{
  "best_metric": 0.35805782675743103,
  "best_model_checkpoint": "./medium-article-titles-engagement/checkpoint-4258",
  "epoch": 2.0,
  "global_step": 4258,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.6818,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.6986,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 3e-06,
      "loss": 0.701,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.7021,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 5e-06,
      "loss": 0.6901,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 6e-06,
      "loss": 0.6941,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.6989,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.7072,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 9e-06,
      "loss": 0.6891,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 1e-05,
      "loss": 0.6904,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.6864,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2e-05,
      "loss": 0.6989,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.7111,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.6903,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5e-05,
      "loss": 0.6937,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.6822,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.6945,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8e-05,
      "loss": 0.6798,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9e-05,
      "loss": 0.6981,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 2e-05,
      "loss": 0.6893,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.1e-05,
      "loss": 0.6848,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.6884,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.6881,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4e-05,
      "loss": 0.6917,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.5e-05,
      "loss": 0.6933,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.6853,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.6998,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.673,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.9e-05,
      "loss": 0.6826,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 3e-05,
      "loss": 0.6645,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.1e-05,
      "loss": 0.6649,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.6866,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.3e-05,
      "loss": 0.7105,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.6383,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.5e-05,
      "loss": 0.6576,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.6e-05,
      "loss": 0.6313,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.7e-05,
      "loss": 0.6826,
      "step": 37
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8e-05,
      "loss": 0.6783,
      "step": 38
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.6609,
      "step": 39
    },
    {
      "epoch": 0.02,
      "learning_rate": 4e-05,
      "loss": 0.6377,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.1e-05,
      "loss": 0.7075,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.2e-05,
      "loss": 0.6811,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.3e-05,
      "loss": 0.6503,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.6835,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.5e-05,
      "loss": 0.6662,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.6585,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.7e-05,
      "loss": 0.6833,
      "step": 47
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.8e-05,
      "loss": 0.652,
      "step": 48
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9e-05,
      "loss": 0.6244,
      "step": 49
    },
    {
      "epoch": 0.02,
      "learning_rate": 5e-05,
      "loss": 0.7156,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.999764595103578e-05,
      "loss": 0.6766,
      "step": 51
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.999529190207157e-05,
      "loss": 0.6881,
      "step": 52
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.999293785310735e-05,
      "loss": 0.6303,
      "step": 53
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9990583804143125e-05,
      "loss": 0.6781,
      "step": 54
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.998822975517891e-05,
      "loss": 0.6836,
      "step": 55
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.998587570621469e-05,
      "loss": 0.6516,
      "step": 56
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.998352165725047e-05,
      "loss": 0.6276,
      "step": 57
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9981167608286253e-05,
      "loss": 0.6594,
      "step": 58
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.997881355932204e-05,
      "loss": 0.6584,
      "step": 59
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.997645951035782e-05,
      "loss": 0.6945,
      "step": 60
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.99741054613936e-05,
      "loss": 0.7133,
      "step": 61
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.997175141242938e-05,
      "loss": 0.6994,
      "step": 62
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9969397363465166e-05,
      "loss": 0.6234,
      "step": 63
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9967043314500943e-05,
      "loss": 0.7266,
      "step": 64
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.996468926553672e-05,
      "loss": 0.678,
      "step": 65
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.996233521657251e-05,
      "loss": 0.684,
      "step": 66
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.995998116760829e-05,
      "loss": 0.6161,
      "step": 67
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9957627118644066e-05,
      "loss": 0.6386,
      "step": 68
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9955273069679856e-05,
      "loss": 0.6564,
      "step": 69
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9952919020715633e-05,
      "loss": 0.71,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.995056497175141e-05,
      "loss": 0.6855,
      "step": 71
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9948210922787195e-05,
      "loss": 0.6838,
      "step": 72
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.994585687382298e-05,
      "loss": 0.6483,
      "step": 73
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.994350282485876e-05,
      "loss": 0.7043,
      "step": 74
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.994114877589454e-05,
      "loss": 0.6453,
      "step": 75
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9938794726930323e-05,
      "loss": 0.6593,
      "step": 76
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.993644067796611e-05,
      "loss": 0.6753,
      "step": 77
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9934086629001885e-05,
      "loss": 0.783,
      "step": 78
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.993173258003767e-05,
      "loss": 0.661,
      "step": 79
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.992937853107345e-05,
      "loss": 0.6719,
      "step": 80
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.992702448210923e-05,
      "loss": 0.7038,
      "step": 81
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.992467043314501e-05,
      "loss": 0.7013,
      "step": 82
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.99223163841808e-05,
      "loss": 0.7031,
      "step": 83
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9919962335216575e-05,
      "loss": 0.6391,
      "step": 84
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.991760828625235e-05,
      "loss": 0.7058,
      "step": 85
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.991525423728814e-05,
      "loss": 0.7178,
      "step": 86
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.991290018832392e-05,
      "loss": 0.6847,
      "step": 87
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.99105461393597e-05,
      "loss": 0.6623,
      "step": 88
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.990819209039548e-05,
      "loss": 0.6607,
      "step": 89
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9905838041431265e-05,
      "loss": 0.6847,
      "step": 90
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.990348399246705e-05,
      "loss": 0.6691,
      "step": 91
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9901129943502826e-05,
      "loss": 0.649,
      "step": 92
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.989877589453861e-05,
      "loss": 0.7197,
      "step": 93
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9896421845574393e-05,
      "loss": 0.6891,
      "step": 94
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.989406779661017e-05,
      "loss": 0.648,
      "step": 95
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9891713747645955e-05,
      "loss": 0.6452,
      "step": 96
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.988935969868174e-05,
      "loss": 0.6559,
      "step": 97
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9887005649717516e-05,
      "loss": 0.6436,
      "step": 98
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.988465160075329e-05,
      "loss": 0.6379,
      "step": 99
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9882297551789083e-05,
      "loss": 0.6937,
      "step": 100
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.987994350282486e-05,
      "loss": 0.613,
      "step": 101
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.987758945386064e-05,
      "loss": 0.6465,
      "step": 102
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.987523540489643e-05,
      "loss": 0.6608,
      "step": 103
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9872881355932206e-05,
      "loss": 0.6617,
      "step": 104
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.987052730696798e-05,
      "loss": 0.6937,
      "step": 105
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.986817325800377e-05,
      "loss": 0.6472,
      "step": 106
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.986581920903955e-05,
      "loss": 0.6862,
      "step": 107
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9863465160075334e-05,
      "loss": 0.6754,
      "step": 108
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.986111111111111e-05,
      "loss": 0.6848,
      "step": 109
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9858757062146896e-05,
      "loss": 0.6845,
      "step": 110
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.985640301318268e-05,
      "loss": 0.631,
      "step": 111
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.985404896421846e-05,
      "loss": 0.5851,
      "step": 112
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.985169491525424e-05,
      "loss": 0.6098,
      "step": 113
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9849340866290024e-05,
      "loss": 0.7105,
      "step": 114
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.98469868173258e-05,
      "loss": 0.6077,
      "step": 115
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.984463276836158e-05,
      "loss": 0.7026,
      "step": 116
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.984227871939737e-05,
      "loss": 0.6297,
      "step": 117
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.983992467043315e-05,
      "loss": 0.7316,
      "step": 118
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9837570621468924e-05,
      "loss": 0.6758,
      "step": 119
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9835216572504714e-05,
      "loss": 0.6324,
      "step": 120
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.983286252354049e-05,
      "loss": 0.719,
      "step": 121
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9830508474576276e-05,
      "loss": 0.6764,
      "step": 122
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.982815442561205e-05,
      "loss": 0.7366,
      "step": 123
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.982580037664784e-05,
      "loss": 0.6851,
      "step": 124
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.982344632768362e-05,
      "loss": 0.6576,
      "step": 125
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.98210922787194e-05,
      "loss": 0.7361,
      "step": 126
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.981873822975518e-05,
      "loss": 0.6496,
      "step": 127
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9816384180790966e-05,
      "loss": 0.68,
      "step": 128
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.981403013182674e-05,
      "loss": 0.6689,
      "step": 129
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.981167608286253e-05,
      "loss": 0.6733,
      "step": 130
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.980932203389831e-05,
      "loss": 0.6602,
      "step": 131
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.980696798493409e-05,
      "loss": 0.6509,
      "step": 132
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.980461393596987e-05,
      "loss": 0.6294,
      "step": 133
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9802259887005656e-05,
      "loss": 0.6351,
      "step": 134
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.979990583804143e-05,
      "loss": 0.6601,
      "step": 135
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.979755178907721e-05,
      "loss": 0.687,
      "step": 136
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9795197740113e-05,
      "loss": 0.6701,
      "step": 137
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.979284369114878e-05,
      "loss": 0.6944,
      "step": 138
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.979048964218456e-05,
      "loss": 0.6964,
      "step": 139
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.978813559322034e-05,
      "loss": 0.625,
      "step": 140
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.978578154425612e-05,
      "loss": 0.6573,
      "step": 141
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9783427495291907e-05,
      "loss": 0.6693,
      "step": 142
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9781073446327684e-05,
      "loss": 0.7238,
      "step": 143
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.977871939736347e-05,
      "loss": 0.6496,
      "step": 144
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.977636534839925e-05,
      "loss": 0.6629,
      "step": 145
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.977401129943503e-05,
      "loss": 0.6345,
      "step": 146
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.977165725047081e-05,
      "loss": 0.6223,
      "step": 147
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9769303201506597e-05,
      "loss": 0.6877,
      "step": 148
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9766949152542374e-05,
      "loss": 0.6346,
      "step": 149
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.976459510357816e-05,
      "loss": 0.6589,
      "step": 150
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.976224105461394e-05,
      "loss": 0.6047,
      "step": 151
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.975988700564972e-05,
      "loss": 0.7097,
      "step": 152
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.97575329566855e-05,
      "loss": 0.6549,
      "step": 153
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9755178907721287e-05,
      "loss": 0.6175,
      "step": 154
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9752824858757064e-05,
      "loss": 0.6594,
      "step": 155
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.975047080979285e-05,
      "loss": 0.6309,
      "step": 156
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9748116760828625e-05,
      "loss": 0.6301,
      "step": 157
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.974576271186441e-05,
      "loss": 0.6198,
      "step": 158
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.974340866290019e-05,
      "loss": 0.6832,
      "step": 159
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.974105461393597e-05,
      "loss": 0.7371,
      "step": 160
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9738700564971754e-05,
      "loss": 0.5944,
      "step": 161
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.973634651600754e-05,
      "loss": 0.6685,
      "step": 162
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9733992467043315e-05,
      "loss": 0.665,
      "step": 163
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.97316384180791e-05,
      "loss": 0.5687,
      "step": 164
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.972928436911488e-05,
      "loss": 0.6496,
      "step": 165
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.972693032015066e-05,
      "loss": 0.6105,
      "step": 166
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9724576271186444e-05,
      "loss": 0.6612,
      "step": 167
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.972222222222223e-05,
      "loss": 0.7072,
      "step": 168
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9719868173258005e-05,
      "loss": 0.6327,
      "step": 169
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.971751412429379e-05,
      "loss": 0.5923,
      "step": 170
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.971516007532957e-05,
      "loss": 0.5961,
      "step": 171
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.971280602636535e-05,
      "loss": 0.6435,
      "step": 172
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9710451977401134e-05,
      "loss": 0.6862,
      "step": 173
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.970809792843691e-05,
      "loss": 0.6389,
      "step": 174
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9705743879472695e-05,
      "loss": 0.5724,
      "step": 175
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.970338983050848e-05,
      "loss": 0.6799,
      "step": 176
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9701035781544256e-05,
      "loss": 0.6455,
      "step": 177
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.969868173258004e-05,
      "loss": 0.6987,
      "step": 178
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9696327683615824e-05,
      "loss": 0.7139,
      "step": 179
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.96939736346516e-05,
      "loss": 0.6064,
      "step": 180
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9691619585687385e-05,
      "loss": 0.6348,
      "step": 181
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.968926553672317e-05,
      "loss": 0.7095,
      "step": 182
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9686911487758946e-05,
      "loss": 0.6013,
      "step": 183
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.968455743879473e-05,
      "loss": 0.6665,
      "step": 184
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9682203389830514e-05,
      "loss": 0.6369,
      "step": 185
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.967984934086629e-05,
      "loss": 0.6243,
      "step": 186
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9677495291902075e-05,
      "loss": 0.6776,
      "step": 187
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.967514124293786e-05,
      "loss": 0.6432,
      "step": 188
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9672787193973636e-05,
      "loss": 0.635,
      "step": 189
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.967043314500942e-05,
      "loss": 0.6561,
      "step": 190
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.96680790960452e-05,
      "loss": 0.6886,
      "step": 191
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.966572504708098e-05,
      "loss": 0.6533,
      "step": 192
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9663370998116765e-05,
      "loss": 0.6599,
      "step": 193
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.966101694915254e-05,
      "loss": 0.5998,
      "step": 194
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9658662900188326e-05,
      "loss": 0.6507,
      "step": 195
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.965630885122411e-05,
      "loss": 0.6565,
      "step": 196
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.965395480225989e-05,
      "loss": 0.6786,
      "step": 197
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.965160075329567e-05,
      "loss": 0.6513,
      "step": 198
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9649246704331455e-05,
      "loss": 0.6497,
      "step": 199
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.964689265536723e-05,
      "loss": 0.5744,
      "step": 200
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9644538606403016e-05,
      "loss": 0.6643,
      "step": 201
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.96421845574388e-05,
      "loss": 0.6969,
      "step": 202
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.963983050847458e-05,
      "loss": 0.6324,
      "step": 203
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.963747645951036e-05,
