {
      "epoch": 18.24,
      "learning_rate": 3.772697904364342e-06,
      "loss": 0.2699,
      "step": 2818
    },
    {
      "epoch": 18.25,
      "learning_rate": 3.7441343776484117e-06,
      "loss": 0.2568,
      "step": 2819
    },
    {
      "epoch": 18.25,
      "learning_rate": 3.7156773280937473e-06,
      "loss": 0.2251,
      "step": 2820
    },
    {
      "epoch": 18.26,
      "learning_rate": 3.687326787179202e-06,
      "loss": 0.2375,
      "step": 2821
    },
    {
      "epoch": 18.27,
      "learning_rate": 3.6590827862658437e-06,
      "loss": 0.2313,
      "step": 2822
    },
    {
      "epoch": 18.27,
      "learning_rate": 3.6309453565968597e-06,
      "loss": 0.2654,
      "step": 2823
    },
    {
      "epoch": 18.28,
      "learning_rate": 3.602914529297563e-06,
      "loss": 0.2137,
      "step": 2824
    },
    {
      "epoch": 18.28,
      "learning_rate": 3.57499033537535e-06,
      "loss": 0.2841,
      "step": 2825
    },
    {
      "epoch": 18.29,
      "learning_rate": 3.547172805719645e-06,
      "loss": 0.2542,
      "step": 2826
    },
    {
      "epoch": 18.3,
      "learning_rate": 3.5194619711018783e-06,
      "loss": 0.3201,
      "step": 2827
    },
    {
      "epoch": 18.3,
      "learning_rate": 3.4918578621754737e-06,
      "loss": 0.2653,
      "step": 2828
    },
    {
      "epoch": 18.31,
      "learning_rate": 3.4643605094757945e-06,
      "loss": 0.2869,
      "step": 2829
    },
    {
      "epoch": 18.32,
      "learning_rate": 3.4369699434200875e-06,
      "loss": 0.2561,
      "step": 2830
    },
    {
      "epoch": 18.32,
      "learning_rate": 3.409686194307504e-06,
      "loss": 0.2628,
      "step": 2831
    },
    {
      "epoch": 18.33,
      "learning_rate": 3.382509292319025e-06,
      "loss": 0.2906,
      "step": 2832
    },
    {
      "epoch": 18.34,
      "learning_rate": 3.355439267517424e-06,
      "loss": 0.2631,
      "step": 2833
    },
    {
      "epoch": 18.34,
      "learning_rate": 3.3284761498472593e-06,
      "loss": 0.283,
      "step": 2834
    },
    {
      "epoch": 18.35,
      "learning_rate": 3.3016199691348504e-06,
      "loss": 0.2273,
      "step": 2835
    },
    {
      "epoch": 18.36,
      "learning_rate": 3.2748707550881665e-06,
      "loss": 0.2837,
      "step": 2836
    },
    {
      "epoch": 18.36,
      "learning_rate": 3.2482285372969157e-06,
      "loss": 0.3051,
      "step": 2837
    },
    {
      "epoch": 18.37,
      "learning_rate": 3.221693345232435e-06,
      "loss": 0.2656,
      "step": 2838
    },
    {
      "epoch": 18.38,
      "learning_rate": 3.1952652082476107e-06,
      "loss": 0.2788,
      "step": 2839
    },
    {
      "epoch": 18.38,
      "learning_rate": 3.1689441555769917e-06,
      "loss": 0.257,
      "step": 2840
    },
    {
      "epoch": 18.39,
      "learning_rate": 3.142730216336609e-06,
      "loss": 0.265,
      "step": 2841
    },
    {
      "epoch": 18.39,
      "learning_rate": 3.116623419524045e-06,
      "loss": 0.2166,
      "step": 2842
    },
    {
      "epoch": 18.4,
      "learning_rate": 3.0906237940183436e-06,
      "loss": 0.27,
      "step": 2843
    },
    {
      "epoch": 18.41,
      "learning_rate": 3.064731368579998e-06,
      "loss": 0.28,
      "step": 2844
    },
    {
      "epoch": 18.41,
      "learning_rate": 3.0389461718509095e-06,
      "loss": 0.2866,
      "step": 2845
    },
    {
      "epoch": 18.42,
      "learning_rate": 3.0132682323543825e-06,
      "loss": 0.283,
      "step": 2846
    },
    {
      "epoch": 18.43,
      "learning_rate": 2.9876975784950633e-06,
      "loss": 0.2494,
      "step": 2847
    },
    {
      "epoch": 18.43,
      "learning_rate": 2.9622342385589254e-06,
      "loss": 0.2592,
      "step": 2848
    },
    {
      "epoch": 18.44,
      "learning_rate": 2.936878240713248e-06,
      "loss": 0.2593,
      "step": 2849
    },
    {
      "epoch": 18.45,
      "learning_rate": 2.911629613006517e-06,
      "loss": 0.2607,
      "step": 2850
    },
    {
      "epoch": 18.45,
