{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.38801261829653,
  "eval_steps": 500,
  "global_step": 2200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 8.403361344537815e-08,
      "loss": 1.8691,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.680672268907563e-07,
      "loss": 1.9322,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.5210084033613445e-07,
      "loss": 1.813,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.361344537815126e-07,
      "loss": 1.8232,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.201680672268908e-07,
      "loss": 1.859,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.042016806722689e-07,
      "loss": 1.8849,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.882352941176471e-07,
      "loss": 1.8436,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.722689075630252e-07,
      "loss": 1.9496,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.563025210084034e-07,
      "loss": 1.8612,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.403361344537816e-07,
      "loss": 1.8148,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.243697478991598e-07,
      "loss": 1.8985,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.0084033613445378e-06,
      "loss": 1.8926,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.092436974789916e-06,
      "loss": 1.9129,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1764705882352942e-06,
      "loss": 1.846,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2605042016806724e-06,
      "loss": 1.8438,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3445378151260504e-06,
      "loss": 1.851,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4285714285714286e-06,
      "loss": 1.7987,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5126050420168068e-06,
      "loss": 1.7278,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5966386554621848e-06,
      "loss": 1.9451,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6806722689075632e-06,
      "loss": 1.9044,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7647058823529414e-06,
      "loss": 1.874,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8487394957983196e-06,
      "loss": 1.8033,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.932773109243698e-06,
      "loss": 1.8778,
      "step": 23
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.0168067226890756e-06,
      "loss": 1.9901,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.100840336134454e-06,
      "loss": 1.8532,
      "step": 25
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.184873949579832e-06,
      "loss": 1.8801,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.2689075630252102e-06,
      "loss": 1.7896,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.3529411764705885e-06,
      "loss": 1.7693,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.4369747899159667e-06,
      "loss": 1.7693,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.521008403361345e-06,
      "loss": 1.8112,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.605042016806723e-06,
      "loss": 1.8509,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.689075630252101e-06,
      "loss": 1.7448,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.7731092436974795e-06,
      "loss": 1.8156,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.8571428571428573e-06,
      "loss": 1.9581,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.9411764705882355e-06,
      "loss": 1.711,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.0252100840336137e-06,
      "loss": 1.6952,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.109243697478992e-06,
      "loss": 1.8371,
      "step": 37
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.1932773109243696e-06,
      "loss": 1.9456,
      "step": 38
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.2773109243697483e-06,
      "loss": 1.7659,
      "step": 39
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.3613445378151265e-06,
      "loss": 1.6756,
      "step": 40
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.4453781512605043e-06,
      "loss": 1.6377,
      "step": 41
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.529411764705883e-06,
      "loss": 1.6489,
      "step": 42
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.6134453781512607e-06,
      "loss": 1.6813,
      "step": 43
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.6974789915966393e-06,
      "loss": 1.5641,
      "step": 44
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.781512605042017e-06,
      "loss": 1.631,
      "step": 45
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.865546218487396e-06,
      "loss": 1.5126,
      "step": 46
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.9495798319327735e-06,
      "loss": 1.6204,
      "step": 47
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.033613445378151e-06,
      "loss": 1.5342,
      "step": 48
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.11764705882353e-06,
      "loss": 1.5312,
      "step": 49
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.201680672268908e-06,
      "loss": 1.5835,
      "step": 50
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.2857142857142855e-06,
      "loss": 1.54,
      "step": 51
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.369747899159664e-06,
      "loss": 1.53,
      "step": 52
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.453781512605043e-06,
      "loss": 1.584,
      "step": 53
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.5378151260504205e-06,
      "loss": 1.5461,
      "step": 54
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.621848739495799e-06,
      "loss": 1.5108,
      "step": 55
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.705882352941177e-06,
      "loss": 1.4494,
      "step": 56
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.7899159663865555e-06,
      "loss": 1.5987,
      "step": 57
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.873949579831933e-06,
      "loss": 1.3851,
      "step": 58
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.957983193277311e-06,
      "loss": 1.4372,
      "step": 59
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.04201680672269e-06,
      "loss": 1.4501,
      "step": 60
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.1260504201680675e-06,
      "loss": 1.3852,
      "step": 61
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.210084033613446e-06,
      "loss": 1.3874,
      "step": 62
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.294117647058824e-06,
      "loss": 1.4378,
      "step": 63
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.378151260504202e-06,
      "loss": 1.3585,
      "step": 64
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.4621848739495795e-06,
      "loss": 1.4164,
      "step": 65
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.546218487394959e-06,
      "loss": 1.3129,
      "step": 66
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.630252100840337e-06,
      "loss": 1.2726,
      "step": 67
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.7142857142857145e-06,
      "loss": 1.2172,
      "step": 68
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.798319327731093e-06,
      "loss": 1.2488,
      "step": 69
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.882352941176471e-06,
      "loss": 1.2259,
      "step": 70
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.9663865546218495e-06,
      "loss": 1.1544,
      "step": 71
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.050420168067227e-06,
      "loss": 1.3055,
      "step": 72
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.134453781512606e-06,
      "loss": 1.257,
      "step": 73
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.218487394957984e-06,
      "loss": 1.2906,
      "step": 74
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.3025210084033615e-06,
      "loss": 1.2083,
      "step": 75
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.386554621848739e-06,
      "loss": 1.1757,
      "step": 76
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.470588235294119e-06,
      "loss": 1.2551,
      "step": 77
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.5546218487394966e-06,
      "loss": 1.0946,
      "step": 78
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.638655462184874e-06,
      "loss": 1.1347,
      "step": 79
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.722689075630253e-06,
      "loss": 1.0241,
      "step": 80
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.806722689075631e-06,
      "loss": 1.035,
      "step": 81
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.8907563025210085e-06,
      "loss": 1.1431,
      "step": 82
    },
    {
      "epoch": 0.05,
      "learning_rate": 6.974789915966387e-06,
      "loss": 1.0455,
      "step": 83
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.058823529411766e-06,
      "loss": 1.0737,
      "step": 84
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.1428571428571436e-06,
      "loss": 1.1082,
      "step": 85
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.226890756302521e-06,
      "loss": 1.0571,
      "step": 86
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.310924369747899e-06,
      "loss": 0.8394,
      "step": 87
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.394957983193279e-06,
      "loss": 1.0506,
      "step": 88
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.478991596638656e-06,
      "loss": 0.9712,
      "step": 89
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.563025210084034e-06,
      "loss": 1.027,
      "step": 90
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.647058823529411e-06,
      "loss": 0.9364,
      "step": 91
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.731092436974791e-06,
      "loss": 0.8227,
      "step": 92
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.815126050420168e-06,
      "loss": 0.8049,
      "step": 93
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.899159663865547e-06,
      "loss": 0.8572,
      "step": 94
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.983193277310926e-06,
      "loss": 0.8572,
      "step": 95
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.067226890756303e-06,
      "loss": 0.7954,
      "step": 96
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.151260504201681e-06,
      "loss": 0.8152,
      "step": 97
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.23529411764706e-06,
      "loss": 0.798,
      "step": 98
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.319327731092438e-06,
      "loss": 0.7755,
      "step": 99
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.403361344537815e-06,
      "loss": 0.6974,
      "step": 100
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.487394957983194e-06,
      "loss": 0.6356,
      "step": 101
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.571428571428571e-06,
      "loss": 0.6874,
      "step": 102
    },
    {
      "epoch": 0.06,
      "learning_rate": 8.655462184873951e-06,
      "loss": 0.8484,
      "step": 103
    },
    {
      "epoch": 0.07,
      "learning_rate": 8.739495798319328e-06,
      "loss": 0.7481,
      "step": 104
    },
    {
      "epoch": 0.07,
      "learning_rate": 8.823529411764707e-06,
      "loss": 0.716,
      "step": 105
    },
    {
      "epoch": 0.07,
      "learning_rate": 8.907563025210085e-06,
      "loss": 0.748,
      "step": 106
    },
    {
      "epoch": 0.07,
      "learning_rate": 8.991596638655462e-06,
      "loss": 0.7601,
      "step": 107
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.075630252100841e-06,
      "loss": 0.7004,
      "step": 108
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.15966386554622e-06,
      "loss": 0.5679,
      "step": 109
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.243697478991598e-06,
      "loss": 0.6666,
      "step": 110
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.327731092436975e-06,
      "loss": 0.5822,
      "step": 111
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.411764705882354e-06,
      "loss": 0.6817,
      "step": 112
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.49579831932773e-06,
      "loss": 0.5127,
      "step": 113
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.579831932773111e-06,
      "loss": 0.5823,
      "step": 114
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.663865546218488e-06,
      "loss": 0.6442,
      "step": 115
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.747899159663867e-06,
      "loss": 0.5335,
      "step": 116
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.831932773109244e-06,
      "loss": 0.5792,
      "step": 117
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.915966386554622e-06,
      "loss": 0.5482,
      "step": 118
    },
    {
      "epoch": 0.08,
      "learning_rate": 1e-05,
      "loss": 0.6763,
      "step": 119
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.008403361344538e-05,
      "loss": 0.6362,
      "step": 120
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0168067226890756e-05,
      "loss": 0.5591,
      "step": 121
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0252100840336135e-05,
      "loss": 0.5528,
      "step": 122
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0336134453781514e-05,
      "loss": 0.5025,
      "step": 123
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0420168067226892e-05,
      "loss": 0.5254,
      "step": 124
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0504201680672271e-05,
      "loss": 0.5157,
      "step": 125
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0588235294117648e-05,
      "loss": 0.5261,
      "step": 126
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0672268907563026e-05,
      "loss": 0.493,
      "step": 127
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0756302521008403e-05,
      "loss": 0.4658,
      "step": 128
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0840336134453782e-05,
      "loss": 0.6546,
      "step": 129
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.0924369747899159e-05,
      "loss": 0.4444,
      "step": 130
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.100840336134454e-05,
      "loss": 0.4153,
      "step": 131
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.1092436974789918e-05,
      "loss": 0.4911,
      "step": 132
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.1176470588235295e-05,
      "loss": 0.5026,
      "step": 133
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.1260504201680673e-05,
      "loss": 0.5123,
      "step": 134
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.134453781512605e-05,
      "loss": 0.4177,
      "step": 135
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.1428571428571429e-05,
      "loss": 0.4184,
      "step": 136
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.1512605042016806e-05,
      "loss": 0.4254,
      "step": 137
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.1596638655462186e-05,
      "loss": 0.4296,
      "step": 138
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.1680672268907565e-05,
      "loss": 0.5156,
      "step": 139
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.1764705882352942e-05,
      "loss": 0.3739,
      "step": 140
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.184873949579832e-05,
      "loss": 0.3774,
      "step": 141
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.1932773109243699e-05,
      "loss": 0.4023,
      "step": 142
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2016806722689076e-05,
      "loss": 0.4624,
      "step": 143
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2100840336134455e-05,
      "loss": 0.5193,
      "step": 144
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2184873949579832e-05,
      "loss": 0.3901,
      "step": 145
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2268907563025212e-05,
      "loss": 0.3578,
      "step": 146
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.235294117647059e-05,
      "loss": 0.4057,
      "step": 147
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2436974789915967e-05,
      "loss": 0.4753,
      "step": 148
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2521008403361346e-05,
      "loss": 0.3559,
      "step": 149
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.2605042016806723e-05,
      "loss": 0.4103,
      "step": 150
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.2689075630252102e-05,
      "loss": 0.4715,
      "step": 151
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.2773109243697479e-05,
      "loss": 0.3896,
      "step": 152
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.2857142857142859e-05,
      "loss": 0.381,
      "step": 153
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.2941176470588238e-05,
      "loss": 0.3592,
      "step": 154
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3025210084033614e-05,
      "loss": 0.3586,
      "step": 155
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3109243697478993e-05,
      "loss": 0.351,
      "step": 156
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.319327731092437e-05,
      "loss": 0.4891,
      "step": 157
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3277310924369749e-05,
      "loss": 0.4245,
      "step": 158
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3361344537815126e-05,
      "loss": 0.3132,
      "step": 159
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3445378151260506e-05,
      "loss": 0.353,
      "step": 160
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3529411764705885e-05,
      "loss": 0.3201,
      "step": 161
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3613445378151261e-05,
      "loss": 0.365,
      "step": 162
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.369747899159664e-05,
      "loss": 0.4156,
      "step": 163
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3781512605042017e-05,
      "loss": 0.3653,
      "step": 164
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3865546218487396e-05,
      "loss": 0.2617,
      "step": 165
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.3949579831932774e-05,
      "loss": 0.2817,
      "step": 166
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4033613445378151e-05,
      "loss": 0.334,
      "step": 167
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4117647058823532e-05,
      "loss": 0.3369,
      "step": 168
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4201680672268908e-05,
      "loss": 0.317,
      "step": 169
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4285714285714287e-05,
      "loss": 0.3698,
      "step": 170
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4369747899159666e-05,
      "loss": 0.309,
      "step": 171
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4453781512605043e-05,
      "loss": 0.3718,
      "step": 172
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4537815126050421e-05,
      "loss": 0.3409,
      "step": 173
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4621848739495798e-05,
      "loss": 0.3431,
      "step": 174
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4705882352941179e-05,
      "loss": 0.2731,
      "step": 175
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4789915966386557e-05,
      "loss": 0.3037,
      "step": 176
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4873949579831934e-05,
      "loss": 0.2258,
      "step": 177
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.4957983193277313e-05,
      "loss": 0.3793,
      "step": 178
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.504201680672269e-05,
      "loss": 0.2701,
      "step": 179
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.5126050420168068e-05,
      "loss": 0.4068,
      "step": 180
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.5210084033613445e-05,
      "loss": 0.3137,
      "step": 181
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.5294117647058822e-05,
      "loss": 0.2372,
      "step": 182
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5378151260504204e-05,
      "loss": 0.3499,
      "step": 183
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5462184873949583e-05,
      "loss": 0.3794,
      "step": 184
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5546218487394958e-05,
      "loss": 0.3609,
      "step": 185
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5630252100840337e-05,
      "loss": 0.2628,
      "step": 186
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5714285714285715e-05,
      "loss": 0.2046,
      "step": 187
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5798319327731094e-05,
      "loss": 0.3348,
      "step": 188
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.5882352941176473e-05,
      "loss": 0.2149,
      "step": 189
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.596638655462185e-05,
      "loss": 0.3018,
      "step": 190
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.605042016806723e-05,
      "loss": 0.2865,
      "step": 191
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.6134453781512605e-05,
      "loss": 0.3261,
      "step": 192
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.6218487394957984e-05,
      "loss": 0.2435,
      "step": 193
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.6302521008403362e-05,
      "loss": 0.2983,
      "step": 194
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.638655462184874e-05,
      "loss": 0.3006,
      "step": 195
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.647058823529412e-05,
      "loss": 0.3597,
      "step": 196
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.6554621848739495e-05,
      "loss": 0.2604,
      "step": 197
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.6638655462184877e-05,
      "loss": 0.3334,
      "step": 198
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.6722689075630255e-05,
      "loss": 0.2514,
      "step": 199
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.680672268907563e-05,
      "loss": 0.2765,
      "step": 200
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.689075630252101e-05,
      "loss": 0.2906,
      "step": 201
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.6974789915966388e-05,
      "loss": 0.3289,
      "step": 202
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7058823529411767e-05,
      "loss": 0.3029,
      "step": 203
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7142857142857142e-05,
      "loss": 0.239,
      "step": 204
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7226890756302524e-05,
      "loss": 0.232,
      "step": 205
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7310924369747902e-05,
      "loss": 0.3154,
      "step": 206
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7394957983193278e-05,
      "loss": 0.3122,
      "step": 207
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7478991596638656e-05,
      "loss": 0.3832,
      "step": 208
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7563025210084035e-05,
      "loss": 0.3167,
      "step": 209
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7647058823529414e-05,
      "loss": 0.2777,
      "step": 210
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.7731092436974792e-05,
      "loss": 0.3369,
      "step": 211
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.781512605042017e-05,
      "loss": 0.2962,
      "step": 212
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.789915966386555e-05,
      "loss": 0.4066,
      "step": 213
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.7983193277310925e-05,
      "loss": 0.1592,
      "step": 214
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8067226890756303e-05,
      "loss": 0.3783,
      "step": 215
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8151260504201682e-05,
      "loss": 0.29,
      "step": 216
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.823529411764706e-05,
      "loss": 0.2008,
      "step": 217
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.831932773109244e-05,
      "loss": 0.2433,
      "step": 218
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8403361344537814e-05,
      "loss": 0.2189,
      "step": 219
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8487394957983196e-05,
      "loss": 0.3552,
      "step": 220
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8571428571428575e-05,
      "loss": 0.281,
      "step": 221
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.865546218487395e-05,
      "loss": 0.2944,
      "step": 222
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.873949579831933e-05,
      "loss": 0.248,
      "step": 223
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8823529411764708e-05,
      "loss": 0.2536,
      "step": 224
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.8907563025210086e-05,
      "loss": 0.1867,
      "step": 225
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.899159663865546e-05,
      "loss": 0.2767,
      "step": 226
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9075630252100844e-05,
      "loss": 0.2358,
      "step": 227
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9159663865546222e-05,
      "loss": 0.303,
      "step": 228
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9243697478991597e-05,
      "loss": 0.2326,
      "step": 229
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9327731092436976e-05,
      "loss": 0.2603,
      "step": 230
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9411764705882355e-05,
      "loss": 0.2697,
      "step": 231
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9495798319327733e-05,
      "loss": 0.3522,
      "step": 232
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.957983193277311e-05,
      "loss": 0.2329,
      "step": 233
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9663865546218487e-05,
      "loss": 0.2916,
      "step": 234
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.974789915966387e-05,
      "loss": 0.3536,
      "step": 235
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9831932773109244e-05,
      "loss": 0.2588,
      "step": 236
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9915966386554623e-05,
      "loss": 0.2458,
      "step": 237
    },
    {
      "epoch": 0.15,
      "learning_rate": 2e-05,
      "loss": 0.3283,
      "step": 238
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9999999164866356e-05,
      "loss": 0.3101,
      "step": 239
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9999996659465552e-05,
      "loss": 0.2086,
      "step": 240
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9999992483798012e-05,
      "loss": 0.3037,
      "step": 241
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.999998663786444e-05,
      "loss": 0.2908,
      "step": 242
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9999979121665797e-05,
      "loss": 0.2185,
      "step": 243
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.999996993520335e-05,
      "loss": 0.3309,
      "step": 244
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9999959078478635e-05,
      "loss": 0.2707,
      "step": 245
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999946551493456e-05,
      "loss": 0.2495,
      "step": 246
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999932354249913e-05,
      "loss": 0.2479,
      "step": 247
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.999991648675037e-05,
      "loss": 0.3172,
      "step": 248
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999898948997485e-05,
      "loss": 0.3356,
      "step": 249
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.999987974099418e-05,
      "loss": 0.2475,
      "step": 250
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.999985886274367e-05,
      "loss": 0.3237,
      "step": 251
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999836314249443e-05,
      "loss": 0.2894,
      "step": 252
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999812095515258e-05,
      "loss": 0.253,
      "step": 253
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999786206545165e-05,
      "loss": 0.2573,
      "step": 254
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999758647343486e-05,
      "loss": 0.3214,
      "step": 255
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999729417914825e-05,
      "loss": 0.1844,
      "step": 256
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999698518264067e-05,
      "loss": 0.2827,
      "step": 257
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999665948396366e-05,
      "loss": 0.291,
      "step": 258
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.999963170831717e-05,
      "loss": 0.2272,
      "step": 259
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.999959579803219e-05,
      "loss": 0.2949,
      "step": 260
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9999558217547432e-05,
      "loss": 0.2524,
      "step": 261
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999518966869166e-05,
      "loss": 0.232,
      "step": 262
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999478046003956e-05,
      "loss": 0.2235,
      "step": 263
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999435454958627e-05,
      "loss": 0.3296,
      "step": 264
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999391193740303e-05,
      "loss": 0.2595,
      "step": 265
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.999934526235637e-05,
      "loss": 0.2743,
      "step": 266
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999297660814497e-05,
      "loss": 0.2552,
      "step": 267
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999248389122644e-05,
      "loss": 0.2187,
      "step": 268
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999197447289034e-05,
      "loss": 0.2743,
      "step": 269
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999144835322178e-05,
      "loss": 0.2933,
      "step": 270
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999090553230864e-05,
      "loss": 0.2864,
      "step": 271
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9999034601024156e-05,
      "loss": 0.2986,
      "step": 272
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9998976978711406e-05,
      "loss": 0.3612,
      "step": 273
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9998917686302227e-05,
      "loss": 0.2684,
      "step": 274
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9998856723806535e-05,
      "loss": 0.2701,
      "step": 275
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.99987940912345e-05,
      "loss": 0.2646,
      "step": 276
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9998729788596596e-05,
      "loss": 0.2868,
      "step": 277
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998663815903556e-05,
      "loss": 0.2736,
      "step": 278
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998596173166397e-05,
      "loss": 0.271,
      "step": 279
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.999852686039642e-05,
      "loss": 0.2924,
      "step": 280
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998455877605204e-05,