      "loss": 0.6223,
      "step": 204
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9635122410546145e-05,
      "loss": 0.6021,
      "step": 205
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.963276836158192e-05,
      "loss": 0.6978,
      "step": 206
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9630414312617706e-05,
      "loss": 0.6619,
      "step": 207
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.962806026365348e-05,
      "loss": 0.6476,
      "step": 208
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.962570621468927e-05,
      "loss": 0.6361,
      "step": 209
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.962335216572505e-05,
      "loss": 0.6816,
      "step": 210
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.962099811676083e-05,
      "loss": 0.6914,
      "step": 211
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.961864406779661e-05,
      "loss": 0.697,
      "step": 212
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9616290018832396e-05,
      "loss": 0.6442,
      "step": 213
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.961393596986817e-05,
      "loss": 0.603,
      "step": 214
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.961158192090396e-05,
      "loss": 0.6668,
      "step": 215
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.960922787193974e-05,
      "loss": 0.6624,
      "step": 216
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.960687382297552e-05,
      "loss": 0.6418,
      "step": 217
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.96045197740113e-05,
      "loss": 0.6617,
      "step": 218
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9602165725047086e-05,
      "loss": 0.6046,
      "step": 219
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.959981167608286e-05,
      "loss": 0.6389,
      "step": 220
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.959745762711865e-05,
      "loss": 0.6413,
      "step": 221
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.959510357815443e-05,
      "loss": 0.6584,
      "step": 222
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.959274952919021e-05,
      "loss": 0.6653,
      "step": 223
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.959039548022599e-05,
      "loss": 0.6143,
      "step": 224
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9588041431261776e-05,
      "loss": 0.641,
      "step": 225
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.958568738229755e-05,
      "loss": 0.6501,
      "step": 226
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.958333333333334e-05,
      "loss": 0.6718,
      "step": 227
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9580979284369114e-05,
      "loss": 0.6535,
      "step": 228
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.95786252354049e-05,
      "loss": 0.6758,
      "step": 229
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.957627118644068e-05,
      "loss": 0.6658,
      "step": 230
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.957391713747646e-05,
      "loss": 0.6803,
      "step": 231
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.957156308851224e-05,
      "loss": 0.6456,
      "step": 232
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.956920903954803e-05,
      "loss": 0.6537,
      "step": 233
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9566854990583804e-05,
      "loss": 0.6759,
      "step": 234
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.956450094161959e-05,
      "loss": 0.6299,
      "step": 235
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.956214689265537e-05,
      "loss": 0.6273,
      "step": 236
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.955979284369115e-05,
      "loss": 0.6789,
      "step": 237
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.955743879472693e-05,
      "loss": 0.6184,
      "step": 238
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.955508474576272e-05,
      "loss": 0.6668,
      "step": 239
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9552730696798494e-05,
      "loss": 0.5536,
      "step": 240
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.955037664783428e-05,
      "loss": 0.5988,
      "step": 241
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.954802259887006e-05,
      "loss": 0.6694,
      "step": 242
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.954566854990584e-05,
      "loss": 0.7239,
      "step": 243
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.954331450094162e-05,
      "loss": 0.617,
      "step": 244
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.95409604519774e-05,
      "loss": 0.5939,
      "step": 245
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9538606403013184e-05,
      "loss": 0.6055,
      "step": 246
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.953625235404897e-05,
      "loss": 0.6361,
      "step": 247
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9533898305084745e-05,
      "loss": 0.6049,
      "step": 248
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.953154425612053e-05,
      "loss": 0.6858,
      "step": 249
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.952919020715631e-05,
      "loss": 0.6994,
      "step": 250
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.952683615819209e-05,
      "loss": 0.6484,
      "step": 251
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9524482109227874e-05,
      "loss": 0.6407,
      "step": 252
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.952212806026366e-05,
      "loss": 0.6611,
      "step": 253
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9519774011299435e-05,
      "loss": 0.6912,
      "step": 254
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.951741996233522e-05,
      "loss": 0.6912,
      "step": 255
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9515065913371e-05,
      "loss": 0.6637,
      "step": 256
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.951271186440678e-05,
      "loss": 0.7289,
      "step": 257
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9510357815442564e-05,
      "loss": 0.6816,
      "step": 258
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.950800376647835e-05,
      "loss": 0.6343,
      "step": 259
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9505649717514125e-05,
      "loss": 0.6059,
      "step": 260
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.950329566854991e-05,
      "loss": 0.6902,
      "step": 261
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9500941619585686e-05,
      "loss": 0.6675,
      "step": 262
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.949858757062147e-05,
      "loss": 0.647,
      "step": 263
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9496233521657254e-05,
      "loss": 0.6575,
      "step": 264
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.949387947269303e-05,
      "loss": 0.6798,
      "step": 265
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9491525423728815e-05,
      "loss": 0.636,
      "step": 266
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.94891713747646e-05,
      "loss": 0.6633,
      "step": 267
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9486817325800376e-05,
      "loss": 0.6894,
      "step": 268
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.948446327683616e-05,
      "loss": 0.6343,
      "step": 269
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9482109227871944e-05,
      "loss": 0.6325,
      "step": 270
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.947975517890772e-05,
      "loss": 0.6715,
      "step": 271
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9477401129943505e-05,
      "loss": 0.6327,
      "step": 272
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.947504708097929e-05,
      "loss": 0.6697,
      "step": 273
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9472693032015066e-05,
      "loss": 0.6723,
      "step": 274
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.947033898305085e-05,
      "loss": 0.6673,
      "step": 275
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9467984934086634e-05,
      "loss": 0.7047,
      "step": 276
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.946563088512241e-05,
      "loss": 0.6927,
      "step": 277
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9463276836158195e-05,
      "loss": 0.6481,
      "step": 278
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.946092278719397e-05,
      "loss": 0.6561,
      "step": 279
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.945856873822976e-05,
      "loss": 0.7036,
      "step": 280
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.945621468926554e-05,
      "loss": 0.6105,
      "step": 281
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.945386064030132e-05,
      "loss": 0.7092,
      "step": 282
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.94515065913371e-05,
      "loss": 0.6625,
      "step": 283
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9449152542372885e-05,
      "loss": 0.6624,
      "step": 284
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.944679849340866e-05,
      "loss": 0.6214,
      "step": 285
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9444444444444446e-05,
      "loss": 0.6551,
      "step": 286
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.944209039548023e-05,
      "loss": 0.6794,
      "step": 287
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.943973634651601e-05,
      "loss": 0.7065,
      "step": 288
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.943738229755179e-05,
      "loss": 0.615,
      "step": 289
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9435028248587575e-05,
      "loss": 0.7009,
      "step": 290
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.943267419962335e-05,
      "loss": 0.6497,
      "step": 291
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9430320150659136e-05,
      "loss": 0.6805,
      "step": 292
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.942796610169492e-05,
      "loss": 0.6581,
      "step": 293
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.94256120527307e-05,
      "loss": 0.6512,
      "step": 294
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.942325800376648e-05,
      "loss": 0.6609,
      "step": 295
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.942090395480226e-05,
      "loss": 0.6107,
      "step": 296
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.941854990583805e-05,
      "loss": 0.6746,
      "step": 297
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9416195856873826e-05,
      "loss": 0.6677,
      "step": 298
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.94138418079096e-05,
      "loss": 0.6349,
      "step": 299
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9411487758945394e-05,
      "loss": 0.6227,
      "step": 300
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.940913370998117e-05,
      "loss": 0.6089,
      "step": 301
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.940677966101695e-05,
      "loss": 0.6874,
      "step": 302
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.940442561205273e-05,
      "loss": 0.6468,
      "step": 303
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9402071563088516e-05,
      "loss": 0.6929,
      "step": 304
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.939971751412429e-05,
      "loss": 0.6779,
      "step": 305
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.939736346516008e-05,
      "loss": 0.6786,
      "step": 306
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.939500941619586e-05,
      "loss": 0.592,
      "step": 307
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.939265536723164e-05,
      "loss": 0.6781,
      "step": 308
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.939030131826742e-05,
      "loss": 0.5925,
      "step": 309
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9387947269303206e-05,
      "loss": 0.6867,
      "step": 310
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.938559322033899e-05,
      "loss": 0.6762,
      "step": 311
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.938323917137477e-05,
      "loss": 0.7195,
      "step": 312
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9380885122410544e-05,
      "loss": 0.5524,
      "step": 313
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9378531073446335e-05,
      "loss": 0.7096,
      "step": 314
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.937617702448211e-05,
      "loss": 0.6225,
      "step": 315
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.937382297551789e-05,
      "loss": 0.6398,
      "step": 316
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.937146892655368e-05,
      "loss": 0.6001,
      "step": 317
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.936911487758946e-05,
      "loss": 0.6783,
      "step": 318
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9366760828625234e-05,
      "loss": 0.6268,
      "step": 319
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.936440677966102e-05,
      "loss": 0.6048,
      "step": 320
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.93620527306968e-05,
      "loss": 0.6385,
      "step": 321
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.935969868173258e-05,
      "loss": 0.6224,
      "step": 322
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.935734463276836e-05,
      "loss": 0.6818,
      "step": 323
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.935499058380415e-05,
      "loss": 0.6835,
      "step": 324
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9352636534839924e-05,
      "loss": 0.6548,
      "step": 325
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.935028248587571e-05,
      "loss": 0.6015,
      "step": 326
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.934792843691149e-05,
      "loss": 0.701,
      "step": 327
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9345574387947276e-05,
      "loss": 0.5609,
      "step": 328
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.934322033898305e-05,
      "loss": 0.5806,
      "step": 329
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.934086629001883e-05,
      "loss": 0.683,
      "step": 330
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.933851224105462e-05,
      "loss": 0.5861,
      "step": 331
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.93361581920904e-05,
      "loss": 0.6397,
      "step": 332
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9333804143126175e-05,
      "loss": 0.6207,
      "step": 333
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9331450094161966e-05,
      "loss": 0.6493,
      "step": 334
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.932909604519774e-05,
      "loss": 0.6784,
      "step": 335
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.932674199623352e-05,
      "loss": 0.6625,
      "step": 336
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9324387947269304e-05,
      "loss": 0.6183,
      "step": 337
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.932203389830509e-05,
      "loss": 0.6148,
      "step": 338
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9319679849340865e-05,
      "loss": 0.659,
      "step": 339
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.931732580037665e-05,
      "loss": 0.669,
      "step": 340
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.931497175141243e-05,
      "loss": 0.6191,
      "step": 341
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.931261770244821e-05,
      "loss": 0.7129,
      "step": 342
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9310263653483994e-05,
      "loss": 0.58,
      "step": 343
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.930790960451978e-05,
      "loss": 0.6315,
      "step": 344
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.930555555555556e-05,
      "loss": 0.687,
      "step": 345
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.930320150659134e-05,
      "loss": 0.6024,
      "step": 346
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9300847457627116e-05,
      "loss": 0.5917,
      "step": 347
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.929849340866291e-05,
      "loss": 0.6047,
      "step": 348
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9296139359698684e-05,
      "loss": 0.7005,
      "step": 349
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.929378531073446e-05,
      "loss": 0.6617,
      "step": 350
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.929143126177025e-05,
      "loss": 0.6486,
      "step": 351
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.928907721280603e-05,
      "loss": 0.685,
      "step": 352
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9286723163841806e-05,
      "loss": 0.6286,
      "step": 353
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.928436911487759e-05,
      "loss": 0.645,
      "step": 354
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9282015065913374e-05,
      "loss": 0.6364,
      "step": 355
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.927966101694915e-05,
      "loss": 0.6552,
      "step": 356
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9277306967984935e-05,
      "loss": 0.6263,
      "step": 357
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.927495291902072e-05,
      "loss": 0.6486,
      "step": 358
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.92725988700565e-05,
      "loss": 0.6295,
      "step": 359
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.927024482109228e-05,
      "loss": 0.6131,
      "step": 360
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9267890772128064e-05,
      "loss": 0.5999,
      "step": 361
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.926553672316385e-05,
      "loss": 0.6821,
      "step": 362
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9263182674199625e-05,
      "loss": 0.632,
      "step": 363
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.92608286252354e-05,
      "loss": 0.7303,
      "step": 364
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.925847457627119e-05,
      "loss": 0.6628,
      "step": 365
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.925612052730697e-05,
      "loss": 0.649,
      "step": 366
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.925376647834275e-05,
      "loss": 0.638,
      "step": 367
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.925141242937854e-05,
      "loss": 0.649,
      "step": 368
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9249058380414315e-05,
      "loss": 0.6283,
      "step": 369
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.924670433145009e-05,
      "loss": 0.6301,
      "step": 370
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.9244350282485876e-05,
      "loss": 0.6963,
      "step": 371
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.924199623352166e-05,
      "loss": 0.6623,
      "step": 372
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.923964218455744e-05,
      "loss": 0.6283,
      "step": 373
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.923728813559322e-05,
      "loss": 0.6544,
      "step": 374
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9234934086629005e-05,
      "loss": 0.6156,
      "step": 375
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.923258003766479e-05,
      "loss": 0.6522,
      "step": 376
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9230225988700566e-05,
      "loss": 0.6009,
      "step": 377
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.922787193973635e-05,
      "loss": 0.6749,
      "step": 378
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9225517890772134e-05,
      "loss": 0.6109,
      "step": 379
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.922316384180791e-05,
      "loss": 0.6607,
      "step": 380
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.922080979284369e-05,
      "loss": 0.6224,
      "step": 381
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.921845574387948e-05,
      "loss": 0.6427,
      "step": 382
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9216101694915256e-05,
      "loss": 0.6386,
      "step": 383
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9213747645951034e-05,
      "loss": 0.6477,
      "step": 384
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9211393596986824e-05,
      "loss": 0.5976,
      "step": 385
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.92090395480226e-05,
      "loss": 0.6395,
      "step": 386
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.920668549905838e-05,
      "loss": 0.5749,
      "step": 387
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.920433145009416e-05,
      "loss": 0.6893,
      "step": 388
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9201977401129946e-05,
      "loss": 0.6841,
      "step": 389
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.919962335216573e-05,
      "loss": 0.664,
      "step": 390