      "learning_rate": 2.886488383368513e-06,
      "loss": 0.2766,
      "step": 2851
    },
    {
      "epoch": 18.46,
      "learning_rate": 2.8614545796101565e-06,
      "loss": 0.3129,
      "step": 2852
    },
    {
      "epoch": 18.47,
      "learning_rate": 2.836528229423574e-06,
      "loss": 0.2354,
      "step": 2853
    },
    {
      "epoch": 18.47,
      "learning_rate": 2.81170936038202e-06,
      "loss": 0.3031,
      "step": 2854
    },
    {
      "epoch": 18.48,
      "learning_rate": 2.7869979999398553e-06,
      "loss": 0.2496,
      "step": 2855
    },
    {
      "epoch": 18.49,
      "learning_rate": 2.762394175432481e-06,
      "loss": 0.3127,
      "step": 2856
    },
    {
      "epoch": 18.49,
      "learning_rate": 2.7378979140763926e-06,
      "loss": 0.2615,
      "step": 2857
    },
    {
      "epoch": 18.5,
      "learning_rate": 2.7135092429690813e-06,
      "loss": 0.2366,
      "step": 2858
    },
    {
      "epoch": 18.5,
      "learning_rate": 2.6892281890890216e-06,
      "loss": 0.2782,
      "step": 2859
    },
    {
      "epoch": 18.51,
      "learning_rate": 2.665054779295639e-06,
      "loss": 0.2751,
      "step": 2860
    },
    {
      "epoch": 18.52,
      "learning_rate": 2.6409890403293004e-06,
      "loss": 0.2573,
      "step": 2861
    },
    {
      "epoch": 18.52,
      "learning_rate": 2.617030998811243e-06,
      "loss": 0.2411,
      "step": 2862
    },
    {
      "epoch": 18.53,
      "learning_rate": 2.593180681243579e-06,
      "loss": 0.3026,
      "step": 2863
    },
    {
      "epoch": 18.54,
      "learning_rate": 2.5694381140092817e-06,
      "loss": 0.2782,
      "step": 2864
    },
    {
      "epoch": 18.54,
      "learning_rate": 2.545803323372109e-06,
      "loss": 0.2623,
      "step": 2865
    },
    {
      "epoch": 18.55,
      "learning_rate": 2.522276335476592e-06,
      "loss": 0.3408,
      "step": 2866
    },
    {
      "epoch": 18.56,
      "learning_rate": 2.498857176348024e-06,
      "loss": 0.2616,
      "step": 2867
    },
    {
      "epoch": 18.56,
      "learning_rate": 2.475545871892426e-06,
      "loss": 0.2629,
      "step": 2868
    },
    {
      "epoch": 18.57,
      "learning_rate": 2.452342447896483e-06,
      "loss": 0.2627,
      "step": 2869
    },
    {
      "epoch": 18.58,
      "learning_rate": 2.4292469300275842e-06,
      "loss": 0.281,
      "step": 2870
    },
    {
      "epoch": 18.58,
      "learning_rate": 2.4062593438337165e-06,
      "loss": 0.2382,
      "step": 2871
    },
    {
      "epoch": 18.59,
      "learning_rate": 2.3833797147435054e-06,
      "loss": 0.2864,
      "step": 2872
    },
    {
      "epoch": 18.6,
      "learning_rate": 2.3606080680661392e-06,
      "loss": 0.2238,
      "step": 2873
    },
    {
      "epoch": 18.6,
      "learning_rate": 2.3379444289913342e-06,
      "loss": 0.2625,
      "step": 2874
    },
    {
      "epoch": 18.61,
      "learning_rate": 2.3153888225893814e-06,
      "loss": 0.2646,
      "step": 2875
    },
    {
      "epoch": 18.61,
      "learning_rate": 2.2929412738110334e-06,
      "loss": 0.2582,
      "step": 2876
    },
    {
      "epoch": 18.62,
      "learning_rate": 2.2706018074875045e-06,
      "loss": 0.2448,
      "step": 2877
    },
    {
      "epoch": 18.63,
      "learning_rate": 2.2483704483304833e-06,
      "loss": 0.2407,
      "step": 2878
    },
    {
      "epoch": 18.63,
      "learning_rate": 2.226247220932043e-06,
      "loss": 0.2448,
      "step": 2879
    },
    {
      "epoch": 18.64,
      "learning_rate": 2.204232149764651e-06,
      "loss": 0.2487,
      "step": 2880
    },
    {
      "epoch": 18.65,
      "learning_rate": 2.1823252591811173e-06,
      "loss": 0.2457,
      "step": 2881
    },
    {
      "epoch": 18.65,
      "learning_rate": 2.160526573414623e-06,
      "loss": 0.266,
      "step": 2882
    },
    {
      "epoch": 18.66,
      "learning_rate": 2.1388361165786017e-06,
      "loss": 0.2375,
      "step": 2883
    },
    {
      "epoch": 18.67,
      "learning_rate": 2.1172539126668167e-06,