      "loss": 0.2682,
      "step": 281
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998383224804606e-05,
      "loss": 0.1784,
      "step": 282
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998308902006754e-05,
      "loss": 0.2137,
      "step": 283
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.999823290922407e-05,
      "loss": 0.2719,
      "step": 284
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998155246469242e-05,
      "loss": 0.1992,
      "step": 285
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9998075913755243e-05,
      "loss": 0.2519,
      "step": 286
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.999799491109532e-05,
      "loss": 0.2143,
      "step": 287
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.999791223850301e-05,
      "loss": 0.2601,
      "step": 288
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.999782789599212e-05,
      "loss": 0.2774,
      "step": 289
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9997741883576732e-05,
      "loss": 0.2742,
      "step": 290
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9997654201271218e-05,
      "loss": 0.213,
      "step": 291
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9997564849090215e-05,
      "loss": 0.1821,
      "step": 292
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9997473827048657e-05,
      "loss": 0.1946,
      "step": 293
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9997381135161746e-05,
      "loss": 0.2356,
      "step": 294
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999728677344496e-05,
      "loss": 0.2035,
      "step": 295
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999719074191406e-05,
      "loss": 0.1997,
      "step": 296
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9997093040585085e-05,
      "loss": 0.2647,
      "step": 297
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999699366947436e-05,
      "loss": 0.2234,
      "step": 298
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9996892628598474e-05,
      "loss": 0.1876,
      "step": 299
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999678991797431e-05,
      "loss": 0.2057,
      "step": 300
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999668553761902e-05,
      "loss": 0.2225,
      "step": 301
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999657948755004e-05,
      "loss": 0.2162,
      "step": 302
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9996471767785085e-05,
      "loss": 0.2427,
      "step": 303
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.999636237834214e-05,
      "loss": 0.1992,
      "step": 304
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9996251319239484e-05,
      "loss": 0.2033,
      "step": 305
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9996138590495664e-05,
      "loss": 0.2075,
      "step": 306
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9996024192129507e-05,
      "loss": 0.2331,
      "step": 307
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9995908124160124e-05,
      "loss": 0.273,
      "step": 308
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9995790386606896e-05,
      "loss": 0.2314,
      "step": 309
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9995670979489497e-05,
      "loss": 0.1937,
      "step": 310
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999554990282786e-05,
      "loss": 0.2338,
      "step": 311
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999542715664222e-05,
      "loss": 0.2182,
      "step": 312
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999530274095307e-05,
      "loss": 0.2046,
      "step": 313
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999517665578119e-05,
      "loss": 0.2303,
      "step": 314
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9995048901147648e-05,
      "loss": 0.2754,
      "step": 315
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9994919477073777e-05,
      "loss": 0.309,
      "step": 316
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9994788383581195e-05,
      "loss": 0.2326,
      "step": 317
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9994655620691795e-05,
      "loss": 0.2177,
      "step": 318
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999452118842776e-05,
      "loss": 0.2025,
      "step": 319
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9994385086811534e-05,
      "loss": 0.2459,
      "step": 320
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9994247315865855e-05,
      "loss": 0.2269,
      "step": 321
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9994107875613738e-05,
      "loss": 0.2144,
      "step": 322
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9993966766078463e-05,
      "loss": 0.2238,
      "step": 323
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.999382398728361e-05,
      "loss": 0.2026,
      "step": 324
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999367953925302e-05,
      "loss": 0.2112,
      "step": 325
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999353342201082e-05,
      "loss": 0.2432,
      "step": 326
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999338563558142e-05,
      "loss": 0.1898,
      "step": 327
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.99932361799895e-05,
      "loss": 0.2556,
      "step": 328
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9993085055260025e-05,
      "loss": 0.2106,
      "step": 329
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9992932261418237e-05,
      "loss": 0.2467,
      "step": 330
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9992777798489655e-05,
      "loss": 0.1955,
      "step": 331
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9992621666500082e-05,
      "loss": 0.2487,
      "step": 332
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999246386547559e-05,
      "loss": 0.2216,
      "step": 333
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999230439544254e-05,
      "loss": 0.1871,
      "step": 334
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999214325642757e-05,
      "loss": 0.2303,
      "step": 335
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9991980448457592e-05,
      "loss": 0.2297,
      "step": 336
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9991815971559796e-05,
      "loss": 0.2596,
      "step": 337
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9991649825761657e-05,
      "loss": 0.241,
      "step": 338
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999148201109093e-05,
      "loss": 0.2109,
      "step": 339
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.999131252757564e-05,
      "loss": 0.2615,
      "step": 340
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9991141375244094e-05,
      "loss": 0.264,
      "step": 341
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999096855412488e-05,
      "loss": 0.2652,
      "step": 342
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999079406424687e-05,
      "loss": 0.2983,
      "step": 343
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9990617905639195e-05,
      "loss": 0.2276,
      "step": 344
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9990440078331293e-05,
      "loss": 0.2743,
      "step": 345
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9990260582352856e-05,
      "loss": 0.2387,
      "step": 346
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.999007941773387e-05,
      "loss": 0.2179,
      "step": 347
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.998989658450459e-05,
      "loss": 0.2198,
      "step": 348
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9989712082695554e-05,
      "loss": 0.2572,
      "step": 349
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9989525912337586e-05,
      "loss": 0.2682,
      "step": 350
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9989338073461778e-05,
      "loss": 0.2348,
      "step": 351
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9989148566099498e-05,
      "loss": 0.2319,
      "step": 352
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9988957390282404e-05,
      "loss": 0.2801,
      "step": 353
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9988764546042424e-05,
      "loss": 0.2307,
      "step": 354
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.998857003341178e-05,
      "loss": 0.2463,
      "step": 355
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9988373852422943e-05,
      "loss": 0.245,
      "step": 356
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998817600310869e-05,
      "loss": 0.2235,
      "step": 357
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998797648550207e-05,
      "loss": 0.2133,
      "step": 358
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9987775299636404e-05,
      "loss": 0.1974,
      "step": 359
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998757244554529e-05,
      "loss": 0.2908,
      "step": 360
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998736792326262e-05,
      "loss": 0.2041,
      "step": 361
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9987161732822552e-05,
      "loss": 0.2717,
      "step": 362
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9986953874259522e-05,
      "loss": 0.189,
      "step": 363
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9986744347608247e-05,
      "loss": 0.1911,
      "step": 364
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9986533152903727e-05,
      "loss": 0.293,
      "step": 365
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998632029018124e-05,
      "loss": 0.2478,
      "step": 366
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998610575947633e-05,
      "loss": 0.2177,
      "step": 367
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998588956082484e-05,
      "loss": 0.1944,
      "step": 368
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9985671694262872e-05,
      "loss": 0.213,
      "step": 369
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9985452159826822e-05,
      "loss": 0.2294,
      "step": 370
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9985230957553357e-05,
      "loss": 0.2054,
      "step": 371
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.998500808747942e-05,
      "loss": 0.23,
      "step": 372
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9984783549642243e-05,
      "loss": 0.2015,
      "step": 373
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9984557344079322e-05,
      "loss": 0.2484,
      "step": 374
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9984329470828445e-05,
      "loss": 0.2999,
      "step": 375
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9984099929927672e-05,
      "loss": 0.1619,
      "step": 376
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9983868721415338e-05,
      "loss": 0.2541,
      "step": 377
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.998363584533007e-05,
      "loss": 0.2051,
      "step": 378
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9983401301710756e-05,
      "loss": 0.2142,
      "step": 379
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9983165090596574e-05,
      "loss": 0.2263,
      "step": 380
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9982927212026975e-05,
      "loss": 0.2824,
      "step": 381
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9982687666041697e-05,
      "loss": 0.2174,
      "step": 382
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9982446452680742e-05,
      "loss": 0.2472,
      "step": 383
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.998220357198441e-05,
      "loss": 0.1873,
      "step": 384
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9981959023993266e-05,
      "loss": 0.2256,
      "step": 385
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9981712808748147e-05,
      "loss": 0.2497,
      "step": 386
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9981464926290183e-05,
      "loss": 0.2448,
      "step": 387
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.998121537666078e-05,
      "loss": 0.2839,
      "step": 388
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9980964159901618e-05,
      "loss": 0.1869,
      "step": 389
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9980711276054655e-05,
      "loss": 0.2087,
      "step": 390
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.998045672516213e-05,
      "loss": 0.2234,
      "step": 391
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9980200507266557e-05,
      "loss": 0.2339,
      "step": 392
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.997994262241074e-05,
      "loss": 0.2468,
      "step": 393
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9979683070637745e-05,
      "loss": 0.2119,
      "step": 394
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9979421851990924e-05,
      "loss": 0.192,
      "step": 395
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.997915896651391e-05,
      "loss": 0.2351,
      "step": 396
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9978894414250614e-05,
      "loss": 0.164,
      "step": 397
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9978628195245216e-05,
      "loss": 0.1831,
      "step": 398
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.997836030954219e-05,
      "loss": 0.2811,
      "step": 399
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9978090757186276e-05,
      "loss": 0.1764,
      "step": 400
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9977819538222498e-05,
      "loss": 0.1726,
      "step": 401
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9977546652696154e-05,
      "loss": 0.2271,
      "step": 402
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9977272100652826e-05,
      "loss": 0.1639,
      "step": 403
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.997699588213837e-05,
      "loss": 0.2023,
      "step": 404
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9976717997198924e-05,
      "loss": 0.214,
      "step": 405
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9976438445880896e-05,
      "loss": 0.2276,
      "step": 406
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9976157228230987e-05,
      "loss": 0.213,
      "step": 407
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.997587434429616e-05,
      "loss": 0.2566,
      "step": 408
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9975589794123673e-05,
      "loss": 0.1884,
      "step": 409
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9975303577761048e-05,
      "loss": 0.1928,
      "step": 410
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9975015695256087e-05,
      "loss": 0.1814,
      "step": 411
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9974726146656885e-05,
      "loss": 0.2004,
      "step": 412
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.997443493201179e-05,
      "loss": 0.1703,
      "step": 413
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.997414205136946e-05,
      "loss": 0.1737,
      "step": 414
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.99738475047788e-05,
      "loss": 0.1935,
      "step": 415
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9973551292289007e-05,
      "loss": 0.228,
      "step": 416
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9973253413949567e-05,
      "loss": 0.2645,
      "step": 417
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9972953869810227e-05,
      "loss": 0.1615,
      "step": 418
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9972652659921016e-05,
      "loss": 0.1586,
      "step": 419
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.997234978433225e-05,
      "loss": 0.1972,
      "step": 420
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9972045243094513e-05,
      "loss": 0.1859,
      "step": 421
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9971739036258676e-05,
      "loss": 0.248,
      "step": 422
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.997143116387588e-05,
      "loss": 0.217,
      "step": 423
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9971121625997548e-05,
      "loss": 0.2336,
      "step": 424
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9970810422675383e-05,
      "loss": 0.2306,
      "step": 425
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9970497553961363e-05,
      "loss": 0.2096,
      "step": 426
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9970183019907745e-05,
      "loss": 0.2048,
      "step": 427
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9969866820567068e-05,
      "loss": 0.2642,
      "step": 428
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9969548955992142e-05,
      "loss": 0.2681,
      "step": 429
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.996922942623606e-05,
      "loss": 0.2047,
      "step": 430
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9968908231352192e-05,
      "loss": 0.2453,
      "step": 431
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9968585371394186e-05,
      "loss": 0.268,
      "step": 432
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.996826084641597e-05,
      "loss": 0.1967,
      "step": 433
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9967934656471743e-05,
      "loss": 0.2844,
      "step": 434
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9967606801615994e-05,
      "loss": 0.2005,
      "step": 435
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.996727728190348e-05,
      "loss": 0.2014,
      "step": 436
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9966946097389242e-05,
      "loss": 0.2236,
      "step": 437
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9966613248128593e-05,
      "loss": 0.2003,
      "step": 438
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9966278734177132e-05,
      "loss": 0.1643,
      "step": 439
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9965942555590726e-05,
      "loss": 0.2513,
      "step": 440
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9965604712425532e-05,
      "loss": 0.2197,
      "step": 441
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9965265204737975e-05,
      "loss": 0.205,
      "step": 442
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9964924032584766e-05,
      "loss": 0.2449,
      "step": 443
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9964581196022882e-05,
      "loss": 0.1878,
      "step": 444
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9964236695109595e-05,
      "loss": 0.1782,
      "step": 445
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9963890529902443e-05,
      "loss": 0.2048,
      "step": 446
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9963542700459237e-05,
      "loss": 0.2239,
      "step": 447
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9963193206838083e-05,
      "loss": 0.2226,
      "step": 448
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9962842049097356e-05,
      "loss": 0.2556,
      "step": 449
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.99624892272957e-05,
      "loss": 0.2024,
      "step": 450
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.996213474149206e-05,
      "loss": 0.2272,
      "step": 451
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.996177859174563e-05,
      "loss": 0.1739,
      "step": 452
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.99614207781159e-05,
      "loss": 0.2492,
      "step": 453
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9961061300662643e-05,
      "loss": 0.2294,
      "step": 454
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9960700159445892e-05,
      "loss": 0.2012,
      "step": 455
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9960337354525973e-05,
      "loss": 0.1727,
      "step": 456
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9959972885963478e-05,
      "loss": 0.2069,
      "step": 457
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.995960675381929e-05,
      "loss": 0.1356,
      "step": 458
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.995923895815456e-05,
      "loss": 0.1814,
      "step": 459
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9958869499030716e-05,
      "loss": 0.2131,
      "step": 460
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9958498376509476e-05,
      "loss": 0.1738,
      "step": 461
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9958125590652818e-05,
      "loss": 0.1689,
      "step": 462
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9957751141523013e-05,
      "loss": 0.2174,
      "step": 463
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.99573750291826e-05,
      "loss": 0.1991,
      "step": 464
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9956997253694405e-05,
      "loss": 0.2903,
      "step": 465
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9956617815121524e-05,
      "loss": 0.192,
      "step": 466
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9956236713527333e-05,
      "loss": 0.2093,
      "step": 467
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9955853948975485e-05,
      "loss": 0.2157,
      "step": 468
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9955469521529913e-05,
      "loss": 0.2189,
      "step": 469
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.995508343125483e-05,
      "loss": 0.2381,
      "step": 470
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9954695678214717e-05,
      "loss": 0.23,
      "step": 471
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9954306262474344e-05,
      "loss": 0.209,
      "step": 472
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9953915184098752e-05,
      "loss": 0.2174,
      "step": 473
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.995352244315326e-05,
      "loss": 0.217,
      "step": 474
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.995312803970347e-05,
      "loss": 0.2016,
      "step": 475
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9952731973815254e-05,
      "loss": 0.1971,
      "step": 476
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9952334245554775e-05,
      "loss": 0.2632,
      "step": 477
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9951934854988448e-05,
      "loss": 0.1839,
      "step": 478
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9951533802182997e-05,
      "loss": 0.2066,
      "step": 479
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9951131087205397e-05,
      "loss": 0.1943,
      "step": 480
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9950726710122922e-05,
      "loss": 0.1753,
      "step": 481
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9950320671003105e-05,
      "loss": 0.1703,
      "step": 482
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.994991296991377e-05,
      "loss": 0.2429,
      "step": 483
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9949503606923018e-05,
      "loss": 0.1767,
      "step": 484
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994909258209922e-05,
      "loss": 0.1826,
      "step": 485
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994867989551102e-05,
      "loss": 0.2219,
      "step": 486
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994826554722736e-05,
      "loss": 0.2554,
      "step": 487
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994784953731744e-05,
      "loss": 0.1696,
      "step": 488
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994743186585075e-05,
      "loss": 0.2252,
      "step": 489
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9947012532897047e-05,
      "loss": 0.176,
      "step": 490
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9946591538526375e-05,
      "loss": 0.1934,
      "step": 491
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994616888280905e-05,
      "loss": 0.1938,
      "step": 492
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9945744565815662e-05,
      "loss": 0.2648,
      "step": 493
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994531858761709e-05,
      "loss": 0.2259,
      "step": 494
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.994489094828448e-05,
      "loss": 0.2151,
      "step": 495
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9944461647889262e-05,
      "loss": 0.2489,
      "step": 496
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9944030686503137e-05,
      "loss": 0.2277,
      "step": 497
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9943598064198092e-05,
      "loss": 0.21,
      "step": 498
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9943163781046378e-05,
      "loss": 0.2033,
      "step": 499
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9942727837120546e-05,
      "loss": 0.196,
      "step": 500
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9942290232493396e-05,
      "loss": 0.18,
      "step": 501
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9941850967238028e-05,
      "loss": 0.1894,
      "step": 502
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9941410041427808e-05,
      "loss": 0.2537,
      "step": 503
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.994096745513638e-05,
      "loss": 0.1832,
      "step": 504
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9940523208437677e-05,
      "loss": 0.1725,
      "step": 505
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.994007730140589e-05,
      "loss": 0.2549,
      "step": 506
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.99396297341155e-05,
      "loss": 0.1949,
      "step": 507
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9939180506641264e-05,
      "loss": 0.2313,
      "step": 508
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9938729619058216e-05,
      "loss": 0.1614,
      "step": 509
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9938277071441665e-05,
      "loss": 0.1762,
      "step": 510
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9937822863867196e-05,
      "loss": 0.164,
      "step": 511
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.993736699641068e-05,
      "loss": 0.2342,
      "step": 512
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9936909469148256e-05,
      "loss": 0.1835,
      "step": 513
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9936450282156342e-05,
      "loss": 0.1627,
      "step": 514
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9935989435511636e-05,
      "loss": 0.1771,
      "step": 515
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.993552692929111e-05,
      "loss": 0.1287,
      "step": 516
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9935062763572014e-05,
      "loss": 0.1594,
      "step": 517
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.993459693843188e-05,
      "loss": 0.1937,
      "step": 518
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9934129453948513e-05,
      "loss": 0.1897,
      "step": 519
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9933660310199992e-05,
      "loss": 0.1834,
      "step": 520
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9933189507264675e-05,
      "loss": 0.2146,
      "step": 521
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9932717045221205e-05,
      "loss": 0.2373,
      "step": 522
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9932242924148494e-05,
      "loss": 0.222,
      "step": 523
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.993176714412573e-05,
      "loss": 0.216,
      "step": 524
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.993128970523238e-05,
      "loss": 0.2331,
      "step": 525
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9930810607548198e-05,
      "loss": 0.1974,
      "step": 526
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.993032985115319e-05,
      "loss": 0.2022,
      "step": 527
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9929847436127675e-05,
      "loss": 0.2075,
      "step": 528
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9929363362552214e-05,
      "loss": 0.1789,
      "step": 529
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9928877630507665e-05,
      "loss": 0.189,
      "step": 530
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9928390240075158e-05,
      "loss": 0.1659,
      "step": 531
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9927901191336105e-05,
      "loss": 0.1694,
      "step": 532
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.992741048437218e-05,
      "loss": 0.2523,
      "step": 533
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9926918119265356e-05,
      "loss": 0.1809,
      "step": 534
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.992642409609786e-05,
      "loss": 0.1828,
      "step": 535
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9925928414952217e-05,
      "loss": 0.2448,
      "step": 536
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9925431075911213e-05,
      "loss": 0.1889,
      "step": 537
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.992493207905792e-05,
      "loss": 0.2289,
      "step": 538
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9924431424475677e-05,
      "loss": 0.1591,
      "step": 539
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9923929112248115e-05,
      "loss": 0.1896,
      "step": 540
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9923425142459132e-05,
      "loss": 0.2248,
      "step": 541
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.99229195151929e-05,
      "loss": 0.2346,
      "step": 542
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9922412230533878e-05,
      "loss": 0.1724,
      "step": 543
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.992190328856679e-05,
      "loss": 0.1736,
      "step": 544
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9921392689376646e-05,
      "loss": 0.2036,
      "step": 545
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.992088043304873e-05,
      "loss": 0.1811,
      "step": 546
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9920366519668605e-05,
      "loss": 0.1857,
      "step": 547
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9919850949322105e-05,
      "loss": 0.1592,
      "step": 548
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.991933372209534e-05,
      "loss": 0.1841,
      "step": 549
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9918814838074708e-05,
      "loss": 0.1731,
      "step": 550
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9918294297346876e-05,
      "loss": 0.2073,
      "step": 551
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9917772099998783e-05,
      "loss": 0.1867,
      "step": 552
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.991724824611766e-05,
      "loss": 0.2107,
      "step": 553
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.991672273579099e-05,
      "loss": 0.2849,
      "step": 554
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9916195569106557e-05,
      "loss": 0.1677,
      "step": 555
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9915666746152408e-05,
      "loss": 0.2274,
      "step": 556
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9915136267016874e-05,
      "loss": 0.1484,