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.919726930320151e-05,
      "loss": 0.6352,
      "step": 391
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.919491525423729e-05,
      "loss": 0.6586,
      "step": 392
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.9192561205273075e-05,
      "loss": 0.6588,
      "step": 393
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.919020715630885e-05,
      "loss": 0.6334,
      "step": 394
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9187853107344636e-05,
      "loss": 0.6313,
      "step": 395
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.918549905838042e-05,
      "loss": 0.6282,
      "step": 396
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.91831450094162e-05,
      "loss": 0.5805,
      "step": 397
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9180790960451975e-05,
      "loss": 0.6239,
      "step": 398
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9178436911487765e-05,
      "loss": 0.6273,
      "step": 399
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.917608286252354e-05,
      "loss": 0.6454,
      "step": 400
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.917372881355932e-05,
      "loss": 0.6149,
      "step": 401
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.917137476459511e-05,
      "loss": 0.5899,
      "step": 402
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.916902071563089e-05,
      "loss": 0.5464,
      "step": 403
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9166666666666665e-05,
      "loss": 0.7368,
      "step": 404
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.916431261770245e-05,
      "loss": 0.6547,
      "step": 405
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.916195856873823e-05,
      "loss": 0.6531,
      "step": 406
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9159604519774016e-05,
      "loss": 0.6269,
      "step": 407
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9157250470809794e-05,
      "loss": 0.6672,
      "step": 408
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.915489642184558e-05,
      "loss": 0.5971,
      "step": 409
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.915254237288136e-05,
      "loss": 0.6463,
      "step": 410
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.915018832391714e-05,
      "loss": 0.7032,
      "step": 411
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.914783427495292e-05,
      "loss": 0.6895,
      "step": 412
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9145480225988706e-05,
      "loss": 0.6581,
      "step": 413
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.9143126177024484e-05,
      "loss": 0.6225,
      "step": 414
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.914077212806026e-05,
      "loss": 0.7009,
      "step": 415
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.913841807909605e-05,
      "loss": 0.5491,
      "step": 416
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.913606403013183e-05,
      "loss": 0.6789,
      "step": 417
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9133709981167606e-05,
      "loss": 0.6756,
      "step": 418
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9131355932203396e-05,
      "loss": 0.7268,
      "step": 419
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9129001883239174e-05,
      "loss": 0.6167,
      "step": 420
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.912664783427495e-05,
      "loss": 0.602,
      "step": 421
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9124293785310735e-05,
      "loss": 0.6567,
      "step": 422
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.912193973634652e-05,
      "loss": 0.6435,
      "step": 423
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.91195856873823e-05,
      "loss": 0.6507,
      "step": 424
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.911723163841808e-05,
      "loss": 0.6893,
      "step": 425
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9114877589453864e-05,
      "loss": 0.6929,
      "step": 426
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.911252354048965e-05,
      "loss": 0.6955,
      "step": 427
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9110169491525425e-05,
      "loss": 0.6161,
      "step": 428
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.910781544256121e-05,
      "loss": 0.5958,
      "step": 429
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.910546139359699e-05,
      "loss": 0.6269,
      "step": 430
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.910310734463277e-05,
      "loss": 0.6261,
      "step": 431
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.910075329566855e-05,
      "loss": 0.6365,
      "step": 432
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.909839924670434e-05,
      "loss": 0.6631,
      "step": 433
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9096045197740115e-05,
      "loss": 0.6346,
      "step": 434
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.909369114877589e-05,
      "loss": 0.698,
      "step": 435
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.909133709981168e-05,
      "loss": 0.6358,
      "step": 436
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.908898305084746e-05,
      "loss": 0.6212,
      "step": 437
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9086629001883244e-05,
      "loss": 0.71,
      "step": 438
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.908427495291902e-05,
      "loss": 0.7036,
      "step": 439
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9081920903954805e-05,
      "loss": 0.7036,
      "step": 440
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.907956685499059e-05,
      "loss": 0.6116,
      "step": 441
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9077212806026366e-05,
      "loss": 0.6079,
      "step": 442
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.907485875706215e-05,
      "loss": 0.5471,
      "step": 443
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9072504708097934e-05,
      "loss": 0.647,
      "step": 444
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.907015065913371e-05,
      "loss": 0.6526,
      "step": 445
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9067796610169495e-05,
      "loss": 0.6373,
      "step": 446
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.906544256120528e-05,
      "loss": 0.6455,
      "step": 447
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9063088512241056e-05,
      "loss": 0.6393,
      "step": 448
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.906073446327684e-05,
      "loss": 0.6509,
      "step": 449
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9058380414312623e-05,
      "loss": 0.6479,
      "step": 450
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.90560263653484e-05,
      "loss": 0.653,
      "step": 451
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.905367231638418e-05,
      "loss": 0.6449,
      "step": 452
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.905131826741997e-05,
      "loss": 0.6267,
      "step": 453
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9048964218455746e-05,
      "loss": 0.621,
      "step": 454
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.904661016949153e-05,
      "loss": 0.5811,
      "step": 455
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.904425612052731e-05,
      "loss": 0.5734,
      "step": 456
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.904190207156309e-05,
      "loss": 0.65,
      "step": 457
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9039548022598875e-05,
      "loss": 0.6375,
      "step": 458
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.903719397363465e-05,
      "loss": 0.5783,
      "step": 459
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9034839924670436e-05,
      "loss": 0.614,
      "step": 460
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.903248587570622e-05,
      "loss": 0.6243,
      "step": 461
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9030131826742e-05,
      "loss": 0.6038,
      "step": 462
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.902777777777778e-05,
      "loss": 0.6104,
      "step": 463
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9025423728813565e-05,
      "loss": 0.6617,
      "step": 464
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.902306967984934e-05,
      "loss": 0.6496,
      "step": 465
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9020715630885126e-05,
      "loss": 0.64,
      "step": 466
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.901836158192091e-05,
      "loss": 0.6478,
      "step": 467
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.901600753295669e-05,
      "loss": 0.603,
      "step": 468
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.901365348399247e-05,
      "loss": 0.6536,
      "step": 469
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9011299435028255e-05,
      "loss": 0.6057,
      "step": 470
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.900894538606403e-05,
      "loss": 0.6066,
      "step": 471
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9006591337099816e-05,
      "loss": 0.6752,
      "step": 472
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.900423728813559e-05,
      "loss": 0.6047,
      "step": 473
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.900188323917138e-05,
      "loss": 0.621,
      "step": 474
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.899952919020716e-05,
      "loss": 0.6663,
      "step": 475
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.899717514124294e-05,
      "loss": 0.6173,
      "step": 476
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.899482109227872e-05,
      "loss": 0.6041,
      "step": 477
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.8992467043314506e-05,
      "loss": 0.6176,
      "step": 478
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.899011299435028e-05,
      "loss": 0.7024,
      "step": 479
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.898775894538607e-05,
      "loss": 0.679,
      "step": 480
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.898540489642185e-05,
      "loss": 0.6845,
      "step": 481
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.898305084745763e-05,
      "loss": 0.6587,
      "step": 482
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.898069679849341e-05,
      "loss": 0.6679,
      "step": 483
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8978342749529196e-05,
      "loss": 0.6254,
      "step": 484
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.897598870056497e-05,
      "loss": 0.6678,
      "step": 485
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.897363465160076e-05,
      "loss": 0.6302,
      "step": 486
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.897128060263654e-05,
      "loss": 0.653,
      "step": 487
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.896892655367232e-05,
      "loss": 0.6054,
      "step": 488
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.89665725047081e-05,
      "loss": 0.581,
      "step": 489
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.896421845574388e-05,
      "loss": 0.6714,
      "step": 490
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.896186440677966e-05,
      "loss": 0.608,
      "step": 491
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.895951035781545e-05,
      "loss": 0.6479,
      "step": 492
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8957156308851224e-05,
      "loss": 0.6083,
      "step": 493
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.895480225988701e-05,
      "loss": 0.6047,
      "step": 494
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.895244821092279e-05,
      "loss": 0.7008,
      "step": 495
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.895009416195857e-05,
      "loss": 0.657,
      "step": 496
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.894774011299435e-05,
      "loss": 0.6485,
      "step": 497
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.894538606403014e-05,
      "loss": 0.5921,
      "step": 498
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8943032015065914e-05,
      "loss": 0.6264,
      "step": 499
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.89406779661017e-05,
      "loss": 0.6239,
      "step": 500
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.893832391713748e-05,
      "loss": 0.5823,
      "step": 501
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.893596986817326e-05,
      "loss": 0.6853,
      "step": 502
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.893361581920904e-05,
      "loss": 0.5856,
      "step": 503
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.893126177024483e-05,
      "loss": 0.6548,
      "step": 504
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8928907721280604e-05,
      "loss": 0.5776,
      "step": 505
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.892655367231639e-05,
      "loss": 0.6032,
      "step": 506
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8924199623352165e-05,
      "loss": 0.6366,
      "step": 507
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.892184557438795e-05,
      "loss": 0.6663,
      "step": 508
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.891949152542373e-05,
      "loss": 0.6061,
      "step": 509
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.891713747645951e-05,
      "loss": 0.6634,
      "step": 510
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8914783427495294e-05,
      "loss": 0.6706,
      "step": 511
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.891242937853108e-05,
      "loss": 0.6159,
      "step": 512
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8910075329566855e-05,
      "loss": 0.7242,
      "step": 513
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.890772128060264e-05,
      "loss": 0.6019,
      "step": 514
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.890536723163842e-05,
      "loss": 0.6863,
      "step": 515
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.89030131826742e-05,
      "loss": 0.6445,
      "step": 516
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8900659133709984e-05,
      "loss": 0.6969,
      "step": 517
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.889830508474577e-05,
      "loss": 0.6388,
      "step": 518
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8895951035781545e-05,
      "loss": 0.6494,
      "step": 519
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.889359698681733e-05,
      "loss": 0.6212,
      "step": 520
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.889124293785311e-05,
      "loss": 0.62,
      "step": 521
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.888888888888889e-05,
      "loss": 0.6494,
      "step": 522
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8886534839924674e-05,
      "loss": 0.6475,
      "step": 523
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.888418079096045e-05,
      "loss": 0.5801,
      "step": 524
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8881826741996235e-05,
      "loss": 0.6253,
      "step": 525
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.887947269303202e-05,
      "loss": 0.6146,
      "step": 526
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8877118644067796e-05,
      "loss": 0.6237,
      "step": 527
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.887476459510358e-05,
      "loss": 0.5503,
      "step": 528
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8872410546139364e-05,
      "loss": 0.6442,
      "step": 529
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.887005649717514e-05,
      "loss": 0.6076,
      "step": 530
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8867702448210925e-05,
      "loss": 0.6669,
      "step": 531
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.886534839924671e-05,
      "loss": 0.6181,
      "step": 532
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8862994350282486e-05,
      "loss": 0.7584,
      "step": 533
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.886064030131827e-05,
      "loss": 0.718,
      "step": 534
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8858286252354054e-05,
      "loss": 0.6001,
      "step": 535
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.885593220338983e-05,
      "loss": 0.5508,
      "step": 536
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8853578154425615e-05,
      "loss": 0.5833,
      "step": 537
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.88512241054614e-05,
      "loss": 0.5685,
      "step": 538
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8848870056497176e-05,
      "loss": 0.5666,
      "step": 539
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.884651600753296e-05,
      "loss": 0.6498,
      "step": 540
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8844161958568744e-05,
      "loss": 0.6808,
      "step": 541
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.884180790960452e-05,
      "loss": 0.5837,
      "step": 542
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8839453860640305e-05,
      "loss": 0.6119,
      "step": 543
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.883709981167608e-05,
      "loss": 0.5911,
      "step": 544
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8834745762711866e-05,
      "loss": 0.6456,
      "step": 545
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.883239171374765e-05,
      "loss": 0.6144,
      "step": 546
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.883003766478343e-05,
      "loss": 0.6007,
      "step": 547
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.882768361581921e-05,
      "loss": 0.6769,
      "step": 548
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8825329566854995e-05,
      "loss": 0.6289,
      "step": 549
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.882297551789077e-05,
      "loss": 0.5484,
      "step": 550
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8820621468926556e-05,
      "loss": 0.6282,
      "step": 551
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.881826741996234e-05,
      "loss": 0.6054,
      "step": 552
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.881591337099812e-05,
      "loss": 0.5775,
      "step": 553
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.88135593220339e-05,
      "loss": 0.5558,
      "step": 554
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8811205273069685e-05,
      "loss": 0.5763,
      "step": 555
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.880885122410546e-05,
      "loss": 0.6572,
      "step": 556
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8806497175141246e-05,
      "loss": 0.5574,
      "step": 557
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.880414312617703e-05,
      "loss": 0.6263,
      "step": 558
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.880178907721281e-05,
      "loss": 0.5932,
      "step": 559
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.879943502824859e-05,
      "loss": 0.6124,
      "step": 560
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.879708097928437e-05,
      "loss": 0.6439,
      "step": 561
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.879472693032015e-05,
      "loss": 0.6052,
      "step": 562
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8792372881355936e-05,
      "loss": 0.6574,
      "step": 563
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.879001883239171e-05,
      "loss": 0.6491,
      "step": 564
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.87876647834275e-05,
      "loss": 0.6192,
      "step": 565
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.878531073446328e-05,
      "loss": 0.7114,
      "step": 566
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.878295668549906e-05,
      "loss": 0.6166,
      "step": 567
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.878060263653484e-05,
      "loss": 0.6218,
      "step": 568
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8778248587570626e-05,
      "loss": 0.6508,
      "step": 569
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.87758945386064e-05,
      "loss": 0.6095,
      "step": 570
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.877354048964219e-05,
      "loss": 0.6798,
      "step": 571
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.877118644067797e-05,
      "loss": 0.6171,
      "step": 572
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.876883239171375e-05,
      "loss": 0.6273,
      "step": 573
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.876647834274953e-05,
      "loss": 0.5677,
      "step": 574
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8764124293785316e-05,
      "loss": 0.6507,
      "step": 575
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.876177024482109e-05,
      "loss": 0.6138,
      "step": 576
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.875941619585688e-05,
      "loss": 0.6194,
      "step": 577