      "loss": 0.2599,
      "step": 2884
    },
    {
      "epoch": 18.67,
      "learning_rate": 2.09577998555327e-06,
      "loss": 0.2506,
      "step": 2885
    },
    {
      "epoch": 18.68,
      "learning_rate": 2.0744143589921494e-06,
      "loss": 0.2932,
      "step": 2886
    },
    {
      "epoch": 18.69,
      "learning_rate": 2.0531570566178828e-06,
      "loss": 0.2379,
      "step": 2887
    },
    {
      "epoch": 18.69,
      "learning_rate": 2.032008101945071e-06,
      "loss": 0.2893,
      "step": 2888
    },
    {
      "epoch": 18.7,
      "learning_rate": 2.010967518368434e-06,
      "loss": 0.2089,
      "step": 2889
    },
    {
      "epoch": 18.71,
      "learning_rate": 1.9900353291628427e-06,
      "loss": 0.2345,
      "step": 2890
    },
    {
      "epoch": 18.71,
      "learning_rate": 1.9692115574832637e-06,
      "loss": 0.2502,
      "step": 2891
    },
    {
      "epoch": 18.72,
      "learning_rate": 1.948496226364682e-06,
      "loss": 0.2645,
      "step": 2892
    },
    {
      "epoch": 18.72,
      "learning_rate": 1.9278893587221904e-06,
      "loss": 0.2596,
      "step": 2893
    },
    {
      "epoch": 18.73,
      "learning_rate": 1.9073909773508757e-06,
      "loss": 0.2008,
      "step": 2894
    },
    {
      "epoch": 18.74,
      "learning_rate": 1.8870011049258118e-06,
      "loss": 0.2653,
      "step": 2895
    },
    {
      "epoch": 18.74,
      "learning_rate": 1.8667197640020561e-06,
      "loss": 0.2692,
      "step": 2896
    },
    {
      "epoch": 18.75,
      "learning_rate": 1.846546977014596e-06,
      "loss": 0.2624,
      "step": 2897
    },
    {
      "epoch": 18.76,
      "learning_rate": 1.8264827662783369e-06,
      "loss": 0.2571,
      "step": 2898
    },
    {
      "epoch": 18.76,
      "learning_rate": 1.8065271539880801e-06,
      "loss": 0.2725,
      "step": 2899
    },
    {
      "epoch": 18.77,
      "learning_rate": 1.7866801622185125e-06,
      "loss": 0.2579,
      "step": 2900
    },
    {
      "epoch": 18.78,
      "learning_rate": 1.76694181292415e-06,
      "loss": 0.2928,
      "step": 2901
    },
    {
      "epoch": 18.78,
      "learning_rate": 1.747312127939349e-06,
      "loss": 0.238,
      "step": 2902
    },
    {
      "epoch": 18.79,
      "learning_rate": 1.7277911289782178e-06,
      "loss": 0.2803,
      "step": 2903
    },
    {
      "epoch": 18.8,
      "learning_rate": 1.7083788376346833e-06,
      "loss": 0.2542,
      "step": 2904
    },
    {
      "epoch": 18.8,
      "learning_rate": 1.6890752753824013e-06,
      "loss": 0.3142,
      "step": 2905
    },
    {
      "epoch": 18.81,
      "learning_rate": 1.6698804635747579e-06,
      "loss": 0.3033,
      "step": 2906
    },
    {
      "epoch": 18.82,
      "learning_rate": 1.6507944234448347e-06,
      "loss": 0.2852,
      "step": 2907
    },
    {
      "epoch": 18.82,
      "learning_rate": 1.6318171761054213e-06,
      "loss": 0.2068,
      "step": 2908
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.612948742548892e-06,
      "loss": 0.2321,
      "step": 2909
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.5941891436473289e-06,
      "loss": 0.2877,
      "step": 2910
    },
    {
      "epoch": 18.84,
      "learning_rate": 1.5755384001523875e-06,
      "loss": 0.2969,
      "step": 2911
    },
    {
      "epoch": 18.85,
      "learning_rate": 1.5569965326952984e-06,
      "loss": 0.2916,
      "step": 2912
    },
    {
      "epoch": 18.85,
      "learning_rate": 1.5385635617868877e-06,
      "loss": 0.2488,
      "step": 2913
    },
    {
      "epoch": 18.86,
      "learning_rate": 1.520239507817467e-06,
      "loss": 0.2701,
      "step": 2914
    },
    {
      "epoch": 18.87,
      "learning_rate": 1.5020243910569442e-06,
      "loss": 0.2692,
      "step": 2915
    },
    {
      "epoch": 18.87,
      "learning_rate": 1.4839182316546462e-06,
      "loss": 0.2449,
      "step": 2916
    },
    {
      "epoch": 18.88,
      "learning_rate": 1.4659210496394072e-06,
      "loss": 0.2494,