      "step": 557
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.991460413178856e-05,
      "loss": 0.2037,
      "step": 558
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9914070340556343e-05,
      "loss": 0.1997,
      "step": 559
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9913534893409378e-05,
      "loss": 0.1732,
      "step": 560
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.991299779043711e-05,
      "loss": 0.1498,
      "step": 561
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9912459031729235e-05,
      "loss": 0.1788,
      "step": 562
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.991191861737575e-05,
      "loss": 0.1917,
      "step": 563
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.991137654746692e-05,
      "loss": 0.2093,
      "step": 564
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9910832822093272e-05,
      "loss": 0.195,
      "step": 565
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.991028744134564e-05,
      "loss": 0.2534,
      "step": 566
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9909740405315107e-05,
      "loss": 0.1847,
      "step": 567
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9909191714093045e-05,
      "loss": 0.192,
      "step": 568
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9908641367771097e-05,
      "loss": 0.2046,
      "step": 569
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.990808936644119e-05,
      "loss": 0.1911,
      "step": 570
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.990753571019552e-05,
      "loss": 0.2428,
      "step": 571
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9906980399126566e-05,
      "loss": 0.2086,
      "step": 572
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.990642343332708e-05,
      "loss": 0.1407,
      "step": 573
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9905864812890083e-05,
      "loss": 0.1886,
      "step": 574
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9905304537908885e-05,
      "loss": 0.205,
      "step": 575
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9904742608477066e-05,
      "loss": 0.204,
      "step": 576
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9904179024688487e-05,
      "loss": 0.1868,
      "step": 577
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.990361378663727e-05,
      "loss": 0.1838,
      "step": 578
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9903046894417842e-05,
      "loss": 0.185,
      "step": 579
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9902478348124874e-05,
      "loss": 0.1779,
      "step": 580
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9901908147853337e-05,
      "loss": 0.2007,
      "step": 581
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9901336293698467e-05,
      "loss": 0.1596,
      "step": 582
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9900762785755775e-05,
      "loss": 0.1822,
      "step": 583
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9900187624121063e-05,
      "loss": 0.2252,
      "step": 584
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9899610808890385e-05,
      "loss": 0.1909,
      "step": 585
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9899032340160095e-05,
      "loss": 0.2158,
      "step": 586
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9898452218026807e-05,
      "loss": 0.1756,
      "step": 587
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.989787044258742e-05,
      "loss": 0.2225,
      "step": 588
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.98972870139391e-05,
      "loss": 0.2088,
      "step": 589
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9896701932179306e-05,
      "loss": 0.2134,
      "step": 590
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9896115197405753e-05,
      "loss": 0.1998,
      "step": 591
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9895526809716447e-05,
      "loss": 0.2058,
      "step": 592
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.989493676920966e-05,
      "loss": 0.1975,
      "step": 593
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9894345075983948e-05,
      "loss": 0.1668,
      "step": 594
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9893751730138137e-05,
      "loss": 0.2023,
      "step": 595
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9893156731771334e-05,
      "loss": 0.2187,
      "step": 596
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9892560080982916e-05,
      "loss": 0.1976,
      "step": 597
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9891961777872543e-05,
      "loss": 0.2534,
      "step": 598
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.989136182254015e-05,
      "loss": 0.1874,
      "step": 599
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9890760215085938e-05,
      "loss": 0.1488,
      "step": 600
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.98901569556104e-05,
      "loss": 0.1713,
      "step": 601
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9889552044214285e-05,
      "loss": 0.2588,
      "step": 602
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9888945480998642e-05,
      "loss": 0.181,
      "step": 603
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.988833726606478e-05,
      "loss": 0.2066,
      "step": 604
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9887727399514277e-05,
      "loss": 0.2188,
      "step": 605
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9887115881449013e-05,
      "loss": 0.186,
      "step": 606
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9886502711971117e-05,
      "loss": 0.1745,
      "step": 607
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9885887891183008e-05,
      "loss": 0.187,
      "step": 608
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9885271419187378e-05,
      "loss": 0.2058,
      "step": 609
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.988465329608719e-05,
      "loss": 0.2208,
      "step": 610
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9884033521985692e-05,
      "loss": 0.1679,
      "step": 611
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9883412096986404e-05,
      "loss": 0.208,
      "step": 612
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9882789021193115e-05,
      "loss": 0.2116,
      "step": 613
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.98821642947099e-05,
      "loss": 0.207,
      "step": 614
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.98815379176411e-05,
      "loss": 0.1639,
      "step": 615
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9880909890091345e-05,
      "loss": 0.2912,
      "step": 616
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9880280212165523e-05,
      "loss": 0.2101,
      "step": 617
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9879648883968815e-05,
      "loss": 0.2022,
      "step": 618
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9879015905606663e-05,
      "loss": 0.1992,
      "step": 619
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9878381277184794e-05,
      "loss": 0.186,
      "step": 620
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.987774499880921e-05,
      "loss": 0.2021,
      "step": 621
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9877107070586185e-05,
      "loss": 0.1499,
      "step": 622
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9876467492622268e-05,
      "loss": 0.1999,
      "step": 623
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9875826265024288e-05,
      "loss": 0.2019,
      "step": 624
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.987518338789935e-05,
      "loss": 0.2236,
      "step": 625
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9874538861354826e-05,
      "loss": 0.1576,
      "step": 626
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.987389268549837e-05,
      "loss": 0.2069,
      "step": 627
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9873244860437916e-05,
      "loss": 0.2029,
      "step": 628
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.987259538628166e-05,
      "loss": 0.2241,
      "step": 629
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.987194426313809e-05,
      "loss": 0.1767,
      "step": 630
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9871291491115953e-05,
      "loss": 0.2615,
      "step": 631
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9870637070324284e-05,
      "loss": 0.1945,
      "step": 632
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9869981000872386e-05,
      "loss": 0.1482,
      "step": 633
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9869323282869844e-05,
      "loss": 0.1761,
      "step": 634
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.986866391642651e-05,
      "loss": 0.1954,
      "step": 635
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.986800290165252e-05,
      "loss": 0.1895,
      "step": 636
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9867340238658284e-05,
      "loss": 0.1695,
      "step": 637
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9866675927554473e-05,
      "loss": 0.2659,
      "step": 638
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9866009968452054e-05,
      "loss": 0.1561,
      "step": 639
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.986534236146226e-05,
      "loss": 0.1632,
      "step": 640
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9864673106696592e-05,
      "loss": 0.1975,
      "step": 641
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9864002204266844e-05,
      "loss": 0.1968,
      "step": 642
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9863329654285063e-05,
      "loss": 0.2108,
      "step": 643
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9862655456863594e-05,
      "loss": 0.2041,
      "step": 644
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9861979612115037e-05,
      "loss": 0.2057,
      "step": 645
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.986130212015228e-05,
      "loss": 0.2059,
      "step": 646
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9860622981088482e-05,
      "loss": 0.2386,
      "step": 647
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9859942195037076e-05,
      "loss": 0.2018,
      "step": 648
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9859259762111778e-05,
      "loss": 0.2073,
      "step": 649
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9858575682426563e-05,
      "loss": 0.1821,
      "step": 650
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9857889956095695e-05,
      "loss": 0.1784,
      "step": 651
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.985720258323371e-05,
      "loss": 0.2023,
      "step": 652
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9856513563955418e-05,
      "loss": 0.1939,
      "step": 653
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.98558228983759e-05,
      "loss": 0.1689,
      "step": 654
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9855130586610517e-05,
      "loss": 0.1746,
      "step": 655
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9854436628774904e-05,
      "loss": 0.2154,
      "step": 656
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9853741024984973e-05,
      "loss": 0.1938,
      "step": 657
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9853043775356903e-05,
      "loss": 0.2052,
      "step": 658
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9852344880007157e-05,
      "loss": 0.2031,
      "step": 659
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.985164433905247e-05,
      "loss": 0.1767,
      "step": 660
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.985094215260985e-05,
      "loss": 0.2015,
      "step": 661
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9850238320796577e-05,
      "loss": 0.2353,
      "step": 662
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9849532843730218e-05,
      "loss": 0.1781,
      "step": 663
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9848825721528598e-05,
      "loss": 0.1262,
      "step": 664
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.984811695430983e-05,
      "loss": 0.187,
      "step": 665
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9847406542192297e-05,
      "loss": 0.1656,
      "step": 666
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9846694485294657e-05,
      "loss": 0.2202,
      "step": 667
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.984598078373584e-05,
      "loss": 0.1578,
      "step": 668
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.984526543763505e-05,
      "loss": 0.1809,
      "step": 669
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9844548447111783e-05,
      "loss": 0.1982,
      "step": 670
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9843829812285783e-05,
      "loss": 0.2368,
      "step": 671
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9843109533277084e-05,
      "loss": 0.1648,
      "step": 672
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9842387610205987e-05,
      "loss": 0.1647,
      "step": 673
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9841664043193085e-05,
      "loss": 0.1731,
      "step": 674
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9840938832359222e-05,
      "loss": 0.213,
      "step": 675
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9840211977825534e-05,
      "loss": 0.1866,
      "step": 676
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.983948347971342e-05,
      "loss": 0.1923,
      "step": 677
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.983875333814456e-05,
      "loss": 0.1956,
      "step": 678
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9838021553240914e-05,
      "loss": 0.1405,
      "step": 679
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.98372881251247e-05,
      "loss": 0.2314,
      "step": 680
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9836553053918425e-05,
      "loss": 0.1593,
      "step": 681
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9835816339744865e-05,
      "loss": 0.1624,
      "step": 682
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9835077982727066e-05,
      "loss": 0.1697,
      "step": 683
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9834337982988365e-05,
      "loss": 0.1842,
      "step": 684
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.983359634065235e-05,
      "loss": 0.2069,
      "step": 685
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.98328530558429e-05,
      "loss": 0.1679,
      "step": 686
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9832108128684164e-05,
      "loss": 0.1966,
      "step": 687
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9831361559300566e-05,
      "loss": 0.1367,
      "step": 688
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9830613347816802e-05,
      "loss": 0.2081,
      "step": 689
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9829863494357837e-05,
      "loss": 0.2068,
      "step": 690
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9829111999048923e-05,
      "loss": 0.1356,
      "step": 691
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.982835886201558e-05,
      "loss": 0.2022,
      "step": 692
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.98276040833836e-05,
      "loss": 0.1771,
      "step": 693
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9826847663279056e-05,
      "loss": 0.2152,
      "step": 694
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9826089601828283e-05,
      "loss": 0.2158,
      "step": 695
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.98253298991579e-05,
      "loss": 0.2349,
      "step": 696
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.98245685553948e-05,
      "loss": 0.15,
      "step": 697
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9823805570666143e-05,
      "loss": 0.2275,
      "step": 698
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9823040945099375e-05,
      "loss": 0.2027,
      "step": 699
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9822274678822203e-05,
      "loss": 0.1665,
      "step": 700
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.982150677196262e-05,
      "loss": 0.1771,
      "step": 701
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9820737224648875e-05,
      "loss": 0.184,
      "step": 702
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9819966037009516e-05,
      "loss": 0.1689,
      "step": 703
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9819193209173345e-05,
      "loss": 0.1935,
      "step": 704
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9818418741269448e-05,
      "loss": 0.1956,
      "step": 705
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9817642633427182e-05,
      "loss": 0.2031,
      "step": 706
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9816864885776175e-05,
      "loss": 0.2162,
      "step": 707
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.981608549844633e-05,
      "loss": 0.1571,
      "step": 708
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.981530447156783e-05,
      "loss": 0.1844,
      "step": 709
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.981452180527113e-05,
      "loss": 0.1663,
      "step": 710
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.981373749968695e-05,
      "loss": 0.1956,
      "step": 711
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9812951554946288e-05,
      "loss": 0.2176,
      "step": 712
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9812163971180424e-05,
      "loss": 0.1883,
      "step": 713
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9811374748520905e-05,
      "loss": 0.2053,
      "step": 714
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.981058388709955e-05,
      "loss": 0.164,
      "step": 715
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9809791387048453e-05,
      "loss": 0.2199,
      "step": 716
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9808997248499986e-05,
      "loss": 0.1669,
      "step": 717
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.980820147158679e-05,
      "loss": 0.1381,
      "step": 718
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.980740405644178e-05,
      "loss": 0.1983,
      "step": 719
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9806605003198144e-05,
      "loss": 0.1629,
      "step": 720
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9805804311989353e-05,
      "loss": 0.1848,
      "step": 721
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.980500198294913e-05,
      "loss": 0.1572,
      "step": 722
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9804198016211503e-05,
      "loss": 0.2065,
      "step": 723
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9803392411910742e-05,
      "loss": 0.1713,
      "step": 724
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.980258517018141e-05,
      "loss": 0.1955,
      "step": 725
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9801776291158338e-05,
      "loss": 0.168,
      "step": 726
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9800965774976627e-05,
      "loss": 0.2355,
      "step": 727
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9800153621771662e-05,
      "loss": 0.2065,
      "step": 728
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9799339831679085e-05,
      "loss": 0.2366,
      "step": 729
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9798524404834826e-05,
      "loss": 0.2287,
      "step": 730
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9797707341375085e-05,
      "loss": 0.1868,
      "step": 731
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.979688864143633e-05,
      "loss": 0.1577,
      "step": 732
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9796068305155306e-05,
      "loss": 0.1337,
      "step": 733
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.979524633266903e-05,
      "loss": 0.2154,
      "step": 734
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.97944227241148e-05,
      "loss": 0.1791,
      "step": 735
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9793597479630172e-05,
      "loss": 0.1646,
      "step": 736
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9792770599352988e-05,
      "loss": 0.1757,
      "step": 737
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9791942083421362e-05,
      "loss": 0.1965,
      "step": 738
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9791111931973673e-05,
      "loss": 0.127,
      "step": 739
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.979028014514858e-05,
      "loss": 0.1881,
      "step": 740
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9789446723085013e-05,
      "loss": 0.1989,
      "step": 741
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.978861166592218e-05,
      "loss": 0.191,
      "step": 742
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9787774973799552e-05,
      "loss": 0.1705,
      "step": 743
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9786936646856883e-05,
      "loss": 0.1781,
      "step": 744
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9786096685234194e-05,
      "loss": 0.2125,
      "step": 745
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9785255089071788e-05,
      "loss": 0.1708,
      "step": 746
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.978441185851022e-05,
      "loss": 0.1507,
      "step": 747
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.978356699369034e-05,
      "loss": 0.1909,
      "step": 748
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.978272049475327e-05,
      "loss": 0.1963,
      "step": 749
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9781872361840384e-05,
      "loss": 0.2289,
      "step": 750
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9781022595093354e-05,
      "loss": 0.1936,
      "step": 751
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.978017119465411e-05,
      "loss": 0.2113,
      "step": 752
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9779318160664855e-05,
      "loss": 0.1908,
      "step": 753
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.977846349326807e-05,
      "loss": 0.1975,
      "step": 754
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9777607192606515e-05,
      "loss": 0.1837,
      "step": 755
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9776749258823203e-05,
      "loss": 0.181,
      "step": 756
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9775889692061444e-05,
      "loss": 0.1767,
      "step": 757
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.97750284924648e-05,
      "loss": 0.1554,
      "step": 758
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9774165660177115e-05,
      "loss": 0.2429,
      "step": 759
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9773301195342507e-05,
      "loss": 0.1987,
      "step": 760
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9772435098105368e-05,
      "loss": 0.2279,
      "step": 761
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.977156736861035e-05,
      "loss": 0.209,
      "step": 762
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.97706980070024e-05,
      "loss": 0.2015,
      "step": 763
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.976982701342671e-05,
      "loss": 0.1911,
      "step": 764
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9768954388028775e-05,
      "loss": 0.1593,
      "step": 765
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9768080130954336e-05,
      "loss": 0.1111,
      "step": 766
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9767204242349418e-05,
      "loss": 0.1871,
      "step": 767
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.976632672236032e-05,
      "loss": 0.1793,
      "step": 768
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9765447571133612e-05,
      "loss": 0.1727,
      "step": 769
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9764566788816136e-05,
      "loss": 0.1954,
      "step": 770
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9763684375555003e-05,
      "loss": 0.1312,
      "step": 771
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9762800331497606e-05,
      "loss": 0.1797,
      "step": 772
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9761914656791593e-05,
      "loss": 0.1718,
      "step": 773
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9761027351584908e-05,
      "loss": 0.1582,
      "step": 774
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9760138416025746e-05,
      "loss": 0.2013,
      "step": 775
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9759247850262583e-05,
      "loss": 0.2067,
      "step": 776
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9758355654444175e-05,
      "loss": 0.1967,
      "step": 777
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9757461828719534e-05,
      "loss": 0.2063,
      "step": 778
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.975656637323796e-05,
      "loss": 0.2085,
      "step": 779
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.975566928814901e-05,
      "loss": 0.1815,
      "step": 780
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.975477057360253e-05,
      "loss": 0.1633,
      "step": 781
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.975387022974862e-05,
      "loss": 0.1746,
      "step": 782
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.975296825673767e-05,
      "loss": 0.1569,
      "step": 783
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9752064654720325e-05,
      "loss": 0.1811,
      "step": 784
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.975115942384752e-05,
      "loss": 0.1953,
      "step": 785
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9750252564270447e-05,
      "loss": 0.1604,
      "step": 786
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.974934407614058e-05,
      "loss": 0.2039,
      "step": 787
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9748433959609653e-05,
      "loss": 0.1684,
      "step": 788
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9747522214829687e-05,
      "loss": 0.2032,
      "step": 789
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9746608841952966e-05,
      "loss": 0.1707,
      "step": 790
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.974569384113205e-05,
      "loss": 0.1894,
      "step": 791
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9744777212519758e-05,
      "loss": 0.1508,
      "step": 792
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9743858956269207e-05,
      "loss": 0.2048,
      "step": 793
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.974293907253376e-05,
      "loss": 0.1572,
      "step": 794
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9742017561467063e-05,
      "loss": 0.2061,
      "step": 795
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9741094423223034e-05,
      "loss": 0.169,
      "step": 796
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9740169657955864e-05,
      "loss": 0.1842,
      "step": 797
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9739243265820013e-05,
      "loss": 0.2061,
      "step": 798
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.973831524697021e-05,
      "loss": 0.2046,
      "step": 799
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9737385601561465e-05,
      "loss": 0.1649,
      "step": 800
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.973645432974905e-05,
      "loss": 0.1697,
      "step": 801
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9735521431688507e-05,
      "loss": 0.2238,
      "step": 802
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9734586907535664e-05,
      "loss": 0.1974,
      "step": 803
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9733650757446606e-05,
      "loss": 0.1943,
      "step": 804
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9732712981577698e-05,
      "loss": 0.1986,
      "step": 805
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.973177358008557e-05,
      "loss": 0.1951,
      "step": 806
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.973083255312713e-05,
      "loss": 0.1929,
      "step": 807
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9729889900859555e-05,
      "loss": 0.2339,
      "step": 808
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9728945623440295e-05,
      "loss": 0.1486,
      "step": 809
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.972799972102706e-05,
      "loss": 0.2334,
      "step": 810
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9727052193777853e-05,
      "loss": 0.19,
      "step": 811
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.972610304185093e-05,
      "loss": 0.2028,
      "step": 812
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9725152265404825e-05,
      "loss": 0.1443,
      "step": 813
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9724199864598344e-05,
      "loss": 0.1582,
      "step": 814
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9723245839590564e-05,
      "loss": 0.173,
      "step": 815
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9722290190540832e-05,
      "loss": 0.2033,
      "step": 816
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9721332917608767e-05,
      "loss": 0.1806,
      "step": 817
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9720374020954264e-05,
      "loss": 0.1501,
      "step": 818
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.971941350073747e-05,
      "loss": 0.2094,
      "step": 819
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9718451357118835e-05,
      "loss": 0.1822,
      "step": 820
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9717487590259053e-05,
      "loss": 0.1777,
      "step": 821
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9716522200319098e-05,
      "loss": 0.2009,
      "step": 822
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9715555187460217e-05,
      "loss": 0.1806,
      "step": 823
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.971458655184393e-05,
      "loss": 0.2106,
      "step": 824
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.971361629363203e-05,
      "loss": 0.1752,
      "step": 825
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9712644412986562e-05,
      "loss": 0.1994,
      "step": 826
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.971167091006987e-05,
      "loss": 0.162,
      "step": 827
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9710695785044542e-05,
      "loss": 0.2096,
      "step": 828
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.970971903807346e-05,
      "loss": 0.1514,
      "step": 829
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9708740669319762e-05,
      "loss": 0.1802,
      "step": 830
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9707760678946864e-05,
      "loss": 0.1952,
      "step": 831
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9706779067118452e-05,
      "loss": 0.1533,
      "step": 832
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9705795833998476e-05,
      "loss": 0.1529,