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8757062146892654e-05,
      "loss": 0.6541,
      "step": 578
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.875470809792844e-05,
      "loss": 0.6229,
      "step": 579
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.875235404896422e-05,
      "loss": 0.6128,
      "step": 580
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.875e-05,
      "loss": 0.6406,
      "step": 581
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.874764595103578e-05,
      "loss": 0.617,
      "step": 582
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.874529190207157e-05,
      "loss": 0.6187,
      "step": 583
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8742937853107344e-05,
      "loss": 0.5935,
      "step": 584
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.874058380414313e-05,
      "loss": 0.5446,
      "step": 585
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.873822975517891e-05,
      "loss": 0.604,
      "step": 586
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.873587570621469e-05,
      "loss": 0.6234,
      "step": 587
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.873352165725047e-05,
      "loss": 0.668,
      "step": 588
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.873116760828626e-05,
      "loss": 0.6124,
      "step": 589
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8728813559322034e-05,
      "loss": 0.7136,
      "step": 590
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.872645951035782e-05,
      "loss": 0.6202,
      "step": 591
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.87241054613936e-05,
      "loss": 0.6154,
      "step": 592
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.872175141242938e-05,
      "loss": 0.6142,
      "step": 593
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.871939736346516e-05,
      "loss": 0.6041,
      "step": 594
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.871704331450094e-05,
      "loss": 0.6913,
      "step": 595
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.871468926553673e-05,
      "loss": 0.7386,
      "step": 596
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.871233521657251e-05,
      "loss": 0.6175,
      "step": 597
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8709981167608285e-05,
      "loss": 0.6201,
      "step": 598
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.870762711864407e-05,
      "loss": 0.6651,
      "step": 599
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.870527306967985e-05,
      "loss": 0.5681,
      "step": 600
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.870291902071563e-05,
      "loss": 0.5604,
      "step": 601
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8700564971751414e-05,
      "loss": 0.5787,
      "step": 602
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.86982109227872e-05,
      "loss": 0.6467,
      "step": 603
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8695856873822975e-05,
      "loss": 0.6555,
      "step": 604
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.869350282485876e-05,
      "loss": 0.6622,
      "step": 605
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.869114877589454e-05,
      "loss": 0.6308,
      "step": 606
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.868879472693032e-05,
      "loss": 0.5625,
      "step": 607
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8686440677966104e-05,
      "loss": 0.5904,
      "step": 608
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.868408662900189e-05,
      "loss": 0.5745,
      "step": 609
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8681732580037665e-05,
      "loss": 0.5678,
      "step": 610
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.867937853107345e-05,
      "loss": 0.6277,
      "step": 611
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8677024482109226e-05,
      "loss": 0.6446,
      "step": 612
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.867467043314502e-05,
      "loss": 0.6801,
      "step": 613
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8672316384180794e-05,
      "loss": 0.5571,
      "step": 614
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.866996233521657e-05,
      "loss": 0.638,
      "step": 615
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.866760828625236e-05,
      "loss": 0.6263,
      "step": 616
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.866525423728814e-05,
      "loss": 0.5882,
      "step": 617
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8662900188323916e-05,
      "loss": 0.5585,
      "step": 618
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.86605461393597e-05,
      "loss": 0.6278,
      "step": 619
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8658192090395484e-05,
      "loss": 0.6429,
      "step": 620
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.865583804143126e-05,
      "loss": 0.5504,
      "step": 621
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8653483992467045e-05,
      "loss": 0.6946,
      "step": 622
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.865112994350283e-05,
      "loss": 0.6207,
      "step": 623
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8648775894538606e-05,
      "loss": 0.5106,
      "step": 624
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.864642184557439e-05,
      "loss": 0.5856,
      "step": 625
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8644067796610174e-05,
      "loss": 0.6857,
      "step": 626
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.864171374764596e-05,
      "loss": 0.6064,
      "step": 627
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8639359698681735e-05,
      "loss": 0.638,
      "step": 628
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.863700564971751e-05,
      "loss": 0.6148,
      "step": 629
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.86346516007533e-05,
      "loss": 0.6616,
      "step": 630
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.863229755178908e-05,
      "loss": 0.5755,
      "step": 631
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.862994350282486e-05,
      "loss": 0.6428,
      "step": 632
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.862758945386065e-05,
      "loss": 0.6739,
      "step": 633
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8625235404896425e-05,
      "loss": 0.5954,
      "step": 634
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.86228813559322e-05,
      "loss": 0.6145,
      "step": 635
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8620527306967986e-05,
      "loss": 0.5675,
      "step": 636
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.861817325800377e-05,
      "loss": 0.6535,
      "step": 637
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.861581920903955e-05,
      "loss": 0.6434,
      "step": 638
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.861346516007533e-05,
      "loss": 0.6126,
      "step": 639
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8611111111111115e-05,
      "loss": 0.6132,
      "step": 640
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.860875706214689e-05,
      "loss": 0.5778,
      "step": 641
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8606403013182676e-05,
      "loss": 0.5499,
      "step": 642
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.860404896421846e-05,
      "loss": 0.6425,
      "step": 643
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8601694915254244e-05,
      "loss": 0.7085,
      "step": 644
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.859934086629002e-05,
      "loss": 0.5235,
      "step": 645
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.85969868173258e-05,
      "loss": 0.6908,
      "step": 646
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.859463276836159e-05,
      "loss": 0.6036,
      "step": 647
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8592278719397366e-05,
      "loss": 0.6394,
      "step": 648
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.858992467043314e-05,
      "loss": 0.7065,
      "step": 649
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8587570621468934e-05,
      "loss": 0.6445,
      "step": 650
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.858521657250471e-05,
      "loss": 0.6741,
      "step": 651
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.858286252354049e-05,
      "loss": 0.6161,
      "step": 652
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.858050847457627e-05,
      "loss": 0.6138,
      "step": 653
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8578154425612056e-05,
      "loss": 0.7495,
      "step": 654
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.857580037664783e-05,
      "loss": 0.5824,
      "step": 655
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.857344632768362e-05,
      "loss": 0.6073,
      "step": 656
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.85710922787194e-05,
      "loss": 0.6253,
      "step": 657
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.856873822975518e-05,
      "loss": 0.5975,
      "step": 658
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.856638418079096e-05,
      "loss": 0.6669,
      "step": 659
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8564030131826746e-05,
      "loss": 0.543,
      "step": 660
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.856167608286253e-05,
      "loss": 0.6545,
      "step": 661
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.855932203389831e-05,
      "loss": 0.5986,
      "step": 662
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8556967984934084e-05,
      "loss": 0.576,
      "step": 663
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8554613935969875e-05,
      "loss": 0.6801,
      "step": 664
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.855225988700565e-05,
      "loss": 0.5862,
      "step": 665
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.854990583804143e-05,
      "loss": 0.5303,
      "step": 666
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.854755178907722e-05,
      "loss": 0.6014,
      "step": 667
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8545197740113e-05,
      "loss": 0.678,
      "step": 668
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8542843691148774e-05,
      "loss": 0.7181,
      "step": 669
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.854048964218456e-05,
      "loss": 0.6667,
      "step": 670
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.853813559322034e-05,
      "loss": 0.6387,
      "step": 671
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.853578154425612e-05,
      "loss": 0.6104,
      "step": 672
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.85334274952919e-05,
      "loss": 0.5413,
      "step": 673
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.853107344632769e-05,
      "loss": 0.6183,
      "step": 674
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.852871939736347e-05,
      "loss": 0.574,
      "step": 675
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.852636534839925e-05,
      "loss": 0.6834,
      "step": 676
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.852401129943503e-05,
      "loss": 0.6254,
      "step": 677
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.8521657250470816e-05,
      "loss": 0.5562,
      "step": 678
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.851930320150659e-05,
      "loss": 0.5903,
      "step": 679
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.851694915254237e-05,
      "loss": 0.5961,
      "step": 680
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.851459510357816e-05,
      "loss": 0.6249,
      "step": 681
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.851224105461394e-05,
      "loss": 0.6372,
      "step": 682
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.8509887005649715e-05,
      "loss": 0.5141,
      "step": 683
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.8507532956685506e-05,
      "loss": 0.6379,
      "step": 684
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.850517890772128e-05,
      "loss": 0.6042,
      "step": 685
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.850282485875706e-05,
      "loss": 0.6922,
      "step": 686
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.8500470809792844e-05,
      "loss": 0.6356,
      "step": 687
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.849811676082863e-05,
      "loss": 0.6297,
      "step": 688
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.8495762711864405e-05,
      "loss": 0.6763,
      "step": 689
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.849340866290019e-05,
      "loss": 0.6197,
      "step": 690
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.849105461393597e-05,
      "loss": 0.6232,
      "step": 691
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.848870056497176e-05,
      "loss": 0.6219,
      "step": 692
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8486346516007534e-05,
      "loss": 0.6014,
      "step": 693
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.848399246704332e-05,
      "loss": 0.6412,
      "step": 694
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.84816384180791e-05,
      "loss": 0.5831,
      "step": 695
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.847928436911488e-05,
      "loss": 0.5874,
      "step": 696
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8476930320150657e-05,
      "loss": 0.6024,
      "step": 697
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.847457627118645e-05,
      "loss": 0.727,
      "step": 698
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8472222222222224e-05,
      "loss": 0.6361,
      "step": 699
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8469868173258e-05,
      "loss": 0.6107,
      "step": 700
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.846751412429379e-05,
      "loss": 0.6151,
      "step": 701
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.846516007532957e-05,
      "loss": 0.5532,
      "step": 702
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8462806026365347e-05,
      "loss": 0.6207,
      "step": 703
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.846045197740113e-05,
      "loss": 0.6307,
      "step": 704
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8458097928436914e-05,
      "loss": 0.6683,
      "step": 705
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.84557438794727e-05,
      "loss": 0.6133,
      "step": 706
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8453389830508475e-05,
      "loss": 0.5911,
      "step": 707
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.845103578154426e-05,
      "loss": 0.5459,
      "step": 708
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.844868173258004e-05,
      "loss": 0.6459,
      "step": 709
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.844632768361582e-05,
      "loss": 0.6029,
      "step": 710
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8443973634651604e-05,
      "loss": 0.6229,
      "step": 711
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.844161958568739e-05,
      "loss": 0.6038,
      "step": 712
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8439265536723165e-05,
      "loss": 0.6234,
      "step": 713
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.843691148775894e-05,
      "loss": 0.5612,
      "step": 714
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.843455743879473e-05,
      "loss": 0.6934,
      "step": 715
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.843220338983051e-05,
      "loss": 0.7028,
      "step": 716
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.842984934086629e-05,
      "loss": 0.685,
      "step": 717
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.842749529190208e-05,
      "loss": 0.6815,
      "step": 718
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8425141242937855e-05,
      "loss": 0.6432,
      "step": 719
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.842278719397363e-05,
      "loss": 0.6114,
      "step": 720
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8420433145009417e-05,
      "loss": 0.7284,
      "step": 721
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.84180790960452e-05,
      "loss": 0.5731,
      "step": 722
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8415725047080984e-05,
      "loss": 0.6295,
      "step": 723
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.841337099811676e-05,
      "loss": 0.5874,
      "step": 724
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8411016949152545e-05,
      "loss": 0.7564,
      "step": 725
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.840866290018833e-05,
      "loss": 0.6438,
      "step": 726
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8406308851224106e-05,
      "loss": 0.6027,
      "step": 727
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.840395480225989e-05,
      "loss": 0.6746,
      "step": 728
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8401600753295674e-05,
      "loss": 0.661,
      "step": 729
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.839924670433145e-05,
      "loss": 0.5698,
      "step": 730
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.839689265536723e-05,
      "loss": 0.6258,
      "step": 731
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.839453860640302e-05,
      "loss": 0.6249,
      "step": 732
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8392184557438796e-05,
      "loss": 0.6278,
      "step": 733
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8389830508474574e-05,
      "loss": 0.5305,
      "step": 734
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8387476459510364e-05,
      "loss": 0.5569,
      "step": 735
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.838512241054614e-05,
      "loss": 0.6538,
      "step": 736
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.838276836158192e-05,
      "loss": 0.5709,
      "step": 737
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.83804143126177e-05,
      "loss": 0.5671,
      "step": 738
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8378060263653486e-05,
      "loss": 0.6532,
      "step": 739
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.837570621468927e-05,
      "loss": 0.6429,
      "step": 740
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.837335216572505e-05,
      "loss": 0.5366,
      "step": 741
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.837099811676083e-05,
      "loss": 0.6145,
      "step": 742
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8368644067796615e-05,
      "loss": 0.6061,
      "step": 743
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.836629001883239e-05,
      "loss": 0.7353,
      "step": 744
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8363935969868176e-05,
      "loss": 0.6444,
      "step": 745
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.836158192090396e-05,
      "loss": 0.5201,
      "step": 746
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.835922787193974e-05,
      "loss": 0.6515,
      "step": 747
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8356873822975515e-05,
      "loss": 0.698,
      "step": 748
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8354519774011305e-05,
      "loss": 0.5823,
      "step": 749
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.835216572504708e-05,
      "loss": 0.5327,
      "step": 750
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.834981167608286e-05,
      "loss": 0.6294,
      "step": 751
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.834745762711865e-05,
      "loss": 0.5886,
      "step": 752
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.834510357815443e-05,
      "loss": 0.5674,
      "step": 753
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.834274952919021e-05,
      "loss": 0.593,
      "step": 754
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.834039548022599e-05,
      "loss": 0.6456,
      "step": 755
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.833804143126177e-05,
      "loss": 0.6764,
      "step": 756
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8335687382297556e-05,
      "loss": 0.6694,
      "step": 757
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8333333333333334e-05,
      "loss": 0.5505,
      "step": 758
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.833097928436912e-05,
      "loss": 0.6387,
      "step": 759
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.83286252354049e-05,
      "loss": 0.6097,
      "step": 760
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.832627118644068e-05,
      "loss": 0.5811,
      "step": 761
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.832391713747646e-05,
      "loss": 0.617,
      "step": 762
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8321563088512246e-05,
      "loss": 0.5836,
      "step": 763
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8319209039548024e-05,
      "loss": 0.6217,
      "step": 764