      "step": 2917
    },
    {
      "epoch": 18.89,
      "learning_rate": 1.4480328649195129e-06,
      "loss": 0.2417,
      "step": 2918
    },
    {
      "epoch": 18.89,
      "learning_rate": 1.4302536972826796e-06,
      "loss": 0.2665,
      "step": 2919
    },
    {
      "epoch": 18.9,
      "learning_rate": 1.4125835663960196e-06,
      "loss": 0.2534,
      "step": 2920
    },
    {
      "epoch": 18.91,
      "learning_rate": 1.3950224918060306e-06,
      "loss": 0.2775,
      "step": 2921
    },
    {
      "epoch": 18.91,
      "learning_rate": 1.3775704929385846e-06,
      "loss": 0.2717,
      "step": 2922
    },
    {
      "epoch": 18.92,
      "learning_rate": 1.3602275890988724e-06,
      "loss": 0.2626,
      "step": 2923
    },
    {
      "epoch": 18.93,
      "learning_rate": 1.3429937994714259e-06,
      "loss": 0.2563,
      "step": 2924
    },
    {
      "epoch": 18.93,
      "learning_rate": 1.3258691431200844e-06,
      "loss": 0.2743,
      "step": 2925
    },
    {
      "epoch": 18.94,
      "learning_rate": 1.308853638987939e-06,
      "loss": 0.2603,
      "step": 2926
    },
    {
      "epoch": 18.94,
      "learning_rate": 1.2919473058973453e-06,
      "loss": 0.2654,
      "step": 2927
    },
    {
      "epoch": 18.95,
      "learning_rate": 1.2751501625499096e-06,
      "loss": 0.2234,
      "step": 2928
    },
    {
      "epoch": 18.96,
      "learning_rate": 1.2584622275264468e-06,
      "loss": 0.2239,
      "step": 2929
    },
    {
      "epoch": 18.96,
      "learning_rate": 1.241883519286946e-06,
      "loss": 0.23,
      "step": 2930
    },
    {
      "epoch": 18.97,
      "learning_rate": 1.2254140561706152e-06,
      "loss": 0.2589,
      "step": 2931
    },
    {
      "epoch": 18.98,
      "learning_rate": 1.2090538563957587e-06,
      "loss": 0.3269,
      "step": 2932
    },
    {
      "epoch": 18.98,
      "learning_rate": 1.1928029380598671e-06,
      "loss": 0.3799,
      "step": 2933
    },
    {
      "epoch": 18.99,
      "learning_rate": 1.176661319139516e-06,
      "loss": 0.223,
      "step": 2934
    },
    {
      "epoch": 19.0,
      "learning_rate": 1.160629017490389e-06,
      "loss": 0.2517,
      "step": 2935
    },
    {
      "epoch": 19.0,
      "learning_rate": 1.144706050847233e-06,
      "loss": 0.2152,
      "step": 2936
    },
    {
      "epoch": 19.01,
      "learning_rate": 1.1288924368238695e-06,
      "loss": 0.2404,
      "step": 2937
    },
    {
      "epoch": 19.02,
      "learning_rate": 1.1131881929131283e-06,
      "loss": 0.2534,
      "step": 2938
    },
    {
      "epoch": 19.02,
      "learning_rate": 1.0975933364868685e-06,
      "loss": 0.2616,
      "step": 2939
    },
    {
      "epoch": 19.03,
      "learning_rate": 1.0821078847959465e-06,
      "loss": 0.2443,
      "step": 2940
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.0667318549702043e-06,
      "loss": 0.2139,
      "step": 2941
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.0514652640184253e-06,
      "loss": 0.2756,
      "step": 2942
    },
    {
      "epoch": 19.05,
      "learning_rate": 1.0363081288283448e-06,
      "loss": 0.2639,
      "step": 2943
    },
    {
      "epoch": 19.06,
      "learning_rate": 1.0212604661666182e-06,
      "loss": 0.2896,
      "step": 2944
    },
    {
      "epoch": 19.06,
      "learning_rate": 1.0063222926787853e-06,
      "loss": 0.2046,
      "step": 2945
    },
    {
      "epoch": 19.07,
      "learning_rate": 9.91493624889317e-07,
      "loss": 0.2511,
      "step": 2946
    },
    {
      "epoch": 19.07,
      "learning_rate": 9.767744792014922e-07,
      "loss": 0.2549,
      "step": 2947
    },
    {
      "epoch": 19.08,
      "learning_rate": 9.621648718974753e-07,
      "loss": 0.2717,
      "step": 2948
    },
    {
      "epoch": 19.09,
      "learning_rate": 9.476648191382498e-07,
      "loss": 0.254,
      "step": 2949
    },
    {
      "epoch": 19.09,
      "learning_rate": 9.332743369636077e-07,
      "loss": 0.291,
      "step": 2950
    },
    {