      "step": 833
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9704810979751164e-05,
      "loss": 0.2023,
      "step": 834
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.970382450454102e-05,
      "loss": 0.1584,
      "step": 835
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9702836408532802e-05,
      "loss": 0.1526,
      "step": 836
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9701846691891552e-05,
      "loss": 0.1872,
      "step": 837
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9700855354782586e-05,
      "loss": 0.1827,
      "step": 838
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.969986239737147e-05,
      "loss": 0.1663,
      "step": 839
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9698867819824062e-05,
      "loss": 0.1567,
      "step": 840
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9697871622306488e-05,
      "loss": 0.1809,
      "step": 841
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.969687380498513e-05,
      "loss": 0.177,
      "step": 842
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.969587436802666e-05,
      "loss": 0.1684,
      "step": 843
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9694873311598003e-05,
      "loss": 0.2078,
      "step": 844
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9693870635866362e-05,
      "loss": 0.203,
      "step": 845
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9692866340999215e-05,
      "loss": 0.1807,
      "step": 846
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9691860427164307e-05,
      "loss": 0.1754,
      "step": 847
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9690852894529646e-05,
      "loss": 0.163,
      "step": 848
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.968984374326352e-05,
      "loss": 0.161,
      "step": 849
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.968883297353449e-05,
      "loss": 0.1766,
      "step": 850
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9687820585511372e-05,
      "loss": 0.1629,
      "step": 851
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9686806579363267e-05,
      "loss": 0.2342,
      "step": 852
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9685790955259543e-05,
      "loss": 0.175,
      "step": 853
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.968477371336983e-05,
      "loss": 0.1837,
      "step": 854
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9683754853864045e-05,
      "loss": 0.1901,
      "step": 855
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9682734376912355e-05,
      "loss": 0.1466,
      "step": 856
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9681712282685206e-05,
      "loss": 0.1762,
      "step": 857
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9680688571353325e-05,
      "loss": 0.1737,
      "step": 858
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.967966324308769e-05,
      "loss": 0.1566,
      "step": 859
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.967863629805956e-05,
      "loss": 0.1677,
      "step": 860
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.967760773644047e-05,
      "loss": 0.203,
      "step": 861
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9676577558402202e-05,
      "loss": 0.1909,
      "step": 862
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9675545764116843e-05,
      "loss": 0.2448,
      "step": 863
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9674512353756713e-05,
      "loss": 0.1835,
      "step": 864
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9673477327494428e-05,
      "loss": 0.1915,
      "step": 865
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.967244068550286e-05,
      "loss": 0.1688,
      "step": 866
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.967140242795516e-05,
      "loss": 0.1493,
      "step": 867
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9670362555024746e-05,
      "loss": 0.2046,
      "step": 868
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9669321066885302e-05,
      "loss": 0.1835,
      "step": 869
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.966827796371078e-05,
      "loss": 0.1872,
      "step": 870
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9667233245675415e-05,
      "loss": 0.189,
      "step": 871
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9666186912953696e-05,
      "loss": 0.1591,
      "step": 872
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9665138965720393e-05,
      "loss": 0.1753,
      "step": 873
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.966408940415054e-05,
      "loss": 0.1781,
      "step": 874
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9663038228419438e-05,
      "loss": 0.1867,
      "step": 875
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9661985438702668e-05,
      "loss": 0.1871,
      "step": 876
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9660931035176064e-05,
      "loss": 0.1985,
      "step": 877
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9659875018015753e-05,
      "loss": 0.1531,
      "step": 878
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9658817387398107e-05,
      "loss": 0.198,
      "step": 879
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9657758143499784e-05,
      "loss": 0.1509,
      "step": 880
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9656697286497703e-05,
      "loss": 0.1621,
      "step": 881
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.965563481656906e-05,
      "loss": 0.209,
      "step": 882
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9654570733891312e-05,
      "loss": 0.1482,
      "step": 883
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.965350503864219e-05,
      "loss": 0.15,
      "step": 884
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9652437730999692e-05,
      "loss": 0.2015,
      "step": 885
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9651368811142093e-05,
      "loss": 0.1691,
      "step": 886
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9650298279247923e-05,
      "loss": 0.1545,
      "step": 887
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9649226135495997e-05,
      "loss": 0.1739,
      "step": 888
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.964815238006539e-05,
      "loss": 0.1602,
      "step": 889
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9647077013135442e-05,
      "loss": 0.1872,
      "step": 890
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.964600003488577e-05,
      "loss": 0.2005,
      "step": 891
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9644921445496263e-05,
      "loss": 0.1449,
      "step": 892
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.964384124514707e-05,
      "loss": 0.1746,
      "step": 893
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9642759434018617e-05,
      "loss": 0.156,
      "step": 894
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9641676012291593e-05,
      "loss": 0.1603,
      "step": 895
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.964059098014696e-05,
      "loss": 0.1872,
      "step": 896
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.963950433776594e-05,
      "loss": 0.1666,
      "step": 897
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.963841608533004e-05,
      "loss": 0.2008,
      "step": 898
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9637326223021028e-05,
      "loss": 0.1748,
      "step": 899
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9636234751020932e-05,
      "loss": 0.1764,
      "step": 900
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9635141669512063e-05,
      "loss": 0.1873,
      "step": 901
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9634046978676997e-05,
      "loss": 0.2035,
      "step": 902
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.963295067869857e-05,
      "loss": 0.1972,
      "step": 903
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9631852769759893e-05,
      "loss": 0.1824,
      "step": 904
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9630753252044354e-05,
      "loss": 0.1815,
      "step": 905
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9629652125735595e-05,
      "loss": 0.1979,
      "step": 906
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9628549391017538e-05,
      "loss": 0.178,
      "step": 907
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.962744504807437e-05,
      "loss": 0.1863,
      "step": 908
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9626339097090537e-05,
      "loss": 0.1992,
      "step": 909
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.962523153825077e-05,
      "loss": 0.1674,
      "step": 910
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.962412237174006e-05,
      "loss": 0.1847,
      "step": 911
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9623011597743665e-05,
      "loss": 0.2146,
      "step": 912
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9621899216447116e-05,
      "loss": 0.1474,
      "step": 913
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9620785228036214e-05,
      "loss": 0.1994,
      "step": 914
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.961966963269702e-05,
      "loss": 0.2003,
      "step": 915
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9618552430615864e-05,
      "loss": 0.2117,
      "step": 916
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.961743362197936e-05,
      "loss": 0.1567,
      "step": 917
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.961631320697437e-05,
      "loss": 0.1687,
      "step": 918
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9615191185788034e-05,
      "loss": 0.2364,
      "step": 919
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9614067558607765e-05,
      "loss": 0.2125,
      "step": 920
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9612942325621232e-05,
      "loss": 0.1622,
      "step": 921
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9611815487016384e-05,
      "loss": 0.1757,
      "step": 922
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.961068704298143e-05,
      "loss": 0.2072,
      "step": 923
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9609556993704856e-05,
      "loss": 0.1916,
      "step": 924
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.96084253393754e-05,
      "loss": 0.2507,
      "step": 925
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9607292080182086e-05,
      "loss": 0.1702,
      "step": 926
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.96061572163142e-05,
      "loss": 0.126,
      "step": 927
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.960502074796129e-05,
      "loss": 0.1639,
      "step": 928
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9603882675313183e-05,
      "loss": 0.1477,
      "step": 929
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9602742998559958e-05,
      "loss": 0.1856,
      "step": 930
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9601601717891978e-05,
      "loss": 0.2123,
      "step": 931
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9600458833499866e-05,
      "loss": 0.1881,
      "step": 932
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9599314345574513e-05,
      "loss": 0.1835,
      "step": 933
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.959816825430708e-05,
      "loss": 0.1895,
      "step": 934
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9597020559888995e-05,
      "loss": 0.2204,
      "step": 935
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9595871262511954e-05,
      "loss": 0.1954,
      "step": 936
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9594720362367924e-05,
      "loss": 0.1911,
      "step": 937
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9593567859649125e-05,
      "loss": 0.1824,
      "step": 938
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9592413754548067e-05,
      "loss": 0.1572,
      "step": 939
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.959125804725751e-05,
      "loss": 0.2125,
      "step": 940
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9590100737970493e-05,
      "loss": 0.2005,
      "step": 941
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9588941826880316e-05,
      "loss": 0.1612,
      "step": 942
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9587781314180546e-05,
      "loss": 0.177,
      "step": 943
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.958661920006502e-05,
      "loss": 0.1851,
      "step": 944
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9585455484727843e-05,
      "loss": 0.1887,
      "step": 945
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9584290168363385e-05,
      "loss": 0.1529,
      "step": 946
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9583123251166287e-05,
      "loss": 0.1687,
      "step": 947
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9581954733331457e-05,
      "loss": 0.1617,
      "step": 948
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9580784615054062e-05,
      "loss": 0.2024,
      "step": 949
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9579612896529552e-05,
      "loss": 0.1949,
      "step": 950
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9578439577953628e-05,
      "loss": 0.2438,
      "step": 951
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9577264659522267e-05,
      "loss": 0.1615,
      "step": 952
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9576088141431718e-05,
      "loss": 0.1855,
      "step": 953
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.957491002387848e-05,
      "loss": 0.2167,
      "step": 954
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.957373030705934e-05,
      "loss": 0.1816,
      "step": 955
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.957254899117134e-05,
      "loss": 0.1668,
      "step": 956
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9571366076411785e-05,
      "loss": 0.1615,
      "step": 957
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9570181562978262e-05,
      "loss": 0.2234,
      "step": 958
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.956899545106861e-05,
      "loss": 0.1419,
      "step": 959
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9567807740880947e-05,
      "loss": 0.2011,
      "step": 960
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9566618432613647e-05,
      "loss": 0.197,
      "step": 961
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.956542752646536e-05,
      "loss": 0.1612,
      "step": 962
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9564235022634998e-05,
      "loss": 0.1787,
      "step": 963
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.956304092132174e-05,
      "loss": 0.1996,
      "step": 964
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9561845222725033e-05,
      "loss": 0.1774,
      "step": 965
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.956064792704459e-05,
      "loss": 0.2282,
      "step": 966
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9559449034480392e-05,
      "loss": 0.1776,
      "step": 967
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.955824854523269e-05,
      "loss": 0.1813,
      "step": 968
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9557046459501993e-05,
      "loss": 0.2339,
      "step": 969
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9555842777489084e-05,
      "loss": 0.1653,
      "step": 970
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.955463749939501e-05,
      "loss": 0.1613,
      "step": 971
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9553430625421078e-05,
      "loss": 0.2019,
      "step": 972
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.955222215576888e-05,
      "loss": 0.2084,
      "step": 973
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9551012090640253e-05,
      "loss": 0.1815,
      "step": 974
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9549800430237315e-05,
      "loss": 0.1589,
      "step": 975
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9548587174762445e-05,
      "loss": 0.2018,
      "step": 976
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9547372324418286e-05,
      "loss": 0.194,
      "step": 977
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.954615587940776e-05,
      "loss": 0.1635,
      "step": 978
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9544937839934033e-05,
      "loss": 0.1847,
      "step": 979
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.954371820620056e-05,
      "loss": 0.1908,
      "step": 980
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9542496978411044e-05,
      "loss": 0.2102,
      "step": 981
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.954127415676947e-05,
      "loss": 0.2049,
      "step": 982
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9540049741480075e-05,
      "loss": 0.1459,
      "step": 983
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9538823732747375e-05,
      "loss": 0.1709,
      "step": 984
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9537596130776147e-05,
      "loss": 0.1936,
      "step": 985
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9536366935771427e-05,
      "loss": 0.1843,
      "step": 986
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9535136147938533e-05,
      "loss": 0.1467,
      "step": 987
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.953390376748303e-05,
      "loss": 0.1821,
      "step": 988
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.953266979461076e-05,
      "loss": 0.1706,
      "step": 989
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9531434229527833e-05,
      "loss": 0.1701,
      "step": 990
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9530197072440626e-05,
      "loss": 0.1585,
      "step": 991
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.952895832355576e-05,
      "loss": 0.179,
      "step": 992
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.952771798308016e-05,
      "loss": 0.2143,
      "step": 993
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.952647605122098e-05,
      "loss": 0.1398,
      "step": 994
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9525232528185668e-05,
      "loss": 0.155,
      "step": 995
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9523987414181916e-05,
      "loss": 0.1523,
      "step": 996
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9522740709417694e-05,
      "loss": 0.1545,
      "step": 997
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.952149241410124e-05,
      "loss": 0.1518,
      "step": 998
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.952024252844105e-05,
      "loss": 0.1659,
      "step": 999
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.951899105264588e-05,
      "loss": 0.1745,
      "step": 1000
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.951773798692477e-05,
      "loss": 0.2117,
      "step": 1001
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9516483331487017e-05,
      "loss": 0.1578,
      "step": 1002
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9515227086542176e-05,
      "loss": 0.2224,
      "step": 1003
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9513969252300068e-05,
      "loss": 0.2034,
      "step": 1004
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9512709828970798e-05,
      "loss": 0.136,
      "step": 1005
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.951144881676472e-05,
      "loss": 0.1718,
      "step": 1006
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.951018621589245e-05,
      "loss": 0.2095,
      "step": 1007
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.950892202656488e-05,
      "loss": 0.1461,
      "step": 1008
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9507656248993164e-05,
      "loss": 0.2026,
      "step": 1009
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.950638888338872e-05,
      "loss": 0.161,
      "step": 1010
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9505119929963235e-05,
      "loss": 0.2002,
      "step": 1011
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9503849388928654e-05,
      "loss": 0.1477,
      "step": 1012
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.950257726049719e-05,
      "loss": 0.1825,
      "step": 1013
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.950130354488133e-05,
      "loss": 0.1564,
      "step": 1014
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9500028242293808e-05,
      "loss": 0.1482,
      "step": 1015
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.949875135294764e-05,
      "loss": 0.1626,
      "step": 1016
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9497472877056106e-05,
      "loss": 0.1463,
      "step": 1017
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9496192814832735e-05,
      "loss": 0.2232,
      "step": 1018
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9494911166491338e-05,
      "loss": 0.191,
      "step": 1019
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.949362793224598e-05,
      "loss": 0.1704,
      "step": 1020
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9492343112311e-05,
      "loss": 0.2393,
      "step": 1021
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9491056706900996e-05,
      "loss": 0.176,
      "step": 1022
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9489768716230826e-05,
      "loss": 0.208,
      "step": 1023
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.948847914051563e-05,
      "loss": 0.1557,
      "step": 1024
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9487187979970797e-05,
      "loss": 0.1734,
      "step": 1025
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.948589523481198e-05,
      "loss": 0.194,
      "step": 1026
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.948460090525511e-05,
      "loss": 0.1693,
      "step": 1027
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9483304991516366e-05,
      "loss": 0.199,
      "step": 1028
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.948200749381221e-05,
      "loss": 0.1928,
      "step": 1029
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9480708412359352e-05,
      "loss": 0.1709,
      "step": 1030
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9479407747374775e-05,
      "loss": 0.1518,
      "step": 1031
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9478105499075725e-05,
      "loss": 0.1764,
      "step": 1032
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9476801667679713e-05,
      "loss": 0.2044,
      "step": 1033
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9475496253404515e-05,
      "loss": 0.1957,
      "step": 1034
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9474189256468167e-05,
      "loss": 0.1791,
      "step": 1035
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9472880677088973e-05,
      "loss": 0.1812,
      "step": 1036
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9471570515485498e-05,
      "loss": 0.2253,
      "step": 1037
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9470258771876582e-05,
      "loss": 0.1477,
      "step": 1038
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9468945446481315e-05,
      "loss": 0.221,
      "step": 1039
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.946763053951906e-05,
      "loss": 0.1606,
      "step": 1040
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.946631405120944e-05,
      "loss": 0.214,
      "step": 1041
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9464995981772342e-05,
      "loss": 0.151,
      "step": 1042
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9463676331427925e-05,
      "loss": 0.2216,
      "step": 1043
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9462355100396598e-05,
      "loss": 0.1843,
      "step": 1044
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.946103228889905e-05,
      "loss": 0.1885,
      "step": 1045
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.945970789715622e-05,
      "loss": 0.1814,
      "step": 1046
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9458381925389317e-05,
      "loss": 0.1376,
      "step": 1047
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.945705437381982e-05,
      "loss": 0.1737,
      "step": 1048
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9455725242669453e-05,
      "loss": 0.1569,
      "step": 1049
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.945439453216023e-05,
      "loss": 0.1704,
      "step": 1050
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.945306224251441e-05,
      "loss": 0.1673,
      "step": 1051
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.945172837395452e-05,
      "loss": 0.1704,
      "step": 1052
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9450392926703353e-05,
      "loss": 0.2083,
      "step": 1053
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9449055900983962e-05,
      "loss": 0.1847,
      "step": 1054
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.944771729701967e-05,
      "loss": 0.1513,
      "step": 1055
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9446377115034055e-05,
      "loss": 0.246,
      "step": 1056
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9445035355250965e-05,
      "loss": 0.1988,
      "step": 1057
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9443692017894513e-05,
      "loss": 0.1629,
      "step": 1058
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9442347103189068e-05,
      "loss": 0.2288,
      "step": 1059
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.944100061135927e-05,
      "loss": 0.168,
      "step": 1060
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9439652542630016e-05,
      "loss": 0.1586,
      "step": 1061
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.943830289722647e-05,
      "loss": 0.1732,
      "step": 1062
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.943695167537406e-05,
      "loss": 0.1861,
      "step": 1063
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9435598877298477e-05,
      "loss": 0.1597,
      "step": 1064
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9434244503225674e-05,
      "loss": 0.197,
      "step": 1065
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9432888553381866e-05,
      "loss": 0.2076,
      "step": 1066
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9431531027993534e-05,
      "loss": 0.1793,
      "step": 1067
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9430171927287418e-05,
      "loss": 0.1892,
      "step": 1068
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9428811251490533e-05,
      "loss": 0.1644,
      "step": 1069
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.942744900083014e-05,
      "loss": 0.1555,
      "step": 1070
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.942608517553377e-05,
      "loss": 0.2126,
      "step": 1071
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9424719775829226e-05,
      "loss": 0.1435,
      "step": 1072
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9423352801944563e-05,
      "loss": 0.1775,
      "step": 1073
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9421984254108096e-05,
      "loss": 0.1675,
      "step": 1074
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.942061413254842e-05,
      "loss": 0.1479,
      "step": 1075
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.941924243749437e-05,
      "loss": 0.181,
      "step": 1076
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.941786916917507e-05,
      "loss": 0.196,
      "step": 1077
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9416494327819877e-05,
      "loss": 0.1836,
      "step": 1078
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9415117913658438e-05,
      "loss": 0.1365,
      "step": 1079
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9413739926920646e-05,
      "loss": 0.1935,
      "step": 1080
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9412360367836666e-05,
      "loss": 0.2131,
      "step": 1081
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.941097923663691e-05,
      "loss": 0.2098,
      "step": 1082
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9409596533552078e-05,
      "loss": 0.1985,
      "step": 1083
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9408212258813107e-05,
      "loss": 0.1626,
      "step": 1084
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9406826412651216e-05,
      "loss": 0.1806,
      "step": 1085
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9405438995297875e-05,
      "loss": 0.1853,
      "step": 1086
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.940405000698482e-05,
      "loss": 0.1473,
      "step": 1087
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9402659447944048e-05,
      "loss": 0.2004,
      "step": 1088
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9401267318407816e-05,
      "loss": 0.1647,
      "step": 1089
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.939987361860866e-05,
      "loss": 0.142,
      "step": 1090
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.939847834877935e-05,
      "loss": 0.1657,
      "step": 1091
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.939708150915294e-05,
      "loss": 0.1726,
      "step": 1092
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9395683099962744e-05,
      "loss": 0.181,
      "step": 1093
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9394283121442327e-05,
      "loss": 0.155,
      "step": 1094
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9392881573825523e-05,
      "loss": 0.1414,
      "step": 1095
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9391478457346436e-05,
      "loss": 0.1642,
      "step": 1096
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.939007377223941e-05,
      "loss": 0.1911,
      "step": 1097
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9388667518739077e-05,
      "loss": 0.2139,
      "step": 1098
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9387259697080317e-05,
      "loss": 0.1824,
      "step": 1099
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9385850307498266e-05,
      "loss": 0.1909,
      "step": 1100
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.938443935022834e-05,
      "loss": 0.2133,
      "step": 1101
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9383026825506196e-05,
      "loss": 0.1673,
      "step": 1102
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9381612733567775e-05,
      "loss": 0.1615,
      "step": 1103
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.938019707464926e-05,
      "loss": 0.2013,
      "step": 1104
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9378779848987107e-05,
      "loss": 0.1977,
      "step": 1105
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.937736105681803e-05,
      "loss": 0.1734,
      "step": 1106
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9375940698379002e-05,
      "loss": 0.1695,
      "step": 1107
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9374518773907266e-05,
      "loss": 0.1749,