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.831685499058381e-05,
      "loss": 0.5317,
      "step": 765
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.831450094161959e-05,
      "loss": 0.5566,
      "step": 766
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.831214689265537e-05,
      "loss": 0.5916,
      "step": 767
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8309792843691146e-05,
      "loss": 0.6032,
      "step": 768
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8307438794726936e-05,
      "loss": 0.5431,
      "step": 769
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8305084745762714e-05,
      "loss": 0.564,
      "step": 770
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.83027306967985e-05,
      "loss": 0.5869,
      "step": 771
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8300376647834275e-05,
      "loss": 0.6008,
      "step": 772
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.829802259887006e-05,
      "loss": 0.5735,
      "step": 773
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.829566854990584e-05,
      "loss": 0.6322,
      "step": 774
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.829331450094162e-05,
      "loss": 0.577,
      "step": 775
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.8290960451977404e-05,
      "loss": 0.5537,
      "step": 776
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.828860640301319e-05,
      "loss": 0.64,
      "step": 777
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8286252354048965e-05,
      "loss": 0.7971,
      "step": 778
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.828389830508475e-05,
      "loss": 0.739,
      "step": 779
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.828154425612053e-05,
      "loss": 0.5542,
      "step": 780
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.827919020715631e-05,
      "loss": 0.5621,
      "step": 781
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8276836158192094e-05,
      "loss": 0.6847,
      "step": 782
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.827448210922788e-05,
      "loss": 0.618,
      "step": 783
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8272128060263655e-05,
      "loss": 0.559,
      "step": 784
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.826977401129944e-05,
      "loss": 0.5478,
      "step": 785
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.826741996233522e-05,
      "loss": 0.5501,
      "step": 786
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8265065913371e-05,
      "loss": 0.647,
      "step": 787
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8262711864406784e-05,
      "loss": 0.6673,
      "step": 788
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.826035781544256e-05,
      "loss": 0.5661,
      "step": 789
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8258003766478345e-05,
      "loss": 0.6418,
      "step": 790
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.825564971751413e-05,
      "loss": 0.5894,
      "step": 791
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8253295668549906e-05,
      "loss": 0.6125,
      "step": 792
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.825094161958569e-05,
      "loss": 0.6207,
      "step": 793
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8248587570621474e-05,
      "loss": 0.6351,
      "step": 794
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.824623352165725e-05,
      "loss": 0.612,
      "step": 795
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8243879472693035e-05,
      "loss": 0.6421,
      "step": 796
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.824152542372882e-05,
      "loss": 0.5526,
      "step": 797
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8239171374764596e-05,
      "loss": 0.6595,
      "step": 798
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.823681732580038e-05,
      "loss": 0.7017,
      "step": 799
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8234463276836164e-05,
      "loss": 0.6451,
      "step": 800
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.823210922787194e-05,
      "loss": 0.618,
      "step": 801
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8229755178907725e-05,
      "loss": 0.5809,
      "step": 802
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.822740112994351e-05,
      "loss": 0.6161,
      "step": 803
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8225047080979286e-05,
      "loss": 0.6176,
      "step": 804
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.822269303201507e-05,
      "loss": 0.6457,
      "step": 805
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.822033898305085e-05,
      "loss": 0.6448,
      "step": 806
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.821798493408663e-05,
      "loss": 0.6331,
      "step": 807
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8215630885122415e-05,
      "loss": 0.645,
      "step": 808
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.821327683615819e-05,
      "loss": 0.5629,
      "step": 809
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8210922787193976e-05,
      "loss": 0.5951,
      "step": 810
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.820856873822976e-05,
      "loss": 0.6019,
      "step": 811
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.820621468926554e-05,
      "loss": 0.6395,
      "step": 812
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.820386064030132e-05,
      "loss": 0.5849,
      "step": 813
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8201506591337105e-05,
      "loss": 0.5712,
      "step": 814
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.819915254237288e-05,
      "loss": 0.5639,
      "step": 815
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8196798493408666e-05,
      "loss": 0.4692,
      "step": 816
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.819444444444445e-05,
      "loss": 0.6298,
      "step": 817
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.819209039548023e-05,
      "loss": 0.6205,
      "step": 818
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.818973634651601e-05,
      "loss": 0.6453,
      "step": 819
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.8187382297551795e-05,
      "loss": 0.6607,
      "step": 820
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.818502824858757e-05,
      "loss": 0.6248,
      "step": 821
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.8182674199623356e-05,
      "loss": 0.5185,
      "step": 822
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.818032015065913e-05,
      "loss": 0.5788,
      "step": 823
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.817796610169492e-05,
      "loss": 0.5363,
      "step": 824
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.81756120527307e-05,
      "loss": 0.5532,
      "step": 825
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.817325800376648e-05,
      "loss": 0.5867,
      "step": 826
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.817090395480226e-05,
      "loss": 0.629,
      "step": 827
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.8168549905838046e-05,
      "loss": 0.5829,
      "step": 828
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.816619585687382e-05,
      "loss": 0.6395,
      "step": 829
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.816384180790961e-05,
      "loss": 0.7104,
      "step": 830
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.816148775894539e-05,
      "loss": 0.6839,
      "step": 831
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.815913370998117e-05,
      "loss": 0.5375,
      "step": 832
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.815677966101695e-05,
      "loss": 0.6644,
      "step": 833
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.8154425612052736e-05,
      "loss": 0.6155,
      "step": 834
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.815207156308851e-05,
      "loss": 0.5561,
      "step": 835
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.81497175141243e-05,
      "loss": 0.6128,
      "step": 836
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.814736346516008e-05,
      "loss": 0.6028,
      "step": 837
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.814500941619586e-05,
      "loss": 0.6076,
      "step": 838
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.814265536723164e-05,
      "loss": 0.5775,
      "step": 839
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.814030131826742e-05,
      "loss": 0.6929,
      "step": 840
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.81379472693032e-05,
      "loss": 0.6453,
      "step": 841
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.813559322033899e-05,
      "loss": 0.5984,
      "step": 842
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.8133239171374764e-05,
      "loss": 0.6715,
      "step": 843
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.813088512241055e-05,
      "loss": 0.6081,
      "step": 844
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.812853107344633e-05,
      "loss": 0.7005,
      "step": 845
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.812617702448211e-05,
      "loss": 0.575,
      "step": 846
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.812382297551789e-05,
      "loss": 0.6672,
      "step": 847
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.812146892655368e-05,
      "loss": 0.6999,
      "step": 848
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.8119114877589454e-05,
      "loss": 0.6442,
      "step": 849
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.811676082862524e-05,
      "loss": 0.5975,
      "step": 850
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.811440677966102e-05,
      "loss": 0.5366,
      "step": 851
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.81120527306968e-05,
      "loss": 0.6808,
      "step": 852
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.810969868173258e-05,
      "loss": 0.6133,
      "step": 853
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.810734463276837e-05,
      "loss": 0.638,
      "step": 854
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.8104990583804144e-05,
      "loss": 0.6319,
      "step": 855
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.810263653483993e-05,
      "loss": 0.6051,
      "step": 856
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.810028248587571e-05,
      "loss": 0.6146,
      "step": 857
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.809792843691149e-05,
      "loss": 0.6383,
      "step": 858
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.809557438794727e-05,
      "loss": 0.5958,
      "step": 859
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.809322033898305e-05,
      "loss": 0.5874,
      "step": 860
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.8090866290018834e-05,
      "loss": 0.5585,
      "step": 861
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.808851224105462e-05,
      "loss": 0.6077,
      "step": 862
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8086158192090395e-05,
      "loss": 0.5857,
      "step": 863
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.808380414312618e-05,
      "loss": 0.6555,
      "step": 864
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.808145009416196e-05,
      "loss": 0.5801,
      "step": 865
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.807909604519774e-05,
      "loss": 0.6116,
      "step": 866
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8076741996233524e-05,
      "loss": 0.6502,
      "step": 867
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.807438794726931e-05,
      "loss": 0.6122,
      "step": 868
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8072033898305085e-05,
      "loss": 0.6623,
      "step": 869
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.806967984934087e-05,
      "loss": 0.5932,
      "step": 870
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.806732580037665e-05,
      "loss": 0.5905,
      "step": 871
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.806497175141243e-05,
      "loss": 0.5558,
      "step": 872
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8062617702448214e-05,
      "loss": 0.5396,
      "step": 873
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8060263653484e-05,
      "loss": 0.5407,
      "step": 874
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8057909604519775e-05,
      "loss": 0.6221,
      "step": 875
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.805555555555556e-05,
      "loss": 0.5977,
      "step": 876
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8053201506591336e-05,
      "loss": 0.5898,
      "step": 877
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.805084745762712e-05,
      "loss": 0.5895,
      "step": 878
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8048493408662904e-05,
      "loss": 0.6666,
      "step": 879
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.804613935969868e-05,
      "loss": 0.5937,
      "step": 880
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8043785310734465e-05,
      "loss": 0.5408,
      "step": 881
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.804143126177025e-05,
      "loss": 0.6329,
      "step": 882
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8039077212806026e-05,
      "loss": 0.5777,
      "step": 883
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.803672316384181e-05,
      "loss": 0.6265,
      "step": 884
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8034369114877594e-05,
      "loss": 0.6415,
      "step": 885
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.803201506591337e-05,
      "loss": 0.5785,
      "step": 886
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8029661016949155e-05,
      "loss": 0.6226,
      "step": 887
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.802730696798494e-05,
      "loss": 0.5555,
      "step": 888
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8024952919020716e-05,
      "loss": 0.5895,
      "step": 889
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.80225988700565e-05,
      "loss": 0.5894,
      "step": 890
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8020244821092284e-05,
      "loss": 0.6404,
      "step": 891
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.801789077212806e-05,
      "loss": 0.6507,
      "step": 892
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8015536723163845e-05,
      "loss": 0.6383,
      "step": 893
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.801318267419962e-05,
      "loss": 0.5818,
      "step": 894
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8010828625235406e-05,
      "loss": 0.6624,
      "step": 895
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.800847457627119e-05,
      "loss": 0.605,
      "step": 896
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.800612052730697e-05,
      "loss": 0.5405,
      "step": 897
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.800376647834275e-05,
      "loss": 0.6288,
      "step": 898
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.8001412429378535e-05,
      "loss": 0.5494,
      "step": 899
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.799905838041431e-05,
      "loss": 0.5406,
      "step": 900
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.7996704331450096e-05,
      "loss": 0.5977,
      "step": 901
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.799435028248588e-05,
      "loss": 0.5726,
      "step": 902
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.799199623352166e-05,
      "loss": 0.6163,
      "step": 903
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.798964218455744e-05,
      "loss": 0.5651,
      "step": 904
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7987288135593225e-05,
      "loss": 0.556,
      "step": 905
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7984934086629e-05,
      "loss": 0.64,
      "step": 906
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7982580037664786e-05,
      "loss": 0.5121,
      "step": 907
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.798022598870057e-05,
      "loss": 0.6552,
      "step": 908
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.797787193973635e-05,
      "loss": 0.5757,
      "step": 909
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.797551789077213e-05,
      "loss": 0.6891,
      "step": 910
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.797316384180791e-05,
      "loss": 0.573,
      "step": 911
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.79708097928437e-05,
      "loss": 0.6991,
      "step": 912
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7968455743879476e-05,
      "loss": 0.5743,
      "step": 913
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.796610169491525e-05,
      "loss": 0.7137,
      "step": 914
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.796374764595104e-05,
      "loss": 0.5736,
      "step": 915
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.796139359698682e-05,
      "loss": 0.5251,
      "step": 916
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.79590395480226e-05,
      "loss": 0.617,
      "step": 917
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.795668549905838e-05,
      "loss": 0.5703,
      "step": 918
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7954331450094166e-05,
      "loss": 0.6639,
      "step": 919
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.795197740112994e-05,
      "loss": 0.6415,
      "step": 920
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.794962335216573e-05,
      "loss": 0.5932,
      "step": 921
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.794726930320151e-05,
      "loss": 0.5338,
      "step": 922
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.794491525423729e-05,
      "loss": 0.5875,
      "step": 923
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.794256120527307e-05,
      "loss": 0.5926,
      "step": 924
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7940207156308856e-05,
      "loss": 0.5815,
      "step": 925
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.793785310734463e-05,
      "loss": 0.5715,
      "step": 926
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.793549905838042e-05,
      "loss": 0.5369,
      "step": 927
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7933145009416194e-05,
      "loss": 0.595,
      "step": 928
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7930790960451985e-05,
      "loss": 0.5691,
      "step": 929
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.792843691148776e-05,
      "loss": 0.6029,
      "step": 930
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.792608286252354e-05,
      "loss": 0.6489,
      "step": 931
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.792372881355933e-05,
      "loss": 0.5891,
      "step": 932
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.792137476459511e-05,
      "loss": 0.5288,
      "step": 933
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7919020715630884e-05,
      "loss": 0.6305,
      "step": 934
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.791666666666667e-05,
      "loss": 0.5949,
      "step": 935
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.791431261770245e-05,
      "loss": 0.6451,
      "step": 936
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.791195856873823e-05,
      "loss": 0.5777,
      "step": 937
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.790960451977401e-05,
      "loss": 0.5907,
      "step": 938
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.79072504708098e-05,
      "loss": 0.5777,
      "step": 939
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7904896421845574e-05,
      "loss": 0.5619,
      "step": 940
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.790254237288136e-05,
      "loss": 0.5803,
      "step": 941
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.790018832391714e-05,
      "loss": 0.5947,
      "step": 942
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7897834274952926e-05,
      "loss": 0.4791,
      "step": 943
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.78954802259887e-05,
      "loss": 0.6261,
      "step": 944
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.789312617702448e-05,
      "loss": 0.6609,
      "step": 945
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.789077212806027e-05,
      "loss": 0.5182,
      "step": 946
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.788841807909605e-05,
      "loss": 0.5193,
      "step": 947
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7886064030131825e-05,
      "loss": 0.5374,
      "step": 948
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7883709981167616e-05,
      "loss": 0.589,
      "step": 949
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.788135593220339e-05,
      "loss": 0.7054,
      "step": 950
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.787900188323917e-05,
      "loss": 0.5411,