      "epoch": 19.1,
      "learning_rate": 9.189934412921486e-07,
      "loss": 0.2406,
      "step": 2951
    },
    {
      "epoch": 19.11,
      "learning_rate": 9.04822147921236e-07,
      "loss": 0.2961,
      "step": 2952
    },
    {
      "epoch": 19.11,
      "learning_rate": 8.907604725269969e-07,
      "loss": 0.2864,
      "step": 2953
    },
    {
      "epoch": 19.12,
      "learning_rate": 8.768084306642888e-07,
      "loss": 0.2654,
      "step": 2954
    },
    {
      "epoch": 19.13,
      "learning_rate": 8.629660377667215e-07,
      "loss": 0.282,
      "step": 2955
    },
    {
      "epoch": 19.13,
      "learning_rate": 8.492333091465799e-07,
      "loss": 0.3025,
      "step": 2956
    },
    {
      "epoch": 19.14,
      "learning_rate": 8.356102599948568e-07,
      "loss": 0.2253,
      "step": 2957
    },
    {
      "epoch": 19.15,
      "learning_rate": 8.22096905381231e-07,
      "loss": 0.2942,
      "step": 2958
    },
    {
      "epoch": 19.15,
      "learning_rate": 8.086932602540009e-07,
      "loss": 0.2212,
      "step": 2959
    },
    {
      "epoch": 19.16,
      "learning_rate": 7.953993394401615e-07,
      "loss": 0.28,
      "step": 2960
    },
    {
      "epoch": 19.17,
      "learning_rate": 7.822151576452607e-07,
      "loss": 0.2621,
      "step": 2961
    },
    {
      "epoch": 19.17,
      "learning_rate": 7.691407294535214e-07,
      "loss": 0.2296,
      "step": 2962
    },
    {
      "epoch": 19.18,
      "learning_rate": 7.561760693277187e-07,
      "loss": 0.299,
      "step": 2963
    },
    {
      "epoch": 19.18,
      "learning_rate": 7.433211916092142e-07,
      "loss": 0.2261,
      "step": 2964
    },
    {
      "epoch": 19.19,
      "learning_rate": 7.305761105179332e-07,
      "loss": 0.2729,
      "step": 2965
    },
    {
      "epoch": 19.2,
      "learning_rate": 7.179408401523313e-07,
      "loss": 0.303,
      "step": 2966
    },
    {
      "epoch": 19.2,
      "learning_rate": 7.054153944894171e-07,
      "loss": 0.2808,
      "step": 2967
    },
    {
      "epoch": 19.21,
      "learning_rate": 6.92999787384696e-07,
      "loss": 0.3069,
      "step": 2968
    },
    {
      "epoch": 19.22,
      "learning_rate": 6.806940325721711e-07,
      "loss": 0.244,
      "step": 2969
    },
    {
      "epoch": 19.22,
      "learning_rate": 6.684981436643201e-07,
      "loss": 0.2582,
      "step": 2970
    },
    {
      "epoch": 19.23,
      "learning_rate": 6.564121341521179e-07,
      "loss": 0.2819,
      "step": 2971
    },
    {
      "epoch": 19.24,
      "learning_rate": 6.444360174049702e-07,
      "loss": 0.2959,
      "step": 2972
    },
    {
      "epoch": 19.24,
      "learning_rate": 6.32569806670702e-07,
      "loss": 0.2576,
      "step": 2973
    },
    {
      "epoch": 19.25,
      "learning_rate": 6.208135150756023e-07,
      "loss": 0.3036,
      "step": 2974
    },
    {
      "epoch": 19.26,
      "learning_rate": 6.091671556243461e-07,
      "loss": 0.25,
      "step": 2975
    },
    {
      "epoch": 19.26,
      "learning_rate": 5.97630741200006e-07,
      "loss": 0.295,
      "step": 2976
    },
    {
      "epoch": 19.27,
      "learning_rate": 5.862042845640403e-07,
      "loss": 0.2646,
      "step": 2977
    },
    {
      "epoch": 19.28,
      "learning_rate": 5.748877983562496e-07,
      "loss": 0.2238,
      "step": 2978
    },
    {
      "epoch": 19.28,
      "learning_rate": 5.636812950948201e-07,
      "loss": 0.2271,
      "step": 2979
    },
    {
      "epoch": 19.29,
      "learning_rate": 5.525847871762579e-07,
      "loss": 0.2561,
      "step": 2980
    },
    {
      "epoch": 19.29,
      "learning_rate": 5.415982868753888e-07,
      "loss": 0.2268,
      "step": 2981
    },
    {
      "epoch": 19.3,
      "learning_rate": 5.307218063453578e-07,
      "loss": 0.2683,
      "step": 2982
    },
    {
      "epoch": 19.31,
      "learning_rate": 5.199553576176075e-07,
      "loss": 0.2928,
      "step": 2983
    },
    {
      "epoch": 19.31,
      "learning_rate": 5.092989526018776e-07,