      "step": 1108
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.937309528364032e-05,
      "loss": 0.199,
      "step": 1109
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9371670227815926e-05,
      "loss": 0.1885,
      "step": 1110
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9370243606672102e-05,
      "loss": 0.1646,
      "step": 1111
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9368815420447136e-05,
      "loss": 0.1526,
      "step": 1112
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.936738566937957e-05,
      "loss": 0.1759,
      "step": 1113
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9365954353708214e-05,
      "loss": 0.1795,
      "step": 1114
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9364521473672132e-05,
      "loss": 0.1857,
      "step": 1115
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9363087029510657e-05,
      "loss": 0.1954,
      "step": 1116
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.936165102146338e-05,
      "loss": 0.1353,
      "step": 1117
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.936021344977015e-05,
      "loss": 0.2129,
      "step": 1118
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.935877431467108e-05,
      "loss": 0.1801,
      "step": 1119
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9357333616406545e-05,
      "loss": 0.193,
      "step": 1120
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.935589135521718e-05,
      "loss": 0.169,
      "step": 1121
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9354447531343877e-05,
      "loss": 0.1833,
      "step": 1122
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9353002145027804e-05,
      "loss": 0.149,
      "step": 1123
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.935155519651037e-05,
      "loss": 0.3029,
      "step": 1124
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9350106686033257e-05,
      "loss": 0.1565,
      "step": 1125
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.93486566138384e-05,
      "loss": 0.1776,
      "step": 1126
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.934720498016801e-05,
      "loss": 0.1755,
      "step": 1127
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9345751785264543e-05,
      "loss": 0.1761,
      "step": 1128
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9344297029370717e-05,
      "loss": 0.1408,
      "step": 1129
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9342840712729524e-05,
      "loss": 0.1649,
      "step": 1130
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.93413828355842e-05,
      "loss": 0.1599,
      "step": 1131
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9339923398178256e-05,
      "loss": 0.159,
      "step": 1132
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.933846240075545e-05,
      "loss": 0.1549,
      "step": 1133
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9336999843559813e-05,
      "loss": 0.1485,
      "step": 1134
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.933553572683563e-05,
      "loss": 0.1783,
      "step": 1135
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9334070050827445e-05,
      "loss": 0.1894,
      "step": 1136
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9332602815780068e-05,
      "loss": 0.1637,
      "step": 1137
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.933113402193857e-05,
      "loss": 0.1585,
      "step": 1138
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9329663669548267e-05,
      "loss": 0.2271,
      "step": 1139
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.932819175885476e-05,
      "loss": 0.1695,
      "step": 1140
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.932671829010389e-05,
      "loss": 0.1918,
      "step": 1141
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9325243263541766e-05,
      "loss": 0.1534,
      "step": 1142
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9323766679414763e-05,
      "loss": 0.1724,
      "step": 1143
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9322288537969504e-05,
      "loss": 0.1537,
      "step": 1144
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9320808839452878e-05,
      "loss": 0.1779,
      "step": 1145
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.931932758411204e-05,
      "loss": 0.1455,
      "step": 1146
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.931784477219439e-05,
      "loss": 0.1626,
      "step": 1147
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9316360403947608e-05,
      "loss": 0.1671,
      "step": 1148
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9314874479619612e-05,
      "loss": 0.1406,
      "step": 1149
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.93133869994586e-05,
      "loss": 0.1643,
      "step": 1150
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.931189796371302e-05,
      "loss": 0.1564,
      "step": 1151
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9310407372631572e-05,
      "loss": 0.176,
      "step": 1152
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9308915226463235e-05,
      "loss": 0.1512,
      "step": 1153
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9307421525457233e-05,
      "loss": 0.1876,
      "step": 1154
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9305926269863052e-05,
      "loss": 0.1747,
      "step": 1155
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9304429459930444e-05,
      "loss": 0.1714,
      "step": 1156
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9302931095909418e-05,
      "loss": 0.1558,
      "step": 1157
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9301431178050228e-05,
      "loss": 0.1741,
      "step": 1158
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9299929706603417e-05,
      "loss": 0.1546,
      "step": 1159
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.929842668181976e-05,
      "loss": 0.174,
      "step": 1160
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9296922103950308e-05,
      "loss": 0.2335,
      "step": 1161
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.929541597324636e-05,
      "loss": 0.1821,
      "step": 1162
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9293908289959483e-05,
      "loss": 0.1623,
      "step": 1163
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9292399054341503e-05,
      "loss": 0.1286,
      "step": 1164
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.92908882666445e-05,
      "loss": 0.2109,
      "step": 1165
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9289375927120817e-05,
      "loss": 0.1937,
      "step": 1166
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9287862036023056e-05,
      "loss": 0.1586,
      "step": 1167
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9286346593604072e-05,
      "loss": 0.1639,
      "step": 1168
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.928482960011699e-05,
      "loss": 0.1994,
      "step": 1169
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9283311055815188e-05,
      "loss": 0.1717,
      "step": 1170
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.92817909609523e-05,
      "loss": 0.1469,
      "step": 1171
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9280269315782225e-05,
      "loss": 0.1613,
      "step": 1172
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.927874612055912e-05,
      "loss": 0.1664,
      "step": 1173
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9277221375537393e-05,
      "loss": 0.195,
      "step": 1174
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9275695080971727e-05,
      "loss": 0.1978,
      "step": 1175
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9274167237117044e-05,
      "loss": 0.1784,
      "step": 1176
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9272637844228543e-05,
      "loss": 0.1741,
      "step": 1177
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9271106902561668e-05,
      "loss": 0.1795,
      "step": 1178
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9269574412372127e-05,
      "loss": 0.2047,
      "step": 1179
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9268040373915893e-05,
      "loss": 0.1508,
      "step": 1180
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9266504787449184e-05,
      "loss": 0.1597,
      "step": 1181
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9264967653228484e-05,
      "loss": 0.1944,
      "step": 1182
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9263428971510543e-05,
      "loss": 0.1664,
      "step": 1183
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.926188874255236e-05,
      "loss": 0.1842,
      "step": 1184
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.926034696661119e-05,
      "loss": 0.1778,
      "step": 1185
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.925880364394455e-05,
      "loss": 0.2058,
      "step": 1186
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.925725877481022e-05,
      "loss": 0.1741,
      "step": 1187
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.925571235946624e-05,
      "loss": 0.1589,
      "step": 1188
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.925416439817089e-05,
      "loss": 0.1486,
      "step": 1189
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9252614891182725e-05,
      "loss": 0.1773,
      "step": 1190
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.925106383876056e-05,
      "loss": 0.1753,
      "step": 1191
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9249511241163456e-05,
      "loss": 0.2392,
      "step": 1192
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.924795709865074e-05,
      "loss": 0.1656,
      "step": 1193
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9246401411482e-05,
      "loss": 0.1944,
      "step": 1194
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9244844179917072e-05,
      "loss": 0.1811,
      "step": 1195
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9243285404216054e-05,
      "loss": 0.1926,
      "step": 1196
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9241725084639307e-05,
      "loss": 0.1996,
      "step": 1197
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9240163221447448e-05,
      "loss": 0.194,
      "step": 1198
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9238599814901345e-05,
      "loss": 0.141,
      "step": 1199
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9237034865262128e-05,
      "loss": 0.144,
      "step": 1200
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9235468372791192e-05,
      "loss": 0.1621,
      "step": 1201
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9233900337750175e-05,
      "loss": 0.1603,
      "step": 1202
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9232330760400987e-05,
      "loss": 0.1676,
      "step": 1203
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9230759641005786e-05,
      "loss": 0.1413,
      "step": 1204
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9229186979826992e-05,
      "loss": 0.1723,
      "step": 1205
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9227612777127286e-05,
      "loss": 0.1907,
      "step": 1206
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.922603703316959e-05,
      "loss": 0.1635,
      "step": 1207
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9224459748217108e-05,
      "loss": 0.1752,
      "step": 1208
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9222880922533284e-05,
      "loss": 0.1546,
      "step": 1209
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.922130055638182e-05,
      "loss": 0.1845,
      "step": 1210
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9219718650026685e-05,
      "loss": 0.1877,
      "step": 1211
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.92181352037321e-05,
      "loss": 0.1712,
      "step": 1212
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9216550217762533e-05,
      "loss": 0.1599,
      "step": 1213
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9214963692382733e-05,
      "loss": 0.1764,
      "step": 1214
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9213375627857683e-05,
      "loss": 0.1985,
      "step": 1215
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9211786024452638e-05,
      "loss": 0.1683,
      "step": 1216
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9210194882433098e-05,
      "loss": 0.1719,
      "step": 1217
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9208602202064832e-05,
      "loss": 0.166,
      "step": 1218
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9207007983613856e-05,
      "loss": 0.1969,
      "step": 1219
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9205412227346452e-05,
      "loss": 0.1791,
      "step": 1220
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9203814933529145e-05,
      "loss": 0.1721,
      "step": 1221
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9202216102428734e-05,
      "loss": 0.1721,
      "step": 1222
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9200615734312267e-05,
      "loss": 0.1852,
      "step": 1223
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.919901382944704e-05,
      "loss": 0.1686,
      "step": 1224
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9197410388100624e-05,
      "loss": 0.1761,
      "step": 1225
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9195805410540827e-05,
      "loss": 0.1631,
      "step": 1226
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9194198897035733e-05,
      "loss": 0.1816,
      "step": 1227
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9192590847853664e-05,
      "loss": 0.2173,
      "step": 1228
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.919098126326321e-05,
      "loss": 0.1533,
      "step": 1229
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.918937014353322e-05,
      "loss": 0.144,
      "step": 1230
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9187757488932786e-05,
      "loss": 0.1697,
      "step": 1231
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.918614329973127e-05,
      "loss": 0.1774,
      "step": 1232
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.918452757619828e-05,
      "loss": 0.1764,
      "step": 1233
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.918291031860369e-05,
      "loss": 0.147,
      "step": 1234
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.918129152721762e-05,
      "loss": 0.1307,
      "step": 1235
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9179671202310458e-05,
      "loss": 0.2051,
      "step": 1236
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9178049344152833e-05,
      "loss": 0.1741,
      "step": 1237
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.917642595301565e-05,
      "loss": 0.1475,
      "step": 1238
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9174801029170047e-05,
      "loss": 0.1909,
      "step": 1239
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9173174572887437e-05,
      "loss": 0.1936,
      "step": 1240
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.917154658443948e-05,
      "loss": 0.1666,
      "step": 1241
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9169917064098094e-05,
      "loss": 0.157,
      "step": 1242
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9168286012135447e-05,
      "loss": 0.1527,
      "step": 1243
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9166653428823977e-05,
      "loss": 0.1615,
      "step": 1244
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9165019314436366e-05,
      "loss": 0.1892,
      "step": 1245
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9163383669245553e-05,
      "loss": 0.1756,
      "step": 1246
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.916174649352473e-05,
      "loss": 0.1212,
      "step": 1247
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9160107787547364e-05,
      "loss": 0.2116,
      "step": 1248
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9158467551587146e-05,
      "loss": 0.1804,
      "step": 1249
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.915682578591805e-05,
      "loss": 0.1487,
      "step": 1250
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.915518249081429e-05,
      "loss": 0.168,
      "step": 1251
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9153537666550338e-05,
      "loss": 0.1994,
      "step": 1252
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.915189131340093e-05,
      "loss": 0.1705,
      "step": 1253
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.915024343164105e-05,
      "loss": 0.1614,
      "step": 1254
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.914859402154593e-05,
      "loss": 0.1893,
      "step": 1255
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9146943083391077e-05,
      "loss": 0.1712,
      "step": 1256
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9145290617452236e-05,
      "loss": 0.1641,
      "step": 1257
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9143636624005414e-05,
      "loss": 0.1539,
      "step": 1258
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.914198110332687e-05,
      "loss": 0.1733,
      "step": 1259
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9140324055693124e-05,
      "loss": 0.2099,
      "step": 1260
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.913866548138094e-05,
      "loss": 0.166,
      "step": 1261
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9137005380667354e-05,
      "loss": 0.1648,
      "step": 1262
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9135343753829644e-05,
      "loss": 0.1426,
      "step": 1263
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9133680601145342e-05,
      "loss": 0.1448,
      "step": 1264
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.913201592289224e-05,
      "loss": 0.1978,
      "step": 1265
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9130349719348388e-05,
      "loss": 0.1622,
      "step": 1266
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.912868199079208e-05,
      "loss": 0.145,
      "step": 1267
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.912701273750188e-05,
      "loss": 0.1849,
      "step": 1268
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.912534195975659e-05,
      "loss": 0.1987,
      "step": 1269
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.912366965783528e-05,
      "loss": 0.1937,
      "step": 1270
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9121995832017263e-05,
      "loss": 0.1903,
      "step": 1271
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9120320482582123e-05,
      "loss": 0.1407,
      "step": 1272
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9118643609809676e-05,
      "loss": 0.1551,
      "step": 1273
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.911696521398001e-05,
      "loss": 0.1591,
      "step": 1274
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9115285295373462e-05,
      "loss": 0.1601,
      "step": 1275
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9113603854270625e-05,
      "loss": 0.1401,
      "step": 1276
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9111920890952345e-05,
      "loss": 0.152,
      "step": 1277
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9110236405699716e-05,
      "loss": 0.1517,
      "step": 1278
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9108550398794098e-05,
      "loss": 0.1611,
      "step": 1279
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.91068628705171e-05,
      "loss": 0.2297,
      "step": 1280
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9105173821150576e-05,
      "loss": 0.1939,
      "step": 1281
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.910348325097665e-05,
      "loss": 0.1821,
      "step": 1282
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.910179116027769e-05,
      "loss": 0.1665,
      "step": 1283
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9100097549336322e-05,
      "loss": 0.1583,
      "step": 1284
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.909840241843542e-05,
      "loss": 0.1697,
      "step": 1285
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.909670576785812e-05,
      "loss": 0.1642,
      "step": 1286
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.909500759788781e-05,
      "loss": 0.1785,
      "step": 1287
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9093307908808122e-05,
      "loss": 0.1813,
      "step": 1288
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9091606700902957e-05,
      "loss": 0.1383,
      "step": 1289
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.908990397445646e-05,
      "loss": 0.1665,
      "step": 1290
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.908819972975303e-05,
      "loss": 0.2211,
      "step": 1291
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9086493967077325e-05,
      "loss": 0.1577,
      "step": 1292
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.908478668671425e-05,
      "loss": 0.1923,
      "step": 1293
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.908307788894897e-05,
      "loss": 0.177,
      "step": 1294
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9081367574066893e-05,
      "loss": 0.1862,
      "step": 1295
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.907965574235369e-05,
      "loss": 0.1877,
      "step": 1296
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9077942394095286e-05,
      "loss": 0.1642,
      "step": 1297
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9076227529577856e-05,
      "loss": 0.1866,
      "step": 1298
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9074511149087823e-05,
      "loss": 0.172,
      "step": 1299
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9072793252911874e-05,
      "loss": 0.1638,
      "step": 1300
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9071073841336943e-05,
      "loss": 0.1759,
      "step": 1301
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.906935291465021e-05,
      "loss": 0.1583,
      "step": 1302
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.906763047313913e-05,
      "loss": 0.189,
      "step": 1303
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.906590651709138e-05,
      "loss": 0.1644,
      "step": 1304
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.906418104679492e-05,
      "loss": 0.1519,
      "step": 1305
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9062454062537943e-05,
      "loss": 0.1725,
      "step": 1306
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9060725564608903e-05,
      "loss": 0.1849,
      "step": 1307
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.905899555329651e-05,
      "loss": 0.1409,
      "step": 1308
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.905726402888971e-05,
      "loss": 0.1639,
      "step": 1309
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.905553099167773e-05,
      "loss": 0.1798,
      "step": 1310
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9053796441950024e-05,
      "loss": 0.1579,
      "step": 1311
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9052060379996304e-05,
      "loss": 0.1873,
      "step": 1312
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9050322806106548e-05,
      "loss": 0.1561,
      "step": 1313
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9048583720570975e-05,
      "loss": 0.2102,
      "step": 1314
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9046843123680054e-05,
      "loss": 0.1446,
      "step": 1315
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9045101015724517e-05,
      "loss": 0.1579,
      "step": 1316
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.904335739699534e-05,
      "loss": 0.1481,
      "step": 1317
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.904161226778375e-05,
      "loss": 0.1738,
      "step": 1318
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.903986562838124e-05,
      "loss": 0.1795,
      "step": 1319
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9038117479079534e-05,
      "loss": 0.2434,
      "step": 1320
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9036367820170628e-05,
      "loss": 0.1688,
      "step": 1321
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.903461665194676e-05,
      "loss": 0.1576,
      "step": 1322
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.903286397470042e-05,
      "loss": 0.1883,
      "step": 1323
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.903110978872435e-05,
      "loss": 0.1762,
      "step": 1324
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9029354094311554e-05,
      "loss": 0.1992,
      "step": 1325
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9027596891755273e-05,
      "loss": 0.14,
      "step": 1326
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.902583818134901e-05,
      "loss": 0.1837,
      "step": 1327
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9024077963386512e-05,
      "loss": 0.1556,
      "step": 1328
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9022316238161788e-05,
      "loss": 0.1508,
      "step": 1329
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9020553005969092e-05,
      "loss": 0.1387,
      "step": 1330
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9018788267102928e-05,
      "loss": 0.1518,
      "step": 1331
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9017022021858056e-05,
      "loss": 0.1616,
      "step": 1332
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9015254270529487e-05,
      "loss": 0.1417,
      "step": 1333
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.901348501341248e-05,
      "loss": 0.1807,
      "step": 1334
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9011714250802554e-05,
      "loss": 0.1767,
      "step": 1335
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.900994198299547e-05,
      "loss": 0.1874,
      "step": 1336
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9008168210287247e-05,
      "loss": 0.1788,
      "step": 1337
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9006392932974146e-05,
      "loss": 0.1826,
      "step": 1338
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9004616151352687e-05,
      "loss": 0.1379,
      "step": 1339
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.9002837865719644e-05,
      "loss": 0.1631,
      "step": 1340
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.900105807637204e-05,
      "loss": 0.1744,
      "step": 1341
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8999276783607144e-05,
      "loss": 0.2088,
      "step": 1342
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8997493987722477e-05,
      "loss": 0.1526,
      "step": 1343
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8995709689015817e-05,
      "loss": 0.2326,
      "step": 1344
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8993923887785187e-05,
      "loss": 0.1948,
      "step": 1345
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.899213658432887e-05,
      "loss": 0.1686,
      "step": 1346
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8990347778945385e-05,
      "loss": 0.1944,
      "step": 1347
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8988557471933516e-05,
      "loss": 0.1841,
      "step": 1348
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.898676566359229e-05,
      "loss": 0.1724,
      "step": 1349
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8984972354220988e-05,
      "loss": 0.1327,
      "step": 1350
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8983177544119137e-05,
      "loss": 0.1955,
      "step": 1351
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8981381233586522e-05,
      "loss": 0.2197,
      "step": 1352
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8979583422923175e-05,
      "loss": 0.1844,
      "step": 1353
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8977784112429376e-05,
      "loss": 0.1808,
      "step": 1354
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.897598330240566e-05,
      "loss": 0.1844,
      "step": 1355
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.897418099315281e-05,
      "loss": 0.1673,
      "step": 1356
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.897237718497186e-05,
      "loss": 0.1786,
      "step": 1357
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.897057187816409e-05,
      "loss": 0.1815,
      "step": 1358
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.896876507303104e-05,
      "loss": 0.2062,
      "step": 1359
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8966956769874496e-05,
      "loss": 0.145,
      "step": 1360
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8965146968996488e-05,
      "loss": 0.1554,
      "step": 1361
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.89633356706993e-05,
      "loss": 0.1521,
      "step": 1362
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8961522875285472e-05,
      "loss": 0.1456,
      "step": 1363
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8959708583057788e-05,
      "loss": 0.1452,
      "step": 1364
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8957892794319287e-05,
      "loss": 0.1568,
      "step": 1365
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8956075509373246e-05,
      "loss": 0.1564,
      "step": 1366
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.89542567285232e-05,
      "loss": 0.1805,
      "step": 1367
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8952436452072943e-05,
      "loss": 0.1664,
      "step": 1368
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8950614680326505e-05,
      "loss": 0.1343,
      "step": 1369
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8948791413588173e-05,
      "loss": 0.1737,
      "step": 1370
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8946966652162474e-05,
      "loss": 0.1359,
      "step": 1371
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.89451403963542e-05,
      "loss": 0.1422,
      "step": 1372
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8943312646468378e-05,
      "loss": 0.1857,
      "step": 1373
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8941483402810298e-05,
      "loss": 0.1887,
      "step": 1374
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.893965266568549e-05,
      "loss": 0.1854,
      "step": 1375
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.893782043539973e-05,
      "loss": 0.1634,
      "step": 1376
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8935986712259056e-05,
      "loss": 0.1808,
      "step": 1377
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.893415149656975e-05,
      "loss": 0.1806,
      "step": 1378
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8932314788638337e-05,
      "loss": 0.188,
      "step": 1379
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.89304765887716e-05,
      "loss": 0.1558,
      "step": 1380
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8928636897276565e-05,
      "loss": 0.1301,
      "step": 1381