      "step": 951
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7876647834274954e-05,
      "loss": 0.6457,
      "step": 952
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.787429378531074e-05,
      "loss": 0.634,
      "step": 953
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7871939736346515e-05,
      "loss": 0.6506,
      "step": 954
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.78695856873823e-05,
      "loss": 0.605,
      "step": 955
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.786723163841808e-05,
      "loss": 0.6222,
      "step": 956
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.786487758945386e-05,
      "loss": 0.4944,
      "step": 957
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7862523540489644e-05,
      "loss": 0.6237,
      "step": 958
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.786016949152543e-05,
      "loss": 0.5886,
      "step": 959
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.785781544256121e-05,
      "loss": 0.6323,
      "step": 960
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.785546139359699e-05,
      "loss": 0.6215,
      "step": 961
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7853107344632766e-05,
      "loss": 0.6327,
      "step": 962
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.785075329566856e-05,
      "loss": 0.6488,
      "step": 963
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7848399246704334e-05,
      "loss": 0.615,
      "step": 964
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.784604519774011e-05,
      "loss": 0.6327,
      "step": 965
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.78436911487759e-05,
      "loss": 0.6419,
      "step": 966
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.784133709981168e-05,
      "loss": 0.5815,
      "step": 967
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7838983050847456e-05,
      "loss": 0.5937,
      "step": 968
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.783662900188324e-05,
      "loss": 0.6221,
      "step": 969
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7834274952919024e-05,
      "loss": 0.6163,
      "step": 970
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.78319209039548e-05,
      "loss": 0.6384,
      "step": 971
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7829566854990585e-05,
      "loss": 0.6095,
      "step": 972
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.782721280602637e-05,
      "loss": 0.6176,
      "step": 973
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7824858757062146e-05,
      "loss": 0.5537,
      "step": 974
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.782250470809793e-05,
      "loss": 0.5937,
      "step": 975
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7820150659133714e-05,
      "loss": 0.5471,
      "step": 976
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.78177966101695e-05,
      "loss": 0.5661,
      "step": 977
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7815442561205275e-05,
      "loss": 0.5721,
      "step": 978
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.781308851224105e-05,
      "loss": 0.5541,
      "step": 979
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.781073446327684e-05,
      "loss": 0.5174,
      "step": 980
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.780838041431262e-05,
      "loss": 0.6759,
      "step": 981
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.78060263653484e-05,
      "loss": 0.5837,
      "step": 982
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.780367231638419e-05,
      "loss": 0.5483,
      "step": 983
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7801318267419965e-05,
      "loss": 0.6631,
      "step": 984
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.779896421845574e-05,
      "loss": 0.576,
      "step": 985
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7796610169491526e-05,
      "loss": 0.6056,
      "step": 986
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.779425612052731e-05,
      "loss": 0.5654,
      "step": 987
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.779190207156309e-05,
      "loss": 0.5246,
      "step": 988
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.778954802259887e-05,
      "loss": 0.6072,
      "step": 989
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7787193973634655e-05,
      "loss": 0.6549,
      "step": 990
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.778483992467044e-05,
      "loss": 0.5565,
      "step": 991
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7782485875706216e-05,
      "loss": 0.5686,
      "step": 992
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7780131826742e-05,
      "loss": 0.6389,
      "step": 993
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7777777777777784e-05,
      "loss": 0.5548,
      "step": 994
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.777542372881356e-05,
      "loss": 0.5707,
      "step": 995
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.777306967984934e-05,
      "loss": 0.6854,
      "step": 996
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.777071563088513e-05,
      "loss": 0.5947,
      "step": 997
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7768361581920906e-05,
      "loss": 0.5621,
      "step": 998
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7766007532956683e-05,
      "loss": 0.6049,
      "step": 999
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7763653483992474e-05,
      "loss": 0.5892,
      "step": 1000
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.776129943502825e-05,
      "loss": 0.5727,
      "step": 1001
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.775894538606403e-05,
      "loss": 0.7136,
      "step": 1002
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.775659133709981e-05,
      "loss": 0.6435,
      "step": 1003
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7754237288135596e-05,
      "loss": 0.6543,
      "step": 1004
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7751883239171373e-05,
      "loss": 0.6426,
      "step": 1005
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.774952919020716e-05,
      "loss": 0.6012,
      "step": 1006
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.774717514124294e-05,
      "loss": 0.562,
      "step": 1007
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7744821092278725e-05,
      "loss": 0.5602,
      "step": 1008
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.77424670433145e-05,
      "loss": 0.5895,
      "step": 1009
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7740112994350286e-05,
      "loss": 0.5967,
      "step": 1010
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.773775894538607e-05,
      "loss": 0.7087,
      "step": 1011
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.773540489642185e-05,
      "loss": 0.5687,
      "step": 1012
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7733050847457624e-05,
      "loss": 0.631,
      "step": 1013
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7730696798493415e-05,
      "loss": 0.4818,
      "step": 1014
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.772834274952919e-05,
      "loss": 0.5714,
      "step": 1015
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.772598870056497e-05,
      "loss": 0.6364,
      "step": 1016
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.772363465160076e-05,
      "loss": 0.5895,
      "step": 1017
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.772128060263654e-05,
      "loss": 0.5855,
      "step": 1018
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7718926553672314e-05,
      "loss": 0.5864,
      "step": 1019
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.77165725047081e-05,
      "loss": 0.5809,
      "step": 1020
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.771421845574388e-05,
      "loss": 0.6844,
      "step": 1021
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7711864406779666e-05,
      "loss": 0.5592,
      "step": 1022
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7709510357815443e-05,
      "loss": 0.6002,
      "step": 1023
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.770715630885123e-05,
      "loss": 0.5927,
      "step": 1024
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.770480225988701e-05,
      "loss": 0.5838,
      "step": 1025
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.770244821092279e-05,
      "loss": 0.6362,
      "step": 1026
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.770009416195857e-05,
      "loss": 0.7157,
      "step": 1027
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7697740112994356e-05,
      "loss": 0.563,
      "step": 1028
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7695386064030133e-05,
      "loss": 0.5688,
      "step": 1029
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.769303201506591e-05,
      "loss": 0.5569,
      "step": 1030
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.76906779661017e-05,
      "loss": 0.5747,
      "step": 1031
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.768832391713748e-05,
      "loss": 0.602,
      "step": 1032
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7685969868173256e-05,
      "loss": 0.6086,
      "step": 1033
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7683615819209046e-05,
      "loss": 0.6181,
      "step": 1034
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7681261770244823e-05,
      "loss": 0.5735,
      "step": 1035
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.76789077212806e-05,
      "loss": 0.5767,
      "step": 1036
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7676553672316384e-05,
      "loss": 0.5419,
      "step": 1037
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.767419962335217e-05,
      "loss": 0.7051,
      "step": 1038
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.767184557438795e-05,
      "loss": 0.653,
      "step": 1039
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.766949152542373e-05,
      "loss": 0.549,
      "step": 1040
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.766713747645951e-05,
      "loss": 0.7132,
      "step": 1041
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.76647834274953e-05,
      "loss": 0.564,
      "step": 1042
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7662429378531074e-05,
      "loss": 0.6448,
      "step": 1043
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.766007532956686e-05,
      "loss": 0.6136,
      "step": 1044
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.765772128060264e-05,
      "loss": 0.6733,
      "step": 1045
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.765536723163842e-05,
      "loss": 0.6189,
      "step": 1046
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7653013182674197e-05,
      "loss": 0.6634,
      "step": 1047
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.765065913370999e-05,
      "loss": 0.5661,
      "step": 1048
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7648305084745764e-05,
      "loss": 0.5251,
      "step": 1049
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.764595103578154e-05,
      "loss": 0.608,
      "step": 1050
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.764359698681733e-05,
      "loss": 0.6729,
      "step": 1051
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.764124293785311e-05,
      "loss": 0.6342,
      "step": 1052
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.7638888888888887e-05,
      "loss": 0.5823,
      "step": 1053
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.763653483992467e-05,
      "loss": 0.5972,
      "step": 1054
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7634180790960454e-05,
      "loss": 0.6726,
      "step": 1055
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.763182674199624e-05,
      "loss": 0.622,
      "step": 1056
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7629472693032016e-05,
      "loss": 0.6376,
      "step": 1057
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.76271186440678e-05,
      "loss": 0.6049,
      "step": 1058
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.762476459510358e-05,
      "loss": 0.6477,
      "step": 1059
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.762241054613936e-05,
      "loss": 0.5329,
      "step": 1060
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7620056497175144e-05,
      "loss": 0.613,
      "step": 1061
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.761770244821093e-05,
      "loss": 0.6095,
      "step": 1062
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7615348399246706e-05,
      "loss": 0.5741,
      "step": 1063
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.761299435028248e-05,
      "loss": 0.6439,
      "step": 1064
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.761064030131827e-05,
      "loss": 0.6045,
      "step": 1065
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.760828625235405e-05,
      "loss": 0.5435,
      "step": 1066
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.760593220338983e-05,
      "loss": 0.5276,
      "step": 1067
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.760357815442562e-05,
      "loss": 0.6162,
      "step": 1068
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7601224105461396e-05,
      "loss": 0.6046,
      "step": 1069
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.759887005649718e-05,
      "loss": 0.529,
      "step": 1070
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7596516007532957e-05,
      "loss": 0.5744,
      "step": 1071
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.759416195856874e-05,
      "loss": 0.6524,
      "step": 1072
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7591807909604524e-05,
      "loss": 0.5446,
      "step": 1073
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.75894538606403e-05,
      "loss": 0.5975,
      "step": 1074
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.7587099811676086e-05,
      "loss": 0.5597,
      "step": 1075
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.758474576271187e-05,
      "loss": 0.6786,
      "step": 1076
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7582391713747647e-05,
      "loss": 0.6574,
      "step": 1077
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.758003766478343e-05,
      "loss": 0.5852,
      "step": 1078
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7577683615819214e-05,
      "loss": 0.512,
      "step": 1079
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.757532956685499e-05,
      "loss": 0.6053,
      "step": 1080
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7572975517890775e-05,
      "loss": 0.5295,
      "step": 1081
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.757062146892656e-05,
      "loss": 0.6789,
      "step": 1082
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7568267419962337e-05,
      "loss": 0.6406,
      "step": 1083
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7565913370998114e-05,
      "loss": 0.5585,
      "step": 1084
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7563559322033904e-05,
      "loss": 0.5694,
      "step": 1085
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.756120527306968e-05,
      "loss": 0.566,
      "step": 1086
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7558851224105465e-05,
      "loss": 0.6614,
      "step": 1087
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.755649717514124e-05,
      "loss": 0.537,
      "step": 1088
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7554143126177027e-05,
      "loss": 0.6308,
      "step": 1089
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.755178907721281e-05,
      "loss": 0.6308,
      "step": 1090
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.754943502824859e-05,
      "loss": 0.554,
      "step": 1091
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.754708097928437e-05,
      "loss": 0.5355,
      "step": 1092
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7544726930320155e-05,
      "loss": 0.5546,
      "step": 1093
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.754237288135593e-05,
      "loss": 0.5967,
      "step": 1094
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.7540018832391717e-05,
      "loss": 0.5931,
      "step": 1095
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.75376647834275e-05,
      "loss": 0.5328,
      "step": 1096
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.753531073446328e-05,
      "loss": 0.7418,
      "step": 1097
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.753295668549906e-05,
      "loss": 0.515,
      "step": 1098
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.7530602636534845e-05,
      "loss": 0.6674,
      "step": 1099
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.752824858757062e-05,
      "loss": 0.4829,
      "step": 1100
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.7525894538606407e-05,
      "loss": 0.5905,
      "step": 1101
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.752354048964219e-05,
      "loss": 0.5042,
      "step": 1102
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.752118644067797e-05,
      "loss": 0.6387,
      "step": 1103
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.751883239171375e-05,
      "loss": 0.4897,
      "step": 1104
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.751647834274953e-05,
      "loss": 0.5269,
      "step": 1105
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.751412429378531e-05,
      "loss": 0.5614,
      "step": 1106
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.7511770244821097e-05,
      "loss": 0.6766,
      "step": 1107
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.7509416195856874e-05,
      "loss": 0.6215,
      "step": 1108
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.750706214689266e-05,
      "loss": 0.7274,
      "step": 1109
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.750470809792844e-05,
      "loss": 0.5437,
      "step": 1110
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.750235404896422e-05,
      "loss": 0.6122,
      "step": 1111
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.75e-05,
      "loss": 0.6177,
      "step": 1112
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.7497645951035787e-05,
      "loss": 0.576,
      "step": 1113
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.7495291902071564e-05,
      "loss": 0.5678,
      "step": 1114
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.749293785310735e-05,
      "loss": 0.7428,
      "step": 1115
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.749058380414313e-05,
      "loss": 0.5649,
      "step": 1116
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.748822975517891e-05,
      "loss": 0.6778,
      "step": 1117
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.748587570621469e-05,
      "loss": 0.5645,
      "step": 1118
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7483521657250477e-05,
      "loss": 0.5718,
      "step": 1119
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7481167608286254e-05,
      "loss": 0.5896,
      "step": 1120
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.747881355932204e-05,
      "loss": 0.6049,
      "step": 1121
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7476459510357815e-05,
      "loss": 0.6298,
      "step": 1122
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.74741054613936e-05,
      "loss": 0.6829,
      "step": 1123
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.747175141242938e-05,
      "loss": 0.5639,
      "step": 1124
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.746939736346516e-05,
      "loss": 0.5878,
      "step": 1125
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7467043314500944e-05,
      "loss": 0.6381,
      "step": 1126
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.746468926553673e-05,
      "loss": 0.6003,
      "step": 1127
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7462335216572505e-05,
      "loss": 0.6087,
      "step": 1128
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.745998116760829e-05,
      "loss": 0.5633,
      "step": 1129
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.745762711864407e-05,
      "loss": 0.5612,
      "step": 1130
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.745527306967985e-05,
      "loss": 0.5216,
      "step": 1131
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7452919020715634e-05,
      "loss": 0.5991,
      "step": 1132
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.745056497175142e-05,
      "loss": 0.5151,
      "step": 1133
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7448210922787195e-05,
      "loss": 0.5961,
      "step": 1134
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.744585687382298e-05,
      "loss": 0.6011,
      "step": 1135
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.744350282485876e-05,
      "loss": 0.5484,
      "step": 1136
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.744114877589454e-05,