      "loss": 0.3021,
      "step": 2984
    },
    {
      "epoch": 19.32,
      "learning_rate": 4.987526030861389e-07,
      "loss": 0.2554,
      "step": 2985
    },
    {
      "epoch": 19.33,
      "learning_rate": 4.883163207366592e-07,
      "loss": 0.2921,
      "step": 2986
    },
    {
      "epoch": 19.33,
      "learning_rate": 4.779901170979373e-07,
      "loss": 0.2704,
      "step": 2987
    },
    {
      "epoch": 19.34,
      "learning_rate": 4.677740035927025e-07,
      "loss": 0.2091,
      "step": 2988
    },
    {
      "epoch": 19.35,
      "learning_rate": 4.5766799152190397e-07,
      "loss": 0.3088,
      "step": 2989
    },
    {
      "epoch": 19.35,
      "learning_rate": 4.476720920646993e-07,
      "loss": 0.2589,
      "step": 2990
    },
    {
      "epoch": 19.36,
      "learning_rate": 4.377863162784435e-07,
      "loss": 0.2268,
      "step": 2991
    },
    {
      "epoch": 19.37,
      "learning_rate": 4.2801067509865567e-07,
      "loss": 0.2542,
      "step": 2992
    },
    {
      "epoch": 19.37,
      "learning_rate": 4.1834517933907467e-07,
      "loss": 0.2018,
      "step": 2993
    },
    {
      "epoch": 19.38,
      "learning_rate": 4.0878983969152574e-07,
      "loss": 0.2351,
      "step": 2994
    },
    {
      "epoch": 19.39,
      "learning_rate": 3.9934466672604256e-07,
      "loss": 0.2598,
      "step": 2995
    },
    {
      "epoch": 19.39,
      "learning_rate": 3.9000967089076743e-07,
      "loss": 0.2056,
      "step": 2996
    },
    {
      "epoch": 19.4,
      "learning_rate": 3.8078486251196255e-07,
      "loss": 0.2244,
      "step": 2997
    },
    {
      "epoch": 19.4,
      "learning_rate": 3.716702517940096e-07,
      "loss": 0.2485,
      "step": 2998
    },
    {
      "epoch": 19.41,
      "learning_rate": 3.626658488193768e-07,
      "loss": 0.2441,
      "step": 2999
    },
    {
      "epoch": 19.42,
      "learning_rate": 3.537716635486521e-07,
      "loss": 0.2908,
      "step": 3000
    },
    {
      "epoch": 19.42,
      "learning_rate": 3.4498770582047646e-07,
      "loss": 0.2477,
      "step": 3001
    },
    {
      "epoch": 19.43,
      "learning_rate": 3.363139853515773e-07,
      "loss": 0.2416,
      "step": 3002
    },
    {
      "epoch": 19.44,
      "learning_rate": 3.277505117367019e-07,
      "loss": 0.236,
      "step": 3003
    },
    {
      "epoch": 19.44,
      "learning_rate": 3.1929729444869494e-07,
      "loss": 0.2686,
      "step": 3004
    },
    {
      "epoch": 19.45,
      "learning_rate": 3.1095434283838764e-07,
      "loss": 0.2271,
      "step": 3005
    },
    {
      "epoch": 19.46,
      "learning_rate": 3.0272166613467544e-07,
      "loss": 0.2339,
      "step": 3006
    },
    {
      "epoch": 19.46,
      "learning_rate": 2.945992734444514e-07,
      "loss": 0.2373,
      "step": 3007
    },
    {
      "epoch": 19.47,
      "learning_rate": 2.86587173752606e-07,
      "loss": 0.2626,
      "step": 3008
    },
    {
      "epoch": 19.48,
      "learning_rate": 2.786853759220498e-07,
      "loss": 0.2596,
      "step": 3009
    },
    {
      "epoch": 19.48,
      "learning_rate": 2.7089388869364627e-07,
      "loss": 0.3069,
      "step": 3010
    },
    {
      "epoch": 19.49,
      "learning_rate": 2.632127206862678e-07,
      "loss": 0.2103,
      "step": 3011
    },
    {
      "epoch": 19.5,
      "learning_rate": 2.5564188039671754e-07,
      "loss": 0.2931,
      "step": 3012
    },
    {
      "epoch": 19.5,
      "learning_rate": 2.4818137619979644e-07,
      "loss": 0.3159,
      "step": 3013
    },
    {
      "epoch": 19.51,
      "learning_rate": 2.4083121634821406e-07,
      "loss": 0.2427,
      "step": 3014
    },
    {
      "epoch": 19.51,
      "learning_rate": 2.335914089726332e-07,
      "loss": 0.2664,
      "step": 3015
    },
    {
      "epoch": 19.52,
      "learning_rate": 2.2646196208165882e-07,
      "loss": 0.2839,
      "step": 3016
    },
    {
      "epoch": 19.53,
      "learning_rate": 2.1944288356180454e-07,
      "loss": 0.2324,