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8926795714460513e-05,
      "loss": 0.1748,
      "step": 1382
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.892495304063097e-05,
      "loss": 0.1538,
      "step": 1383
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8923108876095705e-05,
      "loss": 0.2039,
      "step": 1384
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8921263221162752e-05,
      "loss": 0.1703,
      "step": 1385
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.891941607614038e-05,
      "loss": 0.1913,
      "step": 1386
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.891756744133711e-05,
      "loss": 0.1536,
      "step": 1387
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.891571731706172e-05,
      "loss": 0.1563,
      "step": 1388
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8913865703623225e-05,
      "loss": 0.1794,
      "step": 1389
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8912012601330892e-05,
      "loss": 0.1834,
      "step": 1390
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8910158010494246e-05,
      "loss": 0.1607,
      "step": 1391
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.890830193142305e-05,
      "loss": 0.1777,
      "step": 1392
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.890644436442731e-05,
      "loss": 0.1377,
      "step": 1393
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.89045853098173e-05,
      "loss": 0.1736,
      "step": 1394
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.890272476790353e-05,
      "loss": 0.1783,
      "step": 1395
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.890086273899676e-05,
      "loss": 0.193,
      "step": 1396
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8898999223408e-05,
      "loss": 0.172,
      "step": 1397
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.88971342214485e-05,
      "loss": 0.1824,
      "step": 1398
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.889526773342977e-05,
      "loss": 0.1683,
      "step": 1399
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8893399759663567e-05,
      "loss": 0.1593,
      "step": 1400
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8891530300461888e-05,
      "loss": 0.2167,
      "step": 1401
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.888965935613698e-05,
      "loss": 0.1709,
      "step": 1402
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8887786927001345e-05,
      "loss": 0.1746,
      "step": 1403
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.888591301336773e-05,
      "loss": 0.1858,
      "step": 1404
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8884037615549124e-05,
      "loss": 0.1533,
      "step": 1405
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.888216073385877e-05,
      "loss": 0.1347,
      "step": 1406
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8880282368610162e-05,
      "loss": 0.1659,
      "step": 1407
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8878402520117028e-05,
      "loss": 0.156,
      "step": 1408
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8876521188693366e-05,
      "loss": 0.1401,
      "step": 1409
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8874638374653397e-05,
      "loss": 0.1579,
      "step": 1410
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.887275407831161e-05,
      "loss": 0.2032,
      "step": 1411
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.887086829998272e-05,
      "loss": 0.1856,
      "step": 1412
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8868981039981717e-05,
      "loss": 0.1644,
      "step": 1413
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8867092298623816e-05,
      "loss": 0.1939,
      "step": 1414
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8865202076224487e-05,
      "loss": 0.2191,
      "step": 1415
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8863310373099453e-05,
      "loss": 0.1591,
      "step": 1416
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.886141718956467e-05,
      "loss": 0.1464,
      "step": 1417
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8859522525936362e-05,
      "loss": 0.1859,
      "step": 1418
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.885762638253098e-05,
      "loss": 0.1636,
      "step": 1419
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8855728759665232e-05,
      "loss": 0.1551,
      "step": 1420
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8853829657656074e-05,
      "loss": 0.1461,
      "step": 1421
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8851929076820705e-05,
      "loss": 0.1371,
      "step": 1422
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8850027017476575e-05,
      "loss": 0.1606,
      "step": 1423
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8848123479941373e-05,
      "loss": 0.1895,
      "step": 1424
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8846218464533046e-05,
      "loss": 0.1688,
      "step": 1425
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.884431197156978e-05,
      "loss": 0.1379,
      "step": 1426
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8842404001370017e-05,
      "loss": 0.2025,
      "step": 1427
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8840494554252425e-05,
      "loss": 0.1471,
      "step": 1428
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.883858363053595e-05,
      "loss": 0.1678,
      "step": 1429
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8836671230539752e-05,
      "loss": 0.2095,
      "step": 1430
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8834757354583262e-05,
      "loss": 0.1998,
      "step": 1431
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8832842002986145e-05,
      "loss": 0.164,
      "step": 1432
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.883092517606832e-05,
      "loss": 0.1877,
      "step": 1433
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.882900687414994e-05,
      "loss": 0.1606,
      "step": 1434
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8827087097551418e-05,
      "loss": 0.1554,
      "step": 1435
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8825165846593413e-05,
      "loss": 0.1424,
      "step": 1436
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8823243121596815e-05,
      "loss": 0.1948,
      "step": 1437
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8821318922882777e-05,
      "loss": 0.171,
      "step": 1438
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8819393250772692e-05,
      "loss": 0.1459,
      "step": 1439
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8817466105588194e-05,
      "loss": 0.1991,
      "step": 1440
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.881553748765117e-05,
      "loss": 0.1794,
      "step": 1441
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8813607397283756e-05,
      "loss": 0.1491,
      "step": 1442
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8811675834808317e-05,
      "loss": 0.1396,
      "step": 1443
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.880974280054749e-05,
      "loss": 0.1515,
      "step": 1444
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8807808294824134e-05,
      "loss": 0.1808,
      "step": 1445
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.880587231796136e-05,
      "loss": 0.1972,
      "step": 1446
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.880393487028254e-05,
      "loss": 0.1577,
      "step": 1447
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.880199595211127e-05,
      "loss": 0.165,
      "step": 1448
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.88000555637714e-05,
      "loss": 0.1616,
      "step": 1449
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8798113705587035e-05,
      "loss": 0.1679,
      "step": 1450
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.879617037788251e-05,
      "loss": 0.1511,
      "step": 1451
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8794225580982418e-05,
      "loss": 0.1582,
      "step": 1452
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8792279315211587e-05,
      "loss": 0.1654,
      "step": 1453
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8790331580895097e-05,
      "loss": 0.1371,
      "step": 1454
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8788382378358276e-05,
      "loss": 0.1946,
      "step": 1455
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.878643170792669e-05,
      "loss": 0.1828,
      "step": 1456
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.878447956992615e-05,
      "loss": 0.1558,
      "step": 1457
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8782525964682717e-05,
      "loss": 0.1304,
      "step": 1458
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8780570892522698e-05,
      "loss": 0.16,
      "step": 1459
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.877861435377264e-05,
      "loss": 0.1672,
      "step": 1460
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8776656348759337e-05,
      "loss": 0.1576,
      "step": 1461
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.877469687780983e-05,
      "loss": 0.1908,
      "step": 1462
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.87727359412514e-05,
      "loss": 0.1828,
      "step": 1463
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8770773539411576e-05,
      "loss": 0.1513,
      "step": 1464
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8768809672618137e-05,
      "loss": 0.1917,
      "step": 1465
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8766844341199096e-05,
      "loss": 0.158,
      "step": 1466
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8764877545482717e-05,
      "loss": 0.1428,
      "step": 1467
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.876290928579751e-05,
      "loss": 0.1652,
      "step": 1468
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8760939562472224e-05,
      "loss": 0.1312,
      "step": 1469
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8758968375835854e-05,
      "loss": 0.2004,
      "step": 1470
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8756995726217643e-05,
      "loss": 0.1917,
      "step": 1471
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8755021613947078e-05,
      "loss": 0.1784,
      "step": 1472
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8753046039353887e-05,
      "loss": 0.1422,
      "step": 1473
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8751069002768042e-05,
      "loss": 0.2166,
      "step": 1474
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.874909050451976e-05,
      "loss": 0.1819,
      "step": 1475
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8747110544939513e-05,
      "loss": 0.1658,
      "step": 1476
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8745129124357996e-05,
      "loss": 0.1405,
      "step": 1477
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8743146243106163e-05,
      "loss": 0.2145,
      "step": 1478
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8741161901515207e-05,
      "loss": 0.1757,
      "step": 1479
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.873917609991657e-05,
      "loss": 0.1532,
      "step": 1480
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.873718883864193e-05,
      "loss": 0.168,
      "step": 1481
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8735200118023213e-05,
      "loss": 0.1536,
      "step": 1482
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8733209938392586e-05,
      "loss": 0.2041,
      "step": 1483
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8731218300082472e-05,
      "loss": 0.157,
      "step": 1484
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8729225203425518e-05,
      "loss": 0.1949,
      "step": 1485
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.872723064875463e-05,
      "loss": 0.1768,
      "step": 1486
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.872523463640295e-05,
      "loss": 0.1388,
      "step": 1487
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.872323716670386e-05,
      "loss": 0.1801,
      "step": 1488
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8721238239991004e-05,
      "loss": 0.1531,
      "step": 1489
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8719237856598243e-05,
      "loss": 0.1634,
      "step": 1490
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8717236016859702e-05,
      "loss": 0.1861,
      "step": 1491
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8715232721109737e-05,
      "loss": 0.1797,
      "step": 1492
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8713227969682955e-05,
      "loss": 0.2034,
      "step": 1493
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8711221762914206e-05,
      "loss": 0.1407,
      "step": 1494
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8709214101138572e-05,
      "loss": 0.1883,
      "step": 1495
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8707204984691394e-05,
      "loss": 0.1557,
      "step": 1496
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8705194413908242e-05,
      "loss": 0.153,
      "step": 1497
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8703182389124942e-05,
      "loss": 0.1869,
      "step": 1498
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.870116891067755e-05,
      "loss": 0.1631,
      "step": 1499
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.869915397890237e-05,
      "loss": 0.1712,
      "step": 1500
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8697137594135955e-05,
      "loss": 0.154,
      "step": 1501
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8695119756715097e-05,
      "loss": 0.1532,
      "step": 1502
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.869310046697682e-05,
      "loss": 0.1734,
      "step": 1503
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8691079725258402e-05,
      "loss": 0.1831,
      "step": 1504
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.868905753189736e-05,
      "loss": 0.1604,
      "step": 1505
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.868703388723146e-05,
      "loss": 0.1305,
      "step": 1506
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8685008791598703e-05,
      "loss": 0.1562,
      "step": 1507
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8682982245337328e-05,
      "loss": 0.1619,
      "step": 1508
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.868095424878583e-05,
      "loss": 0.1351,
      "step": 1509
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8678924802282937e-05,
      "loss": 0.1534,
      "step": 1510
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8676893906167616e-05,
      "loss": 0.1581,
      "step": 1511
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8674861560779087e-05,
      "loss": 0.2134,
      "step": 1512
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8672827766456803e-05,
      "loss": 0.1946,
      "step": 1513
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8670792523540462e-05,
      "loss": 0.1608,
      "step": 1514
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8668755832370004e-05,
      "loss": 0.1544,
      "step": 1515
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8666717693285614e-05,
      "loss": 0.1526,
      "step": 1516
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8664678106627703e-05,
      "loss": 0.1551,
      "step": 1517
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8662637072736956e-05,
      "loss": 0.1708,
      "step": 1518
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.866059459195427e-05,
      "loss": 0.1433,
      "step": 1519
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8658550664620788e-05,
      "loss": 0.1286,
      "step": 1520
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8656505291077916e-05,
      "loss": 0.21,
      "step": 1521
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.865445847166727e-05,
      "loss": 0.214,
      "step": 1522
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8652410206730732e-05,
      "loss": 0.2069,
      "step": 1523
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8650360496610417e-05,
      "loss": 0.168,
      "step": 1524
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.864830934164868e-05,
      "loss": 0.1372,
      "step": 1525
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8646256742188118e-05,
      "loss": 0.1499,
      "step": 1526
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.864420269857157e-05,
      "loss": 0.1633,
      "step": 1527
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8642147211142116e-05,
      "loss": 0.1325,
      "step": 1528
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.864009028024308e-05,
      "loss": 0.1422,
      "step": 1529
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8638031906218024e-05,
      "loss": 0.16,
      "step": 1530
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8635972089410747e-05,
      "loss": 0.181,
      "step": 1531
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8633910830165298e-05,
      "loss": 0.148,
      "step": 1532
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.863184812882596e-05,
      "loss": 0.1303,
      "step": 1533
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.862978398573726e-05,
      "loss": 0.1586,
      "step": 1534
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.862771840124397e-05,
      "loss": 0.1465,
      "step": 1535
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.862565137569109e-05,
      "loss": 0.165,
      "step": 1536
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8623582909423866e-05,
      "loss": 0.1914,
      "step": 1537
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.86215130027878e-05,
      "loss": 0.167,
      "step": 1538
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8619441656128613e-05,
      "loss": 0.1923,
      "step": 1539
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.861736886979227e-05,
      "loss": 0.1609,
      "step": 1540
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8615294644124997e-05,
      "loss": 0.1653,
      "step": 1541
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8613218979473233e-05,
      "loss": 0.1742,
      "step": 1542
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8611141876183675e-05,
      "loss": 0.1363,
      "step": 1543
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8609063334603252e-05,
      "loss": 0.1617,
      "step": 1544
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8606983355079136e-05,
      "loss": 0.153,
      "step": 1545
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.860490193795874e-05,
      "loss": 0.1499,
      "step": 1546
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8602819083589717e-05,
      "loss": 0.1683,
      "step": 1547
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8600734792319958e-05,
      "loss": 0.144,
      "step": 1548
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8598649064497597e-05,
      "loss": 0.1624,
      "step": 1549
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8596561900471003e-05,
      "loss": 0.2117,
      "step": 1550
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8594473300588794e-05,
      "loss": 0.156,
      "step": 1551
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8592383265199817e-05,
      "loss": 0.209,
      "step": 1552
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8590291794653164e-05,
      "loss": 0.2158,
      "step": 1553
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.858819888929817e-05,
      "loss": 0.161,
      "step": 1554
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.85861045494844e-05,
      "loss": 0.2192,
      "step": 1555
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.858400877556167e-05,
      "loss": 0.1352,
      "step": 1556
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8581911567880032e-05,
      "loss": 0.1993,
      "step": 1557
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.857981292678977e-05,
      "loss": 0.1781,
      "step": 1558
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.857771285264142e-05,
      "loss": 0.1832,
      "step": 1559
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.857561134578574e-05,
      "loss": 0.1512,
      "step": 1560
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8573508406573747e-05,
      "loss": 0.1609,
      "step": 1561
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8571404035356687e-05,
      "loss": 0.1466,
      "step": 1562
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.856929823248604e-05,
      "loss": 0.1358,
      "step": 1563
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.856719099831354e-05,
      "loss": 0.196,
      "step": 1564
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8565082333191147e-05,
      "loss": 0.1984,
      "step": 1565
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.856297223747106e-05,
      "loss": 0.1775,
      "step": 1566
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8560860711505732e-05,
      "loss": 0.163,
      "step": 1567
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.855874775564783e-05,
      "loss": 0.1705,
      "step": 1568
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8556633370250293e-05,
      "loss": 0.1681,
      "step": 1569
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.855451755566626e-05,
      "loss": 0.2162,
      "step": 1570
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8552400312249147e-05,
      "loss": 0.1745,
      "step": 1571
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8550281640352574e-05,
      "loss": 0.1655,
      "step": 1572
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8548161540330425e-05,
      "loss": 0.195,
      "step": 1573
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8546040012536814e-05,
      "loss": 0.167,
      "step": 1574
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8543917057326086e-05,
      "loss": 0.1519,
      "step": 1575
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8541792675052835e-05,
      "loss": 0.2014,
      "step": 1576
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8539666866071898e-05,
      "loss": 0.1968,
      "step": 1577
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8537539630738327e-05,
      "loss": 0.1818,
      "step": 1578
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8535410969407435e-05,
      "loss": 0.2164,
      "step": 1579
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8533280882434767e-05,
      "loss": 0.2034,
      "step": 1580
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.85311493701761e-05,
      "loss": 0.1464,
      "step": 1581
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8529016432987453e-05,
      "loss": 0.2121,
      "step": 1582
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8526882071225087e-05,
      "loss": 0.1684,
      "step": 1583
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.85247462852455e-05,
      "loss": 0.1943,
      "step": 1584
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.852260907540542e-05,
      "loss": 0.1555,
      "step": 1585
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8520470442061817e-05,
      "loss": 0.1514,
      "step": 1586
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.851833038557191e-05,
      "loss": 0.1195,
      "step": 1587
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8516188906293132e-05,
      "loss": 0.1764,
      "step": 1588
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8514046004583175e-05,
      "loss": 0.156,
      "step": 1589
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.851190168079996e-05,
      "loss": 0.1626,
      "step": 1590
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.850975593530165e-05,
      "loss": 0.1761,
      "step": 1591
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8507608768446634e-05,
      "loss": 0.1329,
      "step": 1592
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.850546018059355e-05,
      "loss": 0.1896,
      "step": 1593
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8503310172101276e-05,
      "loss": 0.1714,
      "step": 1594
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8501158743328908e-05,
      "loss": 0.2009,
      "step": 1595
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8499005894635796e-05,
      "loss": 0.1537,
      "step": 1596
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8496851626381534e-05,
      "loss": 0.1682,
      "step": 1597
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.849469593892593e-05,
      "loss": 0.195,
      "step": 1598
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8492538832629047e-05,
      "loss": 0.1675,
      "step": 1599
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8490380307851177e-05,
      "loss": 0.1797,
      "step": 1600
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8488220364952855e-05,
      "loss": 0.1694,
      "step": 1601
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8486059004294844e-05,
      "loss": 0.1435,
      "step": 1602
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8483896226238153e-05,
      "loss": 0.1973,
      "step": 1603
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8481732031144023e-05,
      "loss": 0.1486,
      "step": 1604
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8479566419373936e-05,
      "loss": 0.1873,
      "step": 1605
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.84773993912896e-05,
      "loss": 0.1378,
      "step": 1606
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8475230947252965e-05,
      "loss": 0.2036,
      "step": 1607
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.8473061087626227e-05,
      "loss": 0.1924,
      "step": 1608
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.847088981277181e-05,
      "loss": 0.1838,
      "step": 1609
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8468717123052367e-05,
      "loss": 0.1889,
      "step": 1610
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8466543018830804e-05,
      "loss": 0.139,
      "step": 1611
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8464367500470252e-05,
      "loss": 0.1325,
      "step": 1612
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.846219056833408e-05,
      "loss": 0.1525,
      "step": 1613
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8460012222785893e-05,
      "loss": 0.1505,
      "step": 1614
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.845783246418953e-05,
      "loss": 0.1964,
      "step": 1615
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8455651292909075e-05,
      "loss": 0.1217,
      "step": 1616
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.845346870930884e-05,
      "loss": 0.2034,
      "step": 1617
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8451284713753374e-05,
      "loss": 0.1527,
      "step": 1618
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.844909930660746e-05,
      "loss": 0.2252,
      "step": 1619
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8446912488236127e-05,
      "loss": 0.1625,
      "step": 1620
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.844472425900463e-05,
      "loss": 0.1781,
      "step": 1621
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.844253461927845e-05,
      "loss": 0.1266,
      "step": 1622
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.8440343569423328e-05,
      "loss": 0.1544,
      "step": 1623
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.843815110980523e-05,
      "loss": 0.1992,
      "step": 1624
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8435957240790342e-05,
      "loss": 0.1286,
      "step": 1625
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.843376196274511e-05,
      "loss": 0.1661,
      "step": 1626
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8431565276036202e-05,
      "loss": 0.1358,
      "step": 1627
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8429367181030517e-05,
      "loss": 0.1565,
      "step": 1628
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8427167678095206e-05,
      "loss": 0.1623,
      "step": 1629
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8424966767597636e-05,
      "loss": 0.1756,
      "step": 1630
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8422764449905423e-05,
      "loss": 0.1456,
      "step": 1631
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8420560725386414e-05,
      "loss": 0.1458,
      "step": 1632
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.841835559440868e-05,
      "loss": 0.136,
      "step": 1633
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.841614905734055e-05,
      "loss": 0.158,
      "step": 1634
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8413941114550566e-05,
      "loss": 0.189,
      "step": 1635
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.841173176640752e-05,
      "loss": 0.1668,
      "step": 1636
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.840952101328042e-05,
      "loss": 0.1486,
      "step": 1637
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8407308855538538e-05,
      "loss": 0.1366,
      "step": 1638
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.840509529355135e-05,
      "loss": 0.1493,
      "step": 1639
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.8402880327688585e-05,
      "loss": 0.1536,
      "step": 1640
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8400663958320202e-05,
      "loss": 0.156,
      "step": 1641
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.839844618581639e-05,
      "loss": 0.1686,
      "step": 1642
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8396227010547585e-05,
      "loss": 0.1609,
      "step": 1643
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8394006432884438e-05,
      "loss": 0.1716,
      "step": 1644
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8391784453197855e-05,
      "loss": 0.1759,
      "step": 1645
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8389561071858956e-05,
      "loss": 0.1764,
      "step": 1646
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8387336289239113e-05,
      "loss": 0.1784,
      "step": 1647
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8385110105709917e-05,
      "loss": 0.1508,
      "step": 1648
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8382882521643205e-05,
      "loss": 0.1479,
      "step": 1649
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8380653537411048e-05,
      "loss": 0.2017,
      "step": 1650
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8378423153385733e-05,
      "loss": 0.156,
      "step": 1651
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8376191369939807e-05,
      "loss": 0.155,
      "step": 1652
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.837395818744603e-05,
      "loss": 0.1772,
      "step": 1653
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.83717236062774e-05,
      "loss": 0.1653,
      "step": 1654
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.836948762680716e-05,
      "loss": 0.1496,