      "loss": 0.5683,
      "step": 1137
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.7438794726930324e-05,
      "loss": 0.5573,
      "step": 1138
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.74364406779661e-05,
      "loss": 0.5757,
      "step": 1139
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7434086629001885e-05,
      "loss": 0.676,
      "step": 1140
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.743173258003767e-05,
      "loss": 0.5208,
      "step": 1141
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7429378531073446e-05,
      "loss": 0.4764,
      "step": 1142
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.742702448210923e-05,
      "loss": 0.5856,
      "step": 1143
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7424670433145014e-05,
      "loss": 0.584,
      "step": 1144
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.742231638418079e-05,
      "loss": 0.5478,
      "step": 1145
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7419962335216575e-05,
      "loss": 0.5687,
      "step": 1146
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.741760828625236e-05,
      "loss": 0.6162,
      "step": 1147
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7415254237288136e-05,
      "loss": 0.638,
      "step": 1148
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.741290018832392e-05,
      "loss": 0.5485,
      "step": 1149
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7410546139359704e-05,
      "loss": 0.536,
      "step": 1150
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.740819209039548e-05,
      "loss": 0.5246,
      "step": 1151
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7405838041431265e-05,
      "loss": 0.5695,
      "step": 1152
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.740348399246705e-05,
      "loss": 0.5687,
      "step": 1153
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7401129943502826e-05,
      "loss": 0.5373,
      "step": 1154
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.739877589453861e-05,
      "loss": 0.5978,
      "step": 1155
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.739642184557439e-05,
      "loss": 0.5824,
      "step": 1156
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.739406779661017e-05,
      "loss": 0.6675,
      "step": 1157
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7391713747645955e-05,
      "loss": 0.5365,
      "step": 1158
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.738935969868173e-05,
      "loss": 0.5596,
      "step": 1159
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.7387005649717516e-05,
      "loss": 0.668,
      "step": 1160
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.73846516007533e-05,
      "loss": 0.639,
      "step": 1161
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.738229755178908e-05,
      "loss": 0.5796,
      "step": 1162
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.737994350282486e-05,
      "loss": 0.5954,
      "step": 1163
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.7377589453860645e-05,
      "loss": 0.5968,
      "step": 1164
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.737523540489642e-05,
      "loss": 0.5679,
      "step": 1165
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.7372881355932206e-05,
      "loss": 0.5627,
      "step": 1166
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.737052730696799e-05,
      "loss": 0.5831,
      "step": 1167
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.736817325800377e-05,
      "loss": 0.558,
      "step": 1168
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.736581920903955e-05,
      "loss": 0.5667,
      "step": 1169
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.7363465160075335e-05,
      "loss": 0.6021,
      "step": 1170
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.736111111111111e-05,
      "loss": 0.5443,
      "step": 1171
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.7358757062146896e-05,
      "loss": 0.5815,
      "step": 1172
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.735640301318268e-05,
      "loss": 0.5241,
      "step": 1173
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.735404896421846e-05,
      "loss": 0.6885,
      "step": 1174
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.735169491525424e-05,
      "loss": 0.6281,
      "step": 1175
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.734934086629002e-05,
      "loss": 0.5531,
      "step": 1176
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.73469868173258e-05,
      "loss": 0.6012,
      "step": 1177
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.7344632768361586e-05,
      "loss": 0.5027,
      "step": 1178
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.734227871939736e-05,
      "loss": 0.6676,
      "step": 1179
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.733992467043315e-05,
      "loss": 0.5181,
      "step": 1180
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.733757062146893e-05,
      "loss": 0.5288,
      "step": 1181
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.733521657250471e-05,
      "loss": 0.636,
      "step": 1182
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.733286252354049e-05,
      "loss": 0.5702,
      "step": 1183
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.7330508474576276e-05,
      "loss": 0.5442,
      "step": 1184
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.732815442561205e-05,
      "loss": 0.5597,
      "step": 1185
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.732580037664784e-05,
      "loss": 0.5684,
      "step": 1186
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.732344632768362e-05,
      "loss": 0.6185,
      "step": 1187
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.73210922787194e-05,
      "loss": 0.566,
      "step": 1188
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.731873822975518e-05,
      "loss": 0.5888,
      "step": 1189
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.7316384180790966e-05,
      "loss": 0.634,
      "step": 1190
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.731403013182674e-05,
      "loss": 0.6445,
      "step": 1191
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.731167608286253e-05,
      "loss": 0.589,
      "step": 1192
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.7309322033898304e-05,
      "loss": 0.5821,
      "step": 1193
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.730696798493409e-05,
      "loss": 0.6936,
      "step": 1194
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.730461393596987e-05,
      "loss": 0.5508,
      "step": 1195
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.730225988700565e-05,
      "loss": 0.6465,
      "step": 1196
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.729990583804143e-05,
      "loss": 0.5483,
      "step": 1197
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.729755178907722e-05,
      "loss": 0.6449,
      "step": 1198
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.7295197740112994e-05,
      "loss": 0.5852,
      "step": 1199
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.729284369114878e-05,
      "loss": 0.6646,
      "step": 1200
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.729048964218456e-05,
      "loss": 0.5391,
      "step": 1201
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.728813559322034e-05,
      "loss": 0.5553,
      "step": 1202
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.728578154425612e-05,
      "loss": 0.7029,
      "step": 1203
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.728342749529191e-05,
      "loss": 0.5199,
      "step": 1204
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.7281073446327684e-05,
      "loss": 0.5793,
      "step": 1205
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.727871939736347e-05,
      "loss": 0.6098,
      "step": 1206
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.727636534839925e-05,
      "loss": 0.6757,
      "step": 1207
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.727401129943503e-05,
      "loss": 0.6629,
      "step": 1208
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.727165725047081e-05,
      "loss": 0.546,
      "step": 1209
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.726930320150659e-05,
      "loss": 0.5626,
      "step": 1210
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.7266949152542374e-05,
      "loss": 0.6212,
      "step": 1211
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.726459510357816e-05,
      "loss": 0.5641,
      "step": 1212
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.7262241054613935e-05,
      "loss": 0.5771,
      "step": 1213
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.725988700564972e-05,
      "loss": 0.6437,
      "step": 1214
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.72575329566855e-05,
      "loss": 0.5391,
      "step": 1215
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.725517890772128e-05,
      "loss": 0.5344,
      "step": 1216
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.7252824858757064e-05,
      "loss": 0.5599,
      "step": 1217
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.725047080979285e-05,
      "loss": 0.6337,
      "step": 1218
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.7248116760828625e-05,
      "loss": 0.5976,
      "step": 1219
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.724576271186441e-05,
      "loss": 0.6028,
      "step": 1220
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.724340866290019e-05,
      "loss": 0.5056,
      "step": 1221
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.724105461393597e-05,
      "loss": 0.492,
      "step": 1222
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.7238700564971754e-05,
      "loss": 0.5863,
      "step": 1223
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.723634651600754e-05,
      "loss": 0.5092,
      "step": 1224
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7233992467043315e-05,
      "loss": 0.5164,
      "step": 1225
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.72316384180791e-05,
      "loss": 0.7087,
      "step": 1226
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7229284369114876e-05,
      "loss": 0.6225,
      "step": 1227
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.722693032015067e-05,
      "loss": 0.6362,
      "step": 1228
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7224576271186444e-05,
      "loss": 0.7482,
      "step": 1229
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.722222222222222e-05,
      "loss": 0.6262,
      "step": 1230
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7219868173258005e-05,
      "loss": 0.5321,
      "step": 1231
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.721751412429379e-05,
      "loss": 0.491,
      "step": 1232
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7215160075329566e-05,
      "loss": 0.5184,
      "step": 1233
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.721280602636535e-05,
      "loss": 0.6358,
      "step": 1234
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7210451977401134e-05,
      "loss": 0.6103,
      "step": 1235
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.720809792843691e-05,
      "loss": 0.5987,
      "step": 1236
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7205743879472695e-05,
      "loss": 0.5571,
      "step": 1237
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.720338983050848e-05,
      "loss": 0.5697,
      "step": 1238
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7201035781544256e-05,
      "loss": 0.609,
      "step": 1239
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.719868173258004e-05,
      "loss": 0.6095,
      "step": 1240
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7196327683615824e-05,
      "loss": 0.6695,
      "step": 1241
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.71939736346516e-05,
      "loss": 0.6838,
      "step": 1242
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.7191619585687385e-05,
      "loss": 0.5875,
      "step": 1243
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.718926553672316e-05,
      "loss": 0.5736,
      "step": 1244
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.718691148775895e-05,
      "loss": 0.5529,
      "step": 1245
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.718455743879473e-05,
      "loss": 0.663,
      "step": 1246
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.718220338983051e-05,
      "loss": 0.5518,
      "step": 1247
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.71798493408663e-05,
      "loss": 0.5669,
      "step": 1248
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7177495291902075e-05,
      "loss": 0.6686,
      "step": 1249
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.717514124293785e-05,
      "loss": 0.6635,
      "step": 1250
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7172787193973636e-05,
      "loss": 0.6225,
      "step": 1251
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.717043314500942e-05,
      "loss": 0.6238,
      "step": 1252
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.71680790960452e-05,
      "loss": 0.7001,
      "step": 1253
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.716572504708098e-05,
      "loss": 0.5673,
      "step": 1254
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7163370998116765e-05,
      "loss": 0.4968,
      "step": 1255
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.716101694915254e-05,
      "loss": 0.698,
      "step": 1256
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7158662900188326e-05,
      "loss": 0.5204,
      "step": 1257
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.715630885122411e-05,
      "loss": 0.5709,
      "step": 1258
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7153954802259894e-05,
      "loss": 0.5903,
      "step": 1259
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.715160075329567e-05,
      "loss": 0.605,
      "step": 1260
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.714924670433145e-05,
      "loss": 0.6389,
      "step": 1261
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.714689265536724e-05,
      "loss": 0.6055,
      "step": 1262
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7144538606403016e-05,
      "loss": 0.5545,
      "step": 1263
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.714218455743879e-05,
      "loss": 0.5946,
      "step": 1264
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.7139830508474584e-05,
      "loss": 0.5867,
      "step": 1265
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.713747645951036e-05,
      "loss": 0.589,
      "step": 1266
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.713512241054614e-05,
      "loss": 0.5086,
      "step": 1267
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.713276836158192e-05,
      "loss": 0.6328,
      "step": 1268
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.7130414312617706e-05,
      "loss": 0.5801,
      "step": 1269
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.712806026365348e-05,
      "loss": 0.5185,
      "step": 1270
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.712570621468927e-05,
      "loss": 0.6443,
      "step": 1271
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.712335216572505e-05,
      "loss": 0.5759,
      "step": 1272
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.712099811676083e-05,
      "loss": 0.5398,
      "step": 1273
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.711864406779661e-05,
      "loss": 0.609,
      "step": 1274
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.7116290018832396e-05,
      "loss": 0.4944,
      "step": 1275
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.711393596986818e-05,
      "loss": 0.5815,
      "step": 1276
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.711158192090396e-05,
      "loss": 0.541,
      "step": 1277
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.7109227871939734e-05,
      "loss": 0.6405,
      "step": 1278
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.7106873822975525e-05,
      "loss": 0.5635,
      "step": 1279
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.71045197740113e-05,
      "loss": 0.6525,
      "step": 1280
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.710216572504708e-05,
      "loss": 0.5163,
      "step": 1281
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.709981167608287e-05,
      "loss": 0.5385,
      "step": 1282
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.709745762711865e-05,
      "loss": 0.5569,
      "step": 1283
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.7095103578154424e-05,
      "loss": 0.6071,
      "step": 1284
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.709274952919021e-05,
      "loss": 0.5255,
      "step": 1285
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.709039548022599e-05,
      "loss": 0.5857,
      "step": 1286
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.708804143126177e-05,
      "loss": 0.5701,
      "step": 1287
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.708568738229755e-05,
      "loss": 0.6878,
      "step": 1288
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.708333333333334e-05,
      "loss": 0.5412,
      "step": 1289
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7080979284369114e-05,
      "loss": 0.6054,
      "step": 1290
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.70786252354049e-05,
      "loss": 0.6227,
      "step": 1291
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.707627118644068e-05,
      "loss": 0.583,
      "step": 1292
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7073917137476466e-05,
      "loss": 0.5403,
      "step": 1293
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.707156308851224e-05,
      "loss": 0.5507,
      "step": 1294
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.706920903954802e-05,
      "loss": 0.5842,
      "step": 1295
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.706685499058381e-05,
      "loss": 0.6716,
      "step": 1296
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.706450094161959e-05,
      "loss": 0.6659,
      "step": 1297
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7062146892655365e-05,
      "loss": 0.5082,
      "step": 1298
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7059792843691156e-05,
      "loss": 0.7321,
      "step": 1299
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.705743879472693e-05,
      "loss": 0.4672,
      "step": 1300
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.705508474576271e-05,
      "loss": 0.615,
      "step": 1301
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7052730696798494e-05,
      "loss": 0.5899,
      "step": 1302
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.705037664783428e-05,
      "loss": 0.6271,
      "step": 1303
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7048022598870055e-05,
      "loss": 0.6246,
      "step": 1304
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.704566854990584e-05,
      "loss": 0.6171,
      "step": 1305
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.704331450094162e-05,
      "loss": 0.6236,
      "step": 1306
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.704096045197741e-05,
      "loss": 0.5576,
      "step": 1307
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.7038606403013184e-05,
      "loss": 0.6073,
      "step": 1308
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.703625235404897e-05,
      "loss": 0.729,
      "step": 1309
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.703389830508475e-05,
      "loss": 0.5811,
      "step": 1310
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.703154425612053e-05,
      "loss": 0.5149,
      "step": 1311
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.7029190207156306e-05,
      "loss": 0.6186,
      "step": 1312
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.70268361581921e-05,
      "loss": 0.5682,
      "step": 1313
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.7024482109227874e-05,
      "loss": 0.5489,
      "step": 1314
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.702212806026365e-05,
      "loss": 0.5885,
      "step": 1315
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.701977401129944e-05,
      "loss": 0.6545,
      "step": 1316
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.701741996233522e-05,
      "loss": 0.5514,
      "step": 1317
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.7015065913370996e-05,
      "loss": 0.5942,
      "step": 1318
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.701271186440678e-05,
      "loss": 0.6081,
      "step": 1319
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.7010357815442564e-05,
      "loss": 0.5668,
      "step": 1320
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.700800376647834e-05,
      "loss": 0.5069,
      "step": 1321
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.7005649717514125e-05,