      "step": 3017
    },
    {
      "epoch": 19.53,
      "learning_rate": 2.125341811774928e-07,
      "loss": 0.2375,
      "step": 3018
    },
    {
      "epoch": 19.54,
      "learning_rate": 2.057358625710437e-07,
      "loss": 0.2336,
      "step": 3019
    },
    {
      "epoch": 19.55,
      "learning_rate": 1.9904793526269726e-07,
      "loss": 0.2366,
      "step": 3020
    },
    {
      "epoch": 19.55,
      "learning_rate": 1.9247040665055782e-07,
      "loss": 0.3032,
      "step": 3021
    },
    {
      "epoch": 19.56,
      "learning_rate": 1.8600328401061629e-07,
      "loss": 0.2717,
      "step": 3022
    },
    {
      "epoch": 19.57,
      "learning_rate": 1.7964657449672795e-07,
      "loss": 0.2916,
      "step": 3023
    },
    {
      "epoch": 19.57,
      "learning_rate": 1.7340028514061245e-07,
      "loss": 0.2355,
      "step": 3024
    },
    {
      "epoch": 19.58,
      "learning_rate": 1.6726442285184275e-07,
      "loss": 0.2712,
      "step": 3025
    },
    {
      "epoch": 19.59,
      "learning_rate": 1.6123899441784496e-07,
      "loss": 0.2404,
      "step": 3026
    },
    {
      "epoch": 19.59,
      "learning_rate": 1.5532400650388745e-07,
      "loss": 0.2808,
      "step": 3027
    },
    {
      "epoch": 19.6,
      "learning_rate": 1.4951946565304742e-07,
      "loss": 0.2292,
      "step": 3028
    },
    {
      "epoch": 19.61,
      "learning_rate": 1.4382537828625531e-07,
      "loss": 0.2283,
      "step": 3029
    },
    {
      "epoch": 19.61,
      "learning_rate": 1.3824175070225042e-07,
      "loss": 0.2292,
      "step": 3030
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.3276858907756983e-07,
      "loss": 0.2646,
      "step": 3031
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.2740589946655946e-07,
      "loss": 0.256,
      "step": 3032
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.2215368780138515e-07,
      "loss": 0.2635,
      "step": 3033
    },
    {
      "epoch": 19.64,
      "learning_rate": 1.1701195989196611e-07,
      "loss": 0.2518,
      "step": 3034
    },
    {
      "epoch": 19.64,
      "learning_rate": 1.1198072142603045e-07,
      "loss": 0.2452,
      "step": 3035
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.0705997796908174e-07,
      "loss": 0.2638,
      "step": 3036
    },
    {
      "epoch": 19.66,
      "learning_rate": 1.0224973496438806e-07,
      "loss": 0.2594,
      "step": 3037
    },
    {
      "epoch": 19.66,
      "learning_rate": 9.75499977329819e-08,
      "loss": 0.2408,
      "step": 3038
    },
    {
      "epoch": 19.67,
      "learning_rate": 9.296077147366023e-08,
      "loss": 0.2484,
      "step": 3039
    },
    {
      "epoch": 19.68,
      "learning_rate": 8.848206126299552e-08,
      "loss": 0.2993,
      "step": 3040
    },
    {
      "epoch": 19.68,
      "learning_rate": 8.411387205525812e-08,
      "loss": 0.2783,
      "step": 3041
    },
    {
      "epoch": 19.69,
      "learning_rate": 7.985620868249388e-08,
      "loss": 0.2615,
      "step": 3042
    },
    {
      "epoch": 19.7,
      "learning_rate": 7.570907585450204e-08,
      "loss": 0.2083,
      "step": 3043
    },
    {
      "epoch": 19.7,
      "learning_rate": 7.167247815877964e-08,
      "loss": 0.2874,
      "step": 3044
    },
    {
      "epoch": 19.71,
      "learning_rate": 6.774642006056597e-08,
      "loss": 0.2241,
      "step": 3045
    },
    {
      "epoch": 19.72,
      "learning_rate": 6.393090590283146e-08,
      "loss": 0.2969,
      "step": 3046
    },
    {
      "epoch": 19.72,
      "learning_rate": 6.022593990624437e-08,
      "loss": 0.3037,
      "step": 3047
    },
    {
      "epoch": 19.73,
      "learning_rate": 5.6631526169192987e-08,
      "loss": 0.1975,
      "step": 3048
    },
    {
      "epoch": 19.73,
      "learning_rate": 5.3147668667796746e-08,
      "loss": 0.2386,
      "step": 3049
    },
    {
      "epoch": 19.74,
      "learning_rate": 4.9774371255839613e-08,
      "loss": 0.2388,
      "step": 3050
    },
    {