      "step": 1655
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.8367250249408776e-05,
      "loss": 0.1972,
      "step": 1656
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.836501147445595e-05,
      "loss": 0.1891,
      "step": 1657
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.836277130232261e-05,
      "loss": 0.1758,
      "step": 1658
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8360529733382933e-05,
      "loss": 0.1637,
      "step": 1659
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.835828676801132e-05,
      "loss": 0.1747,
      "step": 1660
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.83560424065824e-05,
      "loss": 0.1462,
      "step": 1661
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8353796649471053e-05,
      "loss": 0.1533,
      "step": 1662
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8351549497052367e-05,
      "loss": 0.1642,
      "step": 1663
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8349300949701686e-05,
      "loss": 0.1425,
      "step": 1664
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8347051007794575e-05,
      "loss": 0.1597,
      "step": 1665
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8344799671706832e-05,
      "loss": 0.1606,
      "step": 1666
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8342546941814493e-05,
      "loss": 0.1817,
      "step": 1667
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8340292818493825e-05,
      "loss": 0.1795,
      "step": 1668
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.8338037302121322e-05,
      "loss": 0.1798,
      "step": 1669
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.833578039307372e-05,
      "loss": 0.1896,
      "step": 1670
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.833352209172798e-05,
      "loss": 0.1714,
      "step": 1671
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.83312623984613e-05,
      "loss": 0.1786,
      "step": 1672
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8329001313651113e-05,
      "loss": 0.1682,
      "step": 1673
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8326738837675075e-05,
      "loss": 0.1694,
      "step": 1674
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8324474970911085e-05,
      "loss": 0.1608,
      "step": 1675
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8322209713737263e-05,
      "loss": 0.1544,
      "step": 1676
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8319943066531976e-05,
      "loss": 0.1891,
      "step": 1677
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8317675029673804e-05,
      "loss": 0.1579,
      "step": 1678
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8315405603541578e-05,
      "loss": 0.1622,
      "step": 1679
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8313134788514347e-05,
      "loss": 0.1863,
      "step": 1680
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8310862584971406e-05,
      "loss": 0.1405,
      "step": 1681
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8308588993292268e-05,
      "loss": 0.1566,
      "step": 1682
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8306314013856683e-05,
      "loss": 0.1709,
      "step": 1683
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8304037647044634e-05,
      "loss": 0.1206,
      "step": 1684
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8301759893236334e-05,
      "loss": 0.2219,
      "step": 1685
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8299480752812236e-05,
      "loss": 0.2146,
      "step": 1686
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8297200226153008e-05,
      "loss": 0.16,
      "step": 1687
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.8294918313639566e-05,
      "loss": 0.1785,
      "step": 1688
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8292635015653044e-05,
      "loss": 0.1494,
      "step": 1689
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.829035033257482e-05,
      "loss": 0.2329,
      "step": 1690
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.828806426478649e-05,
      "loss": 0.1521,
      "step": 1691
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8285776812669896e-05,
      "loss": 0.1907,
      "step": 1692
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8283487976607097e-05,
      "loss": 0.167,
      "step": 1693
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8281197756980398e-05,
      "loss": 0.1438,
      "step": 1694
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8278906154172318e-05,
      "loss": 0.1678,
      "step": 1695
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.827661316856562e-05,
      "loss": 0.1477,
      "step": 1696
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8274318800543295e-05,
      "loss": 0.1453,
      "step": 1697
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8272023050488564e-05,
      "loss": 0.1394,
      "step": 1698
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8269725918784875e-05,
      "loss": 0.1669,
      "step": 1699
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8267427405815912e-05,
      "loss": 0.1778,
      "step": 1700
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.826512751196559e-05,
      "loss": 0.1609,
      "step": 1701
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8262826237618053e-05,
      "loss": 0.1721,
      "step": 1702
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8260523583157673e-05,
      "loss": 0.1568,
      "step": 1703
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8258219548969052e-05,
      "loss": 0.1548,
      "step": 1704
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8255914135437032e-05,
      "loss": 0.1479,
      "step": 1705
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8253607342946676e-05,
      "loss": 0.1528,
      "step": 1706
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.825129917188328e-05,
      "loss": 0.1635,
      "step": 1707
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.824898962263237e-05,
      "loss": 0.1813,
      "step": 1708
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8246678695579698e-05,
      "loss": 0.1672,
      "step": 1709
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8244366391111258e-05,
      "loss": 0.1547,
      "step": 1710
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8242052709613264e-05,
      "loss": 0.1629,
      "step": 1711
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.823973765147216e-05,
      "loss": 0.1389,
      "step": 1712
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8237421217074626e-05,
      "loss": 0.1455,
      "step": 1713
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.823510340680757e-05,
      "loss": 0.1444,
      "step": 1714
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.823278422105812e-05,
      "loss": 0.187,
      "step": 1715
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.823046366021365e-05,
      "loss": 0.1645,
      "step": 1716
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8228141724661753e-05,
      "loss": 0.1464,
      "step": 1717
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8225818414790256e-05,
      "loss": 0.1582,
      "step": 1718
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8223493730987214e-05,
      "loss": 0.1811,
      "step": 1719
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8221167673640906e-05,
      "loss": 0.1746,
      "step": 1720
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.821884024313985e-05,
      "loss": 0.1421,
      "step": 1721
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.821651143987279e-05,
      "loss": 0.1288,
      "step": 1722
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8214181264228696e-05,
      "loss": 0.198,
      "step": 1723
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.821184971659677e-05,
      "loss": 0.1909,
      "step": 1724
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8209516797366445e-05,
      "loss": 0.1885,
      "step": 1725
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8207182506927377e-05,
      "loss": 0.1899,
      "step": 1726
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.820484684566946e-05,
      "loss": 0.1655,
      "step": 1727
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8202509813982808e-05,
      "loss": 0.1845,
      "step": 1728
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8200171412257766e-05,
      "loss": 0.1788,
      "step": 1729
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8197831640884917e-05,
      "loss": 0.1623,
      "step": 1730
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8195490500255055e-05,
      "loss": 0.1763,
      "step": 1731
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8193147990759225e-05,
      "loss": 0.1356,
      "step": 1732
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8190804112788678e-05,
      "loss": 0.156,
      "step": 1733
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.818845886673491e-05,
      "loss": 0.1458,
      "step": 1734
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8186112252989637e-05,
      "loss": 0.1312,
      "step": 1735
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.818376427194481e-05,
      "loss": 0.1484,
      "step": 1736
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.81814149239926e-05,
      "loss": 0.1686,
      "step": 1737
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8179064209525416e-05,
      "loss": 0.1779,
      "step": 1738
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8176712128935885e-05,
      "loss": 0.1698,
      "step": 1739
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8174358682616868e-05,
      "loss": 0.139,
      "step": 1740
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.817200387096146e-05,
      "loss": 0.1589,
      "step": 1741
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.816964769436297e-05,
      "loss": 0.1403,
      "step": 1742
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8167290153214945e-05,
      "loss": 0.157,
      "step": 1743
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8164931247911156e-05,
      "loss": 0.168,
      "step": 1744
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8162570978845608e-05,
      "loss": 0.1674,
      "step": 1745
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8160209346412525e-05,
      "loss": 0.1505,
      "step": 1746
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8157846351006364e-05,
      "loss": 0.1858,
      "step": 1747
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8155481993021807e-05,
      "loss": 0.1338,
      "step": 1748
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8153116272853762e-05,
      "loss": 0.1354,
      "step": 1749
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.815074919089738e-05,
      "loss": 0.1657,
      "step": 1750
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8148380747548012e-05,
      "loss": 0.1944,
      "step": 1751
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.814601094320126e-05,
      "loss": 0.1599,
      "step": 1752
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8143639778252937e-05,
      "loss": 0.1541,
      "step": 1753
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.81412672530991e-05,
      "loss": 0.1485,
      "step": 1754
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.813889336813602e-05,
      "loss": 0.1845,
      "step": 1755
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.81365181237602e-05,
      "loss": 0.1498,
      "step": 1756
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8134141520368374e-05,
      "loss": 0.161,
      "step": 1757
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8131763558357485e-05,
      "loss": 0.1601,
      "step": 1758
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.812938423812473e-05,
      "loss": 0.1953,
      "step": 1759
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8127003560067508e-05,
      "loss": 0.2039,
      "step": 1760
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8124621524583466e-05,
      "loss": 0.1723,
      "step": 1761
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8122238132070463e-05,
      "loss": 0.1396,
      "step": 1762
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8119853382926588e-05,
      "loss": 0.1565,
      "step": 1763
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.811746727755016e-05,
      "loss": 0.1702,
      "step": 1764
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8115079816339717e-05,
      "loss": 0.1629,
      "step": 1765
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8112690999694037e-05,
      "loss": 0.1919,
      "step": 1766
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8110300828012113e-05,
      "loss": 0.1893,
      "step": 1767
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8107909301693167e-05,
      "loss": 0.1985,
      "step": 1768
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8105516421136645e-05,
      "loss": 0.1236,
      "step": 1769
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8103122186742226e-05,
      "loss": 0.1688,
      "step": 1770
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8100726598909816e-05,
      "loss": 0.1948,
      "step": 1771
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.809832965803953e-05,
      "loss": 0.1618,
      "step": 1772
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.809593136453173e-05,
      "loss": 0.1869,
      "step": 1773
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.809353171878699e-05,
      "loss": 0.1697,
      "step": 1774
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8091130721206117e-05,
      "loss": 0.1434,
      "step": 1775
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8088728372190145e-05,
      "loss": 0.1441,
      "step": 1776
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8086324672140328e-05,
      "loss": 0.1633,
      "step": 1777
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8083919621458146e-05,
      "loss": 0.2127,
      "step": 1778
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.808151322054531e-05,
      "loss": 0.1971,
      "step": 1779
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.807910546980375e-05,
      "loss": 0.1875,
      "step": 1780
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8076696369635632e-05,
      "loss": 0.1596,
      "step": 1781
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8074285920443327e-05,
      "loss": 0.1761,
      "step": 1782
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.807187412262946e-05,
      "loss": 0.1439,
      "step": 1783
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8069460976596853e-05,
      "loss": 0.1269,
      "step": 1784
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8067046482748574e-05,
      "loss": 0.1551,
      "step": 1785
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8064630641487902e-05,
      "loss": 0.1667,
      "step": 1786
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.806221345321835e-05,
      "loss": 0.1619,
      "step": 1787
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8059794918343655e-05,
      "loss": 0.1627,
      "step": 1788
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8057375037267775e-05,
      "loss": 0.1522,
      "step": 1789
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8054953810394893e-05,
      "loss": 0.1728,
      "step": 1790
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8052531238129424e-05,
      "loss": 0.1927,
      "step": 1791
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8050107320875993e-05,
      "loss": 0.1618,
      "step": 1792
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8047682059039468e-05,
      "loss": 0.134,
      "step": 1793
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.804525545302493e-05,
      "loss": 0.1513,
      "step": 1794
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8042827503237687e-05,
      "loss": 0.1506,
      "step": 1795
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.804039821008327e-05,
      "loss": 0.1721,
      "step": 1796
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8037967573967434e-05,
      "loss": 0.1712,
      "step": 1797
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8035535595296164e-05,
      "loss": 0.1668,
      "step": 1798
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8033102274475664e-05,
      "loss": 0.1583,
      "step": 1799
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8030667611912366e-05,
      "loss": 0.1434,
      "step": 1800
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.802823160801292e-05,
      "loss": 0.1619,
      "step": 1801
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.802579426318421e-05,
      "loss": 0.1739,
      "step": 1802
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8023355577833328e-05,
      "loss": 0.172,
      "step": 1803
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8020915552367607e-05,
      "loss": 0.1905,
      "step": 1804
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8018474187194592e-05,
      "loss": 0.142,
      "step": 1805
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8016031482722064e-05,
      "loss": 0.1431,
      "step": 1806
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8013587439358007e-05,
      "loss": 0.1244,
      "step": 1807
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8011142057510653e-05,
      "loss": 0.1492,
      "step": 1808
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8008695337588446e-05,
      "loss": 0.1502,
      "step": 1809
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8006247280000045e-05,
      "loss": 0.1438,
      "step": 1810
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8003797885154343e-05,
      "loss": 0.1827,
      "step": 1811
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.800134715346046e-05,
      "loss": 0.168,
      "step": 1812
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.7998895085327735e-05,
      "loss": 0.1553,
      "step": 1813
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.799644168116572e-05,
      "loss": 0.181,
      "step": 1814
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7993986941384207e-05,
      "loss": 0.1662,
      "step": 1815
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7991530866393197e-05,
      "loss": 0.1643,
      "step": 1816
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.798907345660293e-05,
      "loss": 0.1569,
      "step": 1817
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7986614712423843e-05,
      "loss": 0.1424,
      "step": 1818
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.798415463426663e-05,
      "loss": 0.1306,
      "step": 1819
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7981693222542178e-05,
      "loss": 0.1674,
      "step": 1820
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.797923047766161e-05,
      "loss": 0.1636,
      "step": 1821
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.797676640003627e-05,
      "loss": 0.1528,
      "step": 1822
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7974300990077732e-05,
      "loss": 0.1609,
      "step": 1823
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.797183424819778e-05,
      "loss": 0.1455,
      "step": 1824
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.796936617480842e-05,
      "loss": 0.1467,
      "step": 1825
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.79668967703219e-05,
      "loss": 0.1672,
      "step": 1826
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7964426035150666e-05,
      "loss": 0.1465,
      "step": 1827
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.79619539697074e-05,
      "loss": 0.1235,
      "step": 1828
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7959480574405e-05,
      "loss": 0.1445,
      "step": 1829
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7957005849656596e-05,
      "loss": 0.1724,
      "step": 1830
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7954529795875527e-05,
      "loss": 0.1379,
      "step": 1831
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.795205241347536e-05,
      "loss": 0.1752,
      "step": 1832
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7949573702869886e-05,
      "loss": 0.168,
      "step": 1833
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.794709366447312e-05,
      "loss": 0.1692,
      "step": 1834
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7944612298699288e-05,
      "loss": 0.1758,
      "step": 1835
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.794212960596285e-05,
      "loss": 0.1939,
      "step": 1836
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7939645586678475e-05,
      "loss": 0.1622,
      "step": 1837
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7937160241261067e-05,
      "loss": 0.1281,
      "step": 1838
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7934673570125743e-05,
      "loss": 0.1591,
      "step": 1839
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7932185573687843e-05,
      "loss": 0.1382,
      "step": 1840
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.792969625236293e-05,
      "loss": 0.1377,
      "step": 1841
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7927205606566786e-05,
      "loss": 0.1557,
      "step": 1842
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.792471363671542e-05,
      "loss": 0.1868,
      "step": 1843
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.792222034322505e-05,
      "loss": 0.1801,
      "step": 1844
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.791972572651212e-05,
      "loss": 0.1818,
      "step": 1845
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7917229786993315e-05,
      "loss": 0.1359,
      "step": 1846
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.791473252508551e-05,
      "loss": 0.1856,
      "step": 1847
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.791223394120581e-05,
      "loss": 0.1479,
      "step": 1848
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.790973403577156e-05,
      "loss": 0.1886,
      "step": 1849
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7907232809200303e-05,
      "loss": 0.1336,
      "step": 1850
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7904730261909808e-05,
      "loss": 0.1664,
      "step": 1851
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7902226394318073e-05,
      "loss": 0.1429,
      "step": 1852
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.789972120684331e-05,
      "loss": 0.1768,
      "step": 1853
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7897214699903945e-05,
      "loss": 0.1663,
      "step": 1854
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7894706873918637e-05,
      "loss": 0.1463,
      "step": 1855
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7892197729306264e-05,
      "loss": 0.1542,
      "step": 1856
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7889687266485913e-05,
      "loss": 0.1614,
      "step": 1857
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7887175485876905e-05,
      "loss": 0.1653,
      "step": 1858
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.788466238789877e-05,
      "loss": 0.1378,
      "step": 1859
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.788214797297126e-05,
      "loss": 0.1554,
      "step": 1860
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7879632241514364e-05,
      "loss": 0.1564,
      "step": 1861
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7877115193948254e-05,
      "loss": 0.1693,
      "step": 1862
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.787459683069336e-05,
      "loss": 0.161,
      "step": 1863
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7872077152170314e-05,
      "loss": 0.1893,
      "step": 1864
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7869556158799967e-05,
      "loss": 0.1506,
      "step": 1865
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.786703385100339e-05,
      "loss": 0.1836,
      "step": 1866
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.786451022920188e-05,
      "loss": 0.1611,
      "step": 1867
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7861985293816947e-05,
      "loss": 0.196,
      "step": 1868
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7859459045270325e-05,
      "loss": 0.1695,
      "step": 1869
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7856931483983963e-05,
      "loss": 0.1549,
      "step": 1870
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.785440261038003e-05,
      "loss": 0.1347,
      "step": 1871
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7851872424880923e-05,
      "loss": 0.1818,
      "step": 1872
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7849340927909244e-05,
      "loss": 0.1727,
      "step": 1873
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7846808119887815e-05,
      "loss": 0.1554,
      "step": 1874
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7844274001239697e-05,
      "loss": 0.1422,
      "step": 1875
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7841738572388145e-05,
      "loss": 0.1292,
      "step": 1876
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7839201833756644e-05,
      "loss": 0.1768,
      "step": 1877
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.78366637857689e-05,
      "loss": 0.143,
      "step": 1878
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.783412442884884e-05,
      "loss": 0.1646,
      "step": 1879
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.783158376342059e-05,
      "loss": 0.1669,
      "step": 1880
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7829041789908524e-05,
      "loss": 0.1705,
      "step": 1881
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.782649850873721e-05,
      "loss": 0.1509,
      "step": 1882
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.782395392033145e-05,
      "loss": 0.1507,
      "step": 1883
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7821408025116255e-05,
      "loss": 0.1594,
      "step": 1884
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7818860823516857e-05,
      "loss": 0.151,
      "step": 1885
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7816312315958706e-05,
      "loss": 0.1976,
      "step": 1886
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7813762502867477e-05,
      "loss": 0.142,
      "step": 1887
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.781121138466905e-05,
      "loss": 0.1785,
      "step": 1888
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7808658961789533e-05,
      "loss": 0.1679,
      "step": 1889
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7806105234655248e-05,
      "loss": 0.1541,
      "step": 1890
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7803550203692738e-05,
      "loss": 0.1777,
      "step": 1891
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7800993869328757e-05,
      "loss": 0.1262,
      "step": 1892
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7798436231990287e-05,
      "loss": 0.1784,
      "step": 1893
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7795877292104517e-05,
      "loss": 0.1538,
      "step": 1894
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7793317050098856e-05,
      "loss": 0.1385,
      "step": 1895
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7790755506400944e-05,
      "loss": 0.1389,
      "step": 1896
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7788192661438618e-05,
      "loss": 0.1446,
      "step": 1897
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.778562851563994e-05,
      "loss": 0.1688,
      "step": 1898
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7783063069433196e-05,
      "loss": 0.1855,
      "step": 1899
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7780496323246884e-05,
      "loss": 0.1752,
      "step": 1900
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7777928277509716e-05,
      "loss": 0.1619,
      "step": 1901
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.777535893265063e-05,
      "loss": 0.1419,
      "step": 1902
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7772788289098766e-05,
      "loss": 0.1579,
      "step": 1903
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7770216347283497e-05,
      "loss": 0.1541,
      "step": 1904
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7767643107634408e-05,
      "loss": 0.15,
      "step": 1905
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.776506857058129e-05,
      "loss": 0.1638,
      "step": 1906
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7762492736554172e-05,
      "loss": 0.1443,
      "step": 1907
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7759915605983275e-05,
      "loss": 0.1417,
      "step": 1908
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.7757337179299055e-05,
      "loss": 0.1932,
      "step": 1909
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7754757456932178e-05,
      "loss": 0.171,
      "step": 1910
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7752176439313523e-05,
      "loss": 0.1673,
      "step": 1911
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7749594126874193e-05,
      "loss": 0.1711,
      "step": 1912
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.77470105200455e-05,
      "loss": 0.1646,
      "step": 1913
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7744425619258982e-05,
      "loss": 0.1999,
      "step": 1914
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7741839424946374e-05,
      "loss": 0.1384,
      "step": 1915
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.773925193753965e-05,
      "loss": 0.1712,
      "step": 1916
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7736663157470992e-05,
      "loss": 0.1547,
      "step": 1917
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7734073085172784e-05,
      "loss": 0.1701,
      "step": 1918
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7731481721077646e-05,
      "loss": 0.144,
      "step": 1919
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7728889065618405e-05,
      "loss": 0.155,
      "step": 1920
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7726295119228096e-05,
      "loss": 0.1817,
      "step": 1921
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7723699882339986e-05,
      "loss": 0.1737,
      "step": 1922
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7721103355387544e-05,
      "loss": 0.1696,
      "step": 1923
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.771850553880446e-05,
      "loss": 0.1512,
      "step": 1924
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7715906433024643e-05,
      "loss": 0.1241,
      "step": 1925
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7713306038482206e-05,
      "loss": 0.1727,
      "step": 1926
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7710704355611493e-05,
      "loss": 0.1685,
      "step": 1927
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7708101384847048e-05,
      "loss": 0.1393,
      "step": 1928