      "epoch": 19.75,
      "learning_rate": 4.651163766484779e-08,
      "loss": 0.3192,
      "step": 3051
    },
    {
      "epoch": 19.75,
      "learning_rate": 4.335947150400088e-08,
      "loss": 0.2108,
      "step": 3052
    },
    {
      "epoch": 19.76,
      "learning_rate": 4.031787626018746e-08,
      "loss": 0.2313,
      "step": 3053
    },
    {
      "epoch": 19.77,
      "learning_rate": 3.738685529800501e-08,
      "loss": 0.2499,
      "step": 3054
    },
    {
      "epoch": 19.77,
      "learning_rate": 3.4566411859704436e-08,
      "loss": 0.2546,
      "step": 3055
    },
    {
      "epoch": 19.78,
      "learning_rate": 3.18565490652345e-08,
      "loss": 0.2499,
      "step": 3056
    },
    {
      "epoch": 19.79,
      "learning_rate": 2.925726991220845e-08,
      "loss": 0.2544,
      "step": 3057
    },
    {
      "epoch": 19.79,
      "learning_rate": 2.6768577275915196e-08,
      "loss": 0.2168,
      "step": 3058
    },
    {
      "epoch": 19.8,
      "learning_rate": 2.4390473909330358e-08,
      "loss": 0.2737,
      "step": 3059
    },
    {
      "epoch": 19.81,
      "learning_rate": 2.2122962443082984e-08,
      "loss": 0.281,
      "step": 3060
    },
    {
      "epoch": 19.81,
      "learning_rate": 1.9966045385466648e-08,
      "loss": 0.2178,
      "step": 3061
    },
    {
      "epoch": 19.82,
      "learning_rate": 1.7919725122439446e-08,
      "loss": 0.2682,
      "step": 3062
    },
    {
      "epoch": 19.83,
      "learning_rate": 1.59840039176129e-08,
      "loss": 0.2233,
      "step": 3063
    },
    {
      "epoch": 19.83,
      "learning_rate": 1.4158883912263054e-08,
      "loss": 0.2357,
      "step": 3064
    },
    {
      "epoch": 19.84,
      "learning_rate": 1.244436712533048e-08,
      "loss": 0.2581,
      "step": 3065
    },
    {
      "epoch": 19.84,
      "learning_rate": 1.0840455453375863e-08,
      "loss": 0.2792,
      "step": 3066
    },
    {
      "epoch": 19.85,
      "learning_rate": 9.347150670635518e-09,
      "loss": 0.2621,
      "step": 3067
    },
    {
      "epoch": 19.86,
      "learning_rate": 7.964454428976976e-09,
      "loss": 0.2306,
      "step": 3068
    },
    {
      "epoch": 19.86,
      "learning_rate": 6.6923682579433935e-09,
      "loss": 0.2936,
      "step": 3069
    },
    {
      "epoch": 19.87,
      "learning_rate": 5.530893564686945e-09,
      "loss": 0.2849,
      "step": 3070
    },
    {
      "epoch": 19.88,
      "learning_rate": 4.480031634024329e-09,
      "loss": 0.2233,
      "step": 3071
    },
    {
      "epoch": 19.88,
      "learning_rate": 3.5397836284034593e-09,
      "loss": 0.271,
      "step": 3072
    },
    {
      "epoch": 19.89,
      "learning_rate": 2.710150587925675e-09,
      "loss": 0.2257,
      "step": 3073
    },
    {
      "epoch": 19.9,
      "learning_rate": 1.9911334303013284e-09,
      "loss": 0.2496,
      "step": 3074
    },
    {
      "epoch": 19.9,
      "learning_rate": 1.382732950927501e-09,
      "loss": 0.3008,
      "step": 3075
    },
    {
      "epoch": 19.91,
      "learning_rate": 8.849498227880837e-10,
      "loss": 0.2167,
      "step": 3076
    },
    {
      "epoch": 19.92,
      "learning_rate": 4.977845965203898e-10,
      "loss": 0.2153,
      "step": 3077
    },
    {
      "epoch": 19.92,
      "learning_rate": 2.212377004262578e-10,
      "loss": 0.222,
      "step": 3078
    },
    {
      "epoch": 19.93,
      "learning_rate": 5.530944040543773e-11,
      "loss": 0.2762,
      "step": 3079
    },
    {
      "epoch": 19.94,
      "learning_rate": 0.0,
      "loss": 0.2504,
      "step": 3080
    },
    {
      "epoch": 19.94,
      "step": 3080,
      "total_flos": 2.645806671380611e+18,
      "train_loss": 0.6785610644848316,
      "train_runtime": 13753.5654,
      "train_samples_per_second": 57.386,
      "train_steps_per_second": 0.224
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 3080,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 2.645806671380611e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}