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.770549712662364e-05,
      "loss": 0.1762,
      "step": 1929
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7702891581376244e-05,
      "loss": 0.1724,
      "step": 1930
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7700284749540065e-05,
      "loss": 0.1443,
      "step": 1931
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7697676631550506e-05,
      "loss": 0.1882,
      "step": 1932
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7695067227843195e-05,
      "loss": 0.1257,
      "step": 1933
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7692456538853972e-05,
      "loss": 0.1088,
      "step": 1934
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7689844565018896e-05,
      "loss": 0.1636,
      "step": 1935
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.768723130677423e-05,
      "loss": 0.14,
      "step": 1936
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.768461676455646e-05,
      "loss": 0.1868,
      "step": 1937
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7682000938802283e-05,
      "loss": 0.1524,
      "step": 1938
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7679383829948617e-05,
      "loss": 0.1647,
      "step": 1939
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7676765438432582e-05,
      "loss": 0.1762,
      "step": 1940
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7674145764691526e-05,
      "loss": 0.1926,
      "step": 1941
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7671524809163003e-05,
      "loss": 0.1511,
      "step": 1942
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.766890257228478e-05,
      "loss": 0.1911,
      "step": 1943
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.766627905449484e-05,
      "loss": 0.1672,
      "step": 1944
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7663654256231386e-05,
      "loss": 0.1588,
      "step": 1945
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7661028177932826e-05,
      "loss": 0.1746,
      "step": 1946
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7658400820037784e-05,
      "loss": 0.1608,
      "step": 1947
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.76557721829851e-05,
      "loss": 0.1682,
      "step": 1948
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7653142267213833e-05,
      "loss": 0.1521,
      "step": 1949
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7650511073163236e-05,
      "loss": 0.1617,
      "step": 1950
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.76478786012728e-05,
      "loss": 0.1747,
      "step": 1951
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.764524485198221e-05,
      "loss": 0.1558,
      "step": 1952
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.764260982573138e-05,
      "loss": 0.1561,
      "step": 1953
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7639973522960425e-05,
      "loss": 0.1777,
      "step": 1954
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7637335944109683e-05,
      "loss": 0.1819,
      "step": 1955
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7634697089619695e-05,
      "loss": 0.1892,
      "step": 1956
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.763205695993122e-05,
      "loss": 0.1893,
      "step": 1957
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7629415555485235e-05,
      "loss": 0.1432,
      "step": 1958
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.762677287672292e-05,
      "loss": 0.1695,
      "step": 1959
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7624128924085674e-05,
      "loss": 0.1507,
      "step": 1960
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7621483698015112e-05,
      "loss": 0.1466,
      "step": 1961
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7618837198953053e-05,
      "loss": 0.1539,
      "step": 1962
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7616189427341538e-05,
      "loss": 0.1426,
      "step": 1963
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.761354038362281e-05,
      "loss": 0.1382,
      "step": 1964
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7610890068239333e-05,
      "loss": 0.1592,
      "step": 1965
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.760823848163378e-05,
      "loss": 0.1389,
      "step": 1966
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7605585624249034e-05,
      "loss": 0.1528,
      "step": 1967
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7602931496528198e-05,
      "loss": 0.1831,
      "step": 1968
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.760027609891458e-05,
      "loss": 0.1624,
      "step": 1969
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7597619431851702e-05,
      "loss": 0.1176,
      "step": 1970
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.75949614957833e-05,
      "loss": 0.1465,
      "step": 1971
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7592302291153318e-05,
      "loss": 0.1432,
      "step": 1972
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7589641818405914e-05,
      "loss": 0.2085,
      "step": 1973
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7586980077985464e-05,
      "loss": 0.1748,
      "step": 1974
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7584317070336542e-05,
      "loss": 0.1638,
      "step": 1975
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7581652795903945e-05,
      "loss": 0.1351,
      "step": 1976
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.757898725513268e-05,
      "loss": 0.191,
      "step": 1977
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.757632044846796e-05,
      "loss": 0.1535,
      "step": 1978
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7573652376355217e-05,
      "loss": 0.1653,
      "step": 1979
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7570983039240084e-05,
      "loss": 0.1561,
      "step": 1980
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7568312437568417e-05,
      "loss": 0.1781,
      "step": 1981
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7565640571786277e-05,
      "loss": 0.1924,
      "step": 1982
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7562967442339935e-05,
      "loss": 0.1761,
      "step": 1983
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7560293049675876e-05,
      "loss": 0.1765,
      "step": 1984
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7557617394240795e-05,
      "loss": 0.1465,
      "step": 1985
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7554940476481603e-05,
      "loss": 0.1713,
      "step": 1986
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7552262296845408e-05,
      "loss": 0.155,
      "step": 1987
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.754958285577954e-05,
      "loss": 0.1685,
      "step": 1988
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7546902153731545e-05,
      "loss": 0.1915,
      "step": 1989
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7544220191149165e-05,
      "loss": 0.1493,
      "step": 1990
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7541536968480355e-05,
      "loss": 0.1593,
      "step": 1991
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7538852486173294e-05,
      "loss": 0.1637,
      "step": 1992
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7536166744676358e-05,
      "loss": 0.1653,
      "step": 1993
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.753347974443814e-05,
      "loss": 0.163,
      "step": 1994
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7530791485907437e-05,
      "loss": 0.1653,
      "step": 1995
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7528101969533263e-05,
      "loss": 0.1413,
      "step": 1996
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.752541119576484e-05,
      "loss": 0.1703,
      "step": 1997
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7522719165051593e-05,
      "loss": 0.1394,
      "step": 1998
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.752002587784317e-05,
      "loss": 0.1876,
      "step": 1999
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.751733133458942e-05,
      "loss": 0.1591,
      "step": 2000
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7514635535740403e-05,
      "loss": 0.1602,
      "step": 2001
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.751193848174639e-05,
      "loss": 0.1811,
      "step": 2002
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7509240173057858e-05,
      "loss": 0.1651,
      "step": 2003
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.75065406101255e-05,
      "loss": 0.2181,
      "step": 2004
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.750383979340022e-05,
      "loss": 0.169,
      "step": 2005
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.750113772333312e-05,
      "loss": 0.1788,
      "step": 2006
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.749843440037551e-05,
      "loss": 0.1589,
      "step": 2007
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.749572982497894e-05,
      "loss": 0.1719,
      "step": 2008
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7493023997595123e-05,
      "loss": 0.1714,
      "step": 2009
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7490316918676015e-05,
      "loss": 0.1816,
      "step": 2010
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7487608588673774e-05,
      "loss": 0.1345,
      "step": 2011
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7484899008040758e-05,
      "loss": 0.1847,
      "step": 2012
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.748218817722954e-05,
      "loss": 0.1643,
      "step": 2013
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.74794760966929e-05,
      "loss": 0.1381,
      "step": 2014
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.747676276688383e-05,
      "loss": 0.158,
      "step": 2015
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.747404818825553e-05,
      "loss": 0.1801,
      "step": 2016
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7471332361261403e-05,
      "loss": 0.1676,
      "step": 2017
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.746861528635507e-05,
      "loss": 0.1876,
      "step": 2018
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7465896963990353e-05,
      "loss": 0.1636,
      "step": 2019
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.746317739462128e-05,
      "loss": 0.1361,
      "step": 2020
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.74604565787021e-05,
      "loss": 0.174,
      "step": 2021
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7457734516687255e-05,
      "loss": 0.1263,
      "step": 2022
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7455011209031406e-05,
      "loss": 0.1469,
      "step": 2023
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7452286656189418e-05,
      "loss": 0.139,
      "step": 2024
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.744956085861636e-05,
      "loss": 0.1641,
      "step": 2025
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.744683381676752e-05,
      "loss": 0.1427,
      "step": 2026
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7444105531098377e-05,
      "loss": 0.1686,
      "step": 2027
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.744137600206464e-05,
      "loss": 0.1236,
      "step": 2028
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7438645230122205e-05,
      "loss": 0.1649,
      "step": 2029
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7435913215727185e-05,
      "loss": 0.1458,
      "step": 2030
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7433179959335903e-05,
      "loss": 0.1325,
      "step": 2031
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.743044546140488e-05,
      "loss": 0.1583,
      "step": 2032
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7427709722390854e-05,
      "loss": 0.195,
      "step": 2033
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7424972742750767e-05,
      "loss": 0.1626,
      "step": 2034
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7422234522941767e-05,
      "loss": 0.129,
      "step": 2035
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.7419495063421207e-05,
      "loss": 0.1485,
      "step": 2036
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7416754364646654e-05,
      "loss": 0.1565,
      "step": 2037
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7414012427075878e-05,
      "loss": 0.1454,
      "step": 2038
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.741126925116685e-05,
      "loss": 0.1533,
      "step": 2039
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7408524837377764e-05,
      "loss": 0.1606,
      "step": 2040
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7405779186167e-05,
      "loss": 0.1523,
      "step": 2041
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7403032297993164e-05,
      "loss": 0.1425,
      "step": 2042
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7400284173315053e-05,
      "loss": 0.1569,
      "step": 2043
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.739753481259168e-05,
      "loss": 0.1391,
      "step": 2044
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.739478421628226e-05,
      "loss": 0.1622,
      "step": 2045
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.739203238484622e-05,
      "loss": 0.1771,
      "step": 2046
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7389279318743186e-05,
      "loss": 0.1724,
      "step": 2047
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7386525018432994e-05,
      "loss": 0.1695,
      "step": 2048
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.738376948437569e-05,
      "loss": 0.174,
      "step": 2049
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.738101271703151e-05,
      "loss": 0.1962,
      "step": 2050
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7378254716860924e-05,
      "loss": 0.1727,
      "step": 2051
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.737549548432458e-05,
      "loss": 0.1649,
      "step": 2052
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.737273501988335e-05,
      "loss": 0.1646,
      "step": 2053
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7369973323998302e-05,
      "loss": 0.2016,
      "step": 2054
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.736721039713071e-05,
      "loss": 0.1708,
      "step": 2055
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7364446239742063e-05,
      "loss": 0.1666,
      "step": 2056
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7361680852294047e-05,
      "loss": 0.1808,
      "step": 2057
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7358914235248557e-05,
      "loss": 0.1826,
      "step": 2058
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7356146389067688e-05,
      "loss": 0.1649,
      "step": 2059
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7353377314213746e-05,
      "loss": 0.1687,
      "step": 2060
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7350607011149243e-05,
      "loss": 0.1706,
      "step": 2061
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.734783548033689e-05,
      "loss": 0.1358,
      "step": 2062
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7345062722239607e-05,
      "loss": 0.1526,
      "step": 2063
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7342288737320522e-05,
      "loss": 0.186,
      "step": 2064
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7339513526042962e-05,
      "loss": 0.2059,
      "step": 2065
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7336737088870467e-05,
      "loss": 0.1554,
      "step": 2066
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7333959426266767e-05,
      "loss": 0.1647,
      "step": 2067
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7331180538695807e-05,
      "loss": 0.1667,
      "step": 2068
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7328400426621742e-05,
      "loss": 0.1759,
      "step": 2069
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7325619090508924e-05,
      "loss": 0.1642,
      "step": 2070
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7322836530821905e-05,
      "loss": 0.1175,
      "step": 2071
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7320052748025454e-05,
      "loss": 0.1325,
      "step": 2072
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.731726774258453e-05,
      "loss": 0.1288,
      "step": 2073
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.731448151496431e-05,
      "loss": 0.1623,
      "step": 2074
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7311694065630162e-05,
      "loss": 0.2045,
      "step": 2075
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7308905395047666e-05,
      "loss": 0.1479,
      "step": 2076
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.730611550368261e-05,
      "loss": 0.1388,
      "step": 2077
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7303324392000976e-05,
      "loss": 0.2002,
      "step": 2078
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7300532060468953e-05,
      "loss": 0.1665,
      "step": 2079
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7297738509552937e-05,
      "loss": 0.1771,
      "step": 2080
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7294943739719528e-05,
      "loss": 0.175,
      "step": 2081
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7292147751435523e-05,
      "loss": 0.1397,
      "step": 2082
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.728935054516793e-05,
      "loss": 0.1718,
      "step": 2083
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.728655212138395e-05,
      "loss": 0.1565,
      "step": 2084
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7283752480551008e-05,
      "loss": 0.2082,
      "step": 2085
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.728095162313671e-05,
      "loss": 0.1288,
      "step": 2086
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7278149549608875e-05,
      "loss": 0.1727,
      "step": 2087
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7275346260435522e-05,
      "loss": 0.1632,
      "step": 2088
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7272541756084878e-05,
      "loss": 0.1581,
      "step": 2089
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7269736037025372e-05,
      "loss": 0.1842,
      "step": 2090
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.726692910372563e-05,
      "loss": 0.1793,
      "step": 2091
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.726412095665449e-05,
      "loss": 0.1241,
      "step": 2092
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.726131159628098e-05,
      "loss": 0.1333,
      "step": 2093
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7258501023074347e-05,
      "loss": 0.1728,
      "step": 2094
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7255689237504025e-05,
      "loss": 0.1577,
      "step": 2095
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.725287624003966e-05,
      "loss": 0.1491,
      "step": 2096
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.72500620311511e-05,
      "loss": 0.1382,
      "step": 2097
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7247246611308386e-05,
      "loss": 0.2045,
      "step": 2098
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.724442998098178e-05,
      "loss": 0.1372,
      "step": 2099
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7241612140641724e-05,
      "loss": 0.1488,
      "step": 2100
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7238793090758875e-05,
      "loss": 0.1539,
      "step": 2101
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7235972831804093e-05,
      "loss": 0.1342,
      "step": 2102
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7233151364248435e-05,
      "loss": 0.1509,
      "step": 2103
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.723032868856316e-05,
      "loss": 0.1851,
      "step": 2104
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7227504805219733e-05,
      "loss": 0.1428,
      "step": 2105
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7224679714689813e-05,
      "loss": 0.201,
      "step": 2106
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7221853417445273e-05,
      "loss": 0.1619,
      "step": 2107
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7219025913958176e-05,
      "loss": 0.1874,
      "step": 2108
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.721619720470079e-05,
      "loss": 0.1223,
      "step": 2109
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7213367290145586e-05,
      "loss": 0.1738,
      "step": 2110
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7210536170765236e-05,
      "loss": 0.1366,
      "step": 2111
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.720770384703261e-05,
      "loss": 0.1507,
      "step": 2112
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7204870319420786e-05,
      "loss": 0.1698,
      "step": 2113
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7202035588403036e-05,
      "loss": 0.1653,
      "step": 2114
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7199199654452838e-05,
      "loss": 0.1681,
      "step": 2115
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7196362518043863e-05,
      "loss": 0.1825,
      "step": 2116
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.719352417965e-05,
      "loss": 0.1673,
      "step": 2117
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7190684639745314e-05,
      "loss": 0.185,
      "step": 2118
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.718784389880409e-05,
      "loss": 0.1985,
      "step": 2119
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.718500195730081e-05,
      "loss": 0.1451,
      "step": 2120
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.718215881571015e-05,
      "loss": 0.174,
      "step": 2121
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7179314474506998e-05,
      "loss": 0.1436,
      "step": 2122
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7176468934166427e-05,
      "loss": 0.139,
      "step": 2123
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.717362219516372e-05,
      "loss": 0.2366,
      "step": 2124
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7170774257974364e-05,
      "loss": 0.1462,
      "step": 2125
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7167925123074033e-05,
      "loss": 0.1443,
      "step": 2126
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7165074790938612e-05,
      "loss": 0.1776,
      "step": 2127
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.716222326204418e-05,
      "loss": 0.1773,
      "step": 2128
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7159370536867024e-05,
      "loss": 0.155,
      "step": 2129
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7156516615883624e-05,
      "loss": 0.1668,
      "step": 2130
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7153661499570655e-05,
      "loss": 0.1882,
      "step": 2131
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7150805188405005e-05,
      "loss": 0.1393,
      "step": 2132
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7147947682863753e-05,
      "loss": 0.1997,
      "step": 2133
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7145088983424178e-05,
      "loss": 0.134,
      "step": 2134
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7142229090563756e-05,
      "loss": 0.1529,
      "step": 2135
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.713936800476017e-05,
      "loss": 0.1765,
      "step": 2136
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7136505726491293e-05,
      "loss": 0.1705,
      "step": 2137
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7133642256235208e-05,
      "loss": 0.1368,
      "step": 2138
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7130777594470185e-05,
      "loss": 0.153,
      "step": 2139
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7127911741674706e-05,
      "loss": 0.1395,
      "step": 2140
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.712504469832744e-05,
      "loss": 0.1163,
      "step": 2141
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7122176464907258e-05,
      "loss": 0.166,
      "step": 2142
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.711930704189324e-05,
      "loss": 0.1611,
      "step": 2143
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7116436429764645e-05,
      "loss": 0.2009,
      "step": 2144
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7113564629000954e-05,
      "loss": 0.1453,
      "step": 2145
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.7110691640081826e-05,
      "loss": 0.1794,
      "step": 2146
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.710781746348713e-05,
      "loss": 0.17,
      "step": 2147
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7104942099696928e-05,
      "loss": 0.1595,
      "step": 2148
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7102065549191488e-05,
      "loss": 0.1538,
      "step": 2149
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.709918781245126e-05,
      "loss": 0.1791,
      "step": 2150
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7096308889956917e-05,
      "loss": 0.1608,
      "step": 2151
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7093428782189307e-05,
      "loss": 0.2007,
      "step": 2152
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7090547489629486e-05,
      "loss": 0.1311,
      "step": 2153
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7087665012758708e-05,
      "loss": 0.1813,
      "step": 2154
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7084781352058427e-05,
      "loss": 0.1605,
      "step": 2155
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7081896508010286e-05,
      "loss": 0.1607,
      "step": 2156
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.707901048109613e-05,
      "loss": 0.1728,
      "step": 2157
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.707612327179801e-05,
      "loss": 0.1521,
      "step": 2158
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7073234880598163e-05,
      "loss": 0.1537,
      "step": 2159
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7070345307979026e-05,
      "loss": 0.1573,
      "step": 2160
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7067454554423237e-05,
      "loss": 0.1585,
      "step": 2161
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7064562620413628e-05,
      "loss": 0.162,
      "step": 2162
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.706166950643323e-05,
      "loss": 0.1476,
      "step": 2163
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7058775212965267e-05,
      "loss": 0.1755,
      "step": 2164
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.705587974049317e-05,
      "loss": 0.1878,
      "step": 2165
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7052983089500555e-05,
      "loss": 0.148,
      "step": 2166
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7050085260471244e-05,
      "loss": 0.1688,
      "step": 2167
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.704718625388925e-05,
      "loss": 0.145,
      "step": 2168
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.704428607023878e-05,
      "loss": 0.1606,
      "step": 2169
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.704138471000425e-05,
      "loss": 0.1692,
      "step": 2170
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7038482173670262e-05,
      "loss": 0.1943,
      "step": 2171
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7035578461721613e-05,
      "loss": 0.1164,
      "step": 2172
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7032673574643306e-05,
      "loss": 0.1611,
      "step": 2173
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7029767512920534e-05,
      "loss": 0.1552,
      "step": 2174
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7026860277038685e-05,
      "loss": 0.1804,
      "step": 2175
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7023951867483342e-05,
      "loss": 0.1479,
      "step": 2176
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7021042284740296e-05,
      "loss": 0.17,
      "step": 2177
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7018131529295517e-05,
      "loss": 0.1585,
      "step": 2178
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7015219601635188e-05,
      "loss": 0.1903,
      "step": 2179
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7012306502245665e-05,
      "loss": 0.1475,
      "step": 2180
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7009392231613528e-05,
      "loss": 0.1603,
      "step": 2181
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7006476790225526e-05,
      "loss": 0.1606,
      "step": 2182
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.700356017856862e-05,
      "loss": 0.1525,
      "step": 2183
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7000642397129967e-05,
      "loss": 0.1221,
      "step": 2184
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6997723446396908e-05,
      "loss": 0.1641,
      "step": 2185
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6994803326856994e-05,
      "loss": 0.1703,
      "step": 2186
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6991882038997952e-05,
      "loss": 0.1537,
      "step": 2187
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6988959583307722e-05,
      "loss": 0.1441,
      "step": 2188
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.698603596027443e-05,
      "loss": 0.1651,
      "step": 2189
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.69831111703864e-05,
      "loss": 0.1716,
      "step": 2190
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6980185214132148e-05,
      "loss": 0.1828,
      "step": 2191
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.697725809200039e-05,
      "loss": 0.1723,
      "step": 2192
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6974329804480035e-05,
      "loss": 0.1408,
      "step": 2193
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6971400352060183e-05,
      "loss": 0.1484,
      "step": 2194
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.6968469735230128e-05,
      "loss": 0.1227,
      "step": 2195
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.6965537954479368e-05,
      "loss": 0.1291,
      "step": 2196
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.696260501029758e-05,
      "loss": 0.1706,
      "step": 2197
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.695967090317465e-05,
      "loss": 0.1661,
      "step": 2198
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.695673563360065e-05,
      "loss": 0.1718,
      "step": 2199
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.6953799202065856e-05,
      "loss": 0.163,
      "step": 2200
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 7925,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 200,
  "total_flos": 2.955774406372098e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}