      "loss": 0.5465,
      "step": 1322
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.700329566854991e-05,
      "loss": 0.5564,
      "step": 1323
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.700094161958569e-05,
      "loss": 0.5975,
      "step": 1324
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.699858757062147e-05,
      "loss": 0.5838,
      "step": 1325
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.6996233521657254e-05,
      "loss": 0.6247,
      "step": 1326
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.699387947269304e-05,
      "loss": 0.5565,
      "step": 1327
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.6991525423728815e-05,
      "loss": 0.5917,
      "step": 1328
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.698917137476459e-05,
      "loss": 0.5666,
      "step": 1329
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.698681732580038e-05,
      "loss": 0.4917,
      "step": 1330
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.698446327683616e-05,
      "loss": 0.5886,
      "step": 1331
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.698210922787194e-05,
      "loss": 0.5358,
      "step": 1332
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.697975517890773e-05,
      "loss": 0.6064,
      "step": 1333
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6977401129943505e-05,
      "loss": 0.5263,
      "step": 1334
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.697504708097928e-05,
      "loss": 0.5591,
      "step": 1335
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6972693032015066e-05,
      "loss": 0.6053,
      "step": 1336
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.697033898305085e-05,
      "loss": 0.517,
      "step": 1337
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6967984934086634e-05,
      "loss": 0.4913,
      "step": 1338
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.696563088512241e-05,
      "loss": 0.5474,
      "step": 1339
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6963276836158195e-05,
      "loss": 0.665,
      "step": 1340
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.696092278719398e-05,
      "loss": 0.6737,
      "step": 1341
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6958568738229756e-05,
      "loss": 0.5278,
      "step": 1342
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.695621468926554e-05,
      "loss": 0.6599,
      "step": 1343
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6953860640301324e-05,
      "loss": 0.5446,
      "step": 1344
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.69515065913371e-05,
      "loss": 0.6459,
      "step": 1345
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.694915254237288e-05,
      "loss": 0.527,
      "step": 1346
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.694679849340867e-05,
      "loss": 0.6077,
      "step": 1347
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6944444444444446e-05,
      "loss": 0.4746,
      "step": 1348
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6942090395480224e-05,
      "loss": 0.5136,
      "step": 1349
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6939736346516014e-05,
      "loss": 0.6612,
      "step": 1350
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.693738229755179e-05,
      "loss": 0.6032,
      "step": 1351
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.693502824858757e-05,
      "loss": 0.4972,
      "step": 1352
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.693267419962335e-05,
      "loss": 0.5301,
      "step": 1353
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6930320150659136e-05,
      "loss": 0.5326,
      "step": 1354
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.692796610169492e-05,
      "loss": 0.5338,
      "step": 1355
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.69256120527307e-05,
      "loss": 0.6023,
      "step": 1356
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.692325800376648e-05,
      "loss": 0.6037,
      "step": 1357
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6920903954802265e-05,
      "loss": 0.5917,
      "step": 1358
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.691854990583804e-05,
      "loss": 0.623,
      "step": 1359
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6916195856873826e-05,
      "loss": 0.5088,
      "step": 1360
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.691384180790961e-05,
      "loss": 0.5978,
      "step": 1361
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.691148775894539e-05,
      "loss": 0.5528,
      "step": 1362
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6909133709981165e-05,
      "loss": 0.595,
      "step": 1363
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6906779661016955e-05,
      "loss": 0.508,
      "step": 1364
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.690442561205273e-05,
      "loss": 0.6445,
      "step": 1365
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.690207156308851e-05,
      "loss": 0.5,
      "step": 1366
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.68997175141243e-05,
      "loss": 0.6556,
      "step": 1367
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.689736346516008e-05,
      "loss": 0.614,
      "step": 1368
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6895009416195855e-05,
      "loss": 0.6806,
      "step": 1369
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.689265536723164e-05,
      "loss": 0.5394,
      "step": 1370
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.689030131826742e-05,
      "loss": 0.5802,
      "step": 1371
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6887947269303206e-05,
      "loss": 0.557,
      "step": 1372
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6885593220338983e-05,
      "loss": 0.5691,
      "step": 1373
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.688323917137477e-05,
      "loss": 0.4973,
      "step": 1374
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.688088512241055e-05,
      "loss": 0.5939,
      "step": 1375
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.687853107344633e-05,
      "loss": 0.6216,
      "step": 1376
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.687617702448211e-05,
      "loss": 0.5828,
      "step": 1377
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6873822975517896e-05,
      "loss": 0.5673,
      "step": 1378
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6871468926553673e-05,
      "loss": 0.5171,
      "step": 1379
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.686911487758945e-05,
      "loss": 0.5813,
      "step": 1380
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.686676082862524e-05,
      "loss": 0.5932,
      "step": 1381
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.686440677966102e-05,
      "loss": 0.544,
      "step": 1382
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6862052730696796e-05,
      "loss": 0.4942,
      "step": 1383
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6859698681732586e-05,
      "loss": 0.5674,
      "step": 1384
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6857344632768363e-05,
      "loss": 0.5899,
      "step": 1385
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.685499058380415e-05,
      "loss": 0.6145,
      "step": 1386
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6852636534839925e-05,
      "loss": 0.6462,
      "step": 1387
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.685028248587571e-05,
      "loss": 0.6346,
      "step": 1388
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.684792843691149e-05,
      "loss": 0.5969,
      "step": 1389
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.684557438794727e-05,
      "loss": 0.5284,
      "step": 1390
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6843220338983053e-05,
      "loss": 0.5359,
      "step": 1391
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.684086629001884e-05,
      "loss": 0.5487,
      "step": 1392
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.6838512241054615e-05,
      "loss": 0.6146,
      "step": 1393
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.68361581920904e-05,
      "loss": 0.5824,
      "step": 1394
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.683380414312618e-05,
      "loss": 0.5488,
      "step": 1395
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.683145009416196e-05,
      "loss": 0.5291,
      "step": 1396
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6829096045197743e-05,
      "loss": 0.5399,
      "step": 1397
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.682674199623353e-05,
      "loss": 0.6184,
      "step": 1398
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6824387947269305e-05,
      "loss": 0.5806,
      "step": 1399
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.682203389830508e-05,
      "loss": 0.5305,
      "step": 1400
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.681967984934087e-05,
      "loss": 0.4339,
      "step": 1401
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.681732580037665e-05,
      "loss": 0.6025,
      "step": 1402
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6814971751412433e-05,
      "loss": 0.5851,
      "step": 1403
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.681261770244821e-05,
      "loss": 0.5213,
      "step": 1404
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6810263653483995e-05,
      "loss": 0.6232,
      "step": 1405
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.680790960451978e-05,
      "loss": 0.4642,
      "step": 1406
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6805555555555556e-05,
      "loss": 0.5419,
      "step": 1407
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.680320150659134e-05,
      "loss": 0.5069,
      "step": 1408
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6800847457627123e-05,
      "loss": 0.5736,
      "step": 1409
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.67984934086629e-05,
      "loss": 0.5456,
      "step": 1410
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6796139359698685e-05,
      "loss": 0.5564,
      "step": 1411
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.679378531073447e-05,
      "loss": 0.6641,
      "step": 1412
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6791431261770246e-05,
      "loss": 0.5577,
      "step": 1413
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.678907721280603e-05,
      "loss": 0.5436,
      "step": 1414
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.6786723163841813e-05,
      "loss": 0.5574,
      "step": 1415
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.678436911487759e-05,
      "loss": 0.5577,
      "step": 1416
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.678201506591337e-05,
      "loss": 0.7128,
      "step": 1417
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.677966101694916e-05,
      "loss": 0.5329,
      "step": 1418
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6777306967984936e-05,
      "loss": 0.5772,
      "step": 1419
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.677495291902072e-05,
      "loss": 0.5135,
      "step": 1420
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.67725988700565e-05,
      "loss": 0.5366,
      "step": 1421
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.677024482109228e-05,
      "loss": 0.5643,
      "step": 1422
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6767890772128065e-05,
      "loss": 0.5443,
      "step": 1423
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.676553672316384e-05,
      "loss": 0.6548,
      "step": 1424
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6763182674199626e-05,
      "loss": 0.6396,
      "step": 1425
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.676082862523541e-05,
      "loss": 0.5966,
      "step": 1426
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.675847457627119e-05,
      "loss": 0.5314,
      "step": 1427
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.675612052730697e-05,
      "loss": 0.5912,
      "step": 1428
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6753766478342755e-05,
      "loss": 0.5156,
      "step": 1429
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.675141242937853e-05,
      "loss": 0.652,
      "step": 1430
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6749058380414316e-05,
      "loss": 0.5895,
      "step": 1431
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.67467043314501e-05,
      "loss": 0.4965,
      "step": 1432
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.674435028248588e-05,
      "loss": 0.5837,
      "step": 1433
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.674199623352166e-05,
      "loss": 0.538,
      "step": 1434
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6739642184557444e-05,
      "loss": 0.5588,
      "step": 1435
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.673728813559322e-05,
      "loss": 0.5677,
      "step": 1436
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.6734934086629006e-05,
      "loss": 0.5259,
      "step": 1437
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.673258003766478e-05,
      "loss": 0.5868,
      "step": 1438
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.673022598870057e-05,
      "loss": 0.5405,
      "step": 1439
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.672787193973635e-05,
      "loss": 0.4885,
      "step": 1440
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.672551789077213e-05,
      "loss": 0.582,
      "step": 1441
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.672316384180791e-05,
      "loss": 0.5372,
      "step": 1442
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.6720809792843696e-05,
      "loss": 0.4826,
      "step": 1443
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.671845574387947e-05,
      "loss": 0.5186,
      "step": 1444
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.671610169491526e-05,
      "loss": 0.5739,
      "step": 1445
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.671374764595104e-05,
      "loss": 0.6121,
      "step": 1446
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.671139359698682e-05,
      "loss": 0.6453,
      "step": 1447
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.67090395480226e-05,
      "loss": 0.5759,
      "step": 1448
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.6706685499058386e-05,
      "loss": 0.6055,
      "step": 1449
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.670433145009416e-05,
      "loss": 0.6069,
      "step": 1450
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.670197740112995e-05,
      "loss": 0.588,
      "step": 1451
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.669962335216573e-05,
      "loss": 0.5474,
      "step": 1452
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.669726930320151e-05,
      "loss": 0.4365,
      "step": 1453
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.669491525423729e-05,
      "loss": 0.5833,
      "step": 1454
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.669256120527307e-05,
      "loss": 0.5309,
      "step": 1455
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.669020715630885e-05,
      "loss": 0.5438,
      "step": 1456
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.668785310734464e-05,
      "loss": 0.5377,
      "step": 1457
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.6685499058380414e-05,
      "loss": 0.5706,
      "step": 1458
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.66831450094162e-05,
      "loss": 0.5105,
      "step": 1459
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.668079096045198e-05,
      "loss": 0.5592,
      "step": 1460
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.667843691148776e-05,
      "loss": 0.5307,
      "step": 1461
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.667608286252354e-05,
      "loss": 0.5415,
      "step": 1462
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.667372881355933e-05,
      "loss": 0.6803,
      "step": 1463
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6671374764595104e-05,
      "loss": 0.6094,
      "step": 1464
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.666902071563089e-05,
      "loss": 0.5512,
      "step": 1465
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.666666666666667e-05,
      "loss": 0.5559,
      "step": 1466
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.666431261770245e-05,
      "loss": 0.6089,
      "step": 1467
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.666195856873823e-05,
      "loss": 0.5936,
      "step": 1468
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6659604519774017e-05,
      "loss": 0.5317,
      "step": 1469
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6657250470809794e-05,
      "loss": 0.5147,
      "step": 1470
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.665489642184558e-05,
      "loss": 0.5708,
      "step": 1471
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6652542372881355e-05,
      "loss": 0.5279,
      "step": 1472
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.665018832391714e-05,
      "loss": 0.5711,
      "step": 1473
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.664783427495292e-05,
      "loss": 0.6292,
      "step": 1474
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.66454802259887e-05,
      "loss": 0.4902,
      "step": 1475
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6643126177024484e-05,
      "loss": 0.5894,
      "step": 1476
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.664077212806027e-05,
      "loss": 0.5197,
      "step": 1477
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6638418079096045e-05,
      "loss": 0.5927,
      "step": 1478
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.663606403013183e-05,
      "loss": 0.4987,
      "step": 1479
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.663370998116761e-05,
      "loss": 0.518,
      "step": 1480
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.663135593220339e-05,
      "loss": 0.5807,
      "step": 1481
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6629001883239174e-05,
      "loss": 0.6554,
      "step": 1482
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.662664783427496e-05,
      "loss": 0.5715,
      "step": 1483
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6624293785310735e-05,
      "loss": 0.5947,
      "step": 1484
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.662193973634652e-05,
      "loss": 0.6982,
      "step": 1485
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.66195856873823e-05,
      "loss": 0.6181,
      "step": 1486
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.661723163841808e-05,
      "loss": 0.5241,
      "step": 1487
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6614877589453864e-05,
      "loss": 0.6341,
      "step": 1488
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.661252354048965e-05,
      "loss": 0.558,
      "step": 1489
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6610169491525425e-05,
      "loss": 0.5709,
      "step": 1490
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.660781544256121e-05,
      "loss": 0.5886,
      "step": 1491
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6605461393596986e-05,
      "loss": 0.538,
      "step": 1492
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.660310734463277e-05,
      "loss": 0.6379,
      "step": 1493
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6600753295668554e-05,
      "loss": 0.5893,
      "step": 1494
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.659839924670433e-05,
      "loss": 0.5824,
      "step": 1495
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6596045197740115e-05,
      "loss": 0.7418,
      "step": 1496
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.65936911487759e-05,
      "loss": 0.5193,
      "step": 1497
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6591337099811676e-05,
      "loss": 0.6861,
      "step": 1498
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.658898305084746e-05,
      "loss": 0.4468,
      "step": 1499
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6586629001883244e-05,
      "loss": 0.5916,
      "step": 1500
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.658427495291902e-05,
      "loss": 0.5572,
      "step": 1501
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6581920903954805e-05,
      "loss": 0.6361,
      "step": 1502
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.657956685499059e-05,
      "loss": 0.5733,
      "step": 1503
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6577212806026366e-05,
      "loss": 0.6152,
      "step": 1504
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.657485875706215e-05,
      "loss": 0.6057,
      "step": 1505
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6572504708097934e-05,
      "loss": 0.592,
      "step": 1506
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.657015065913371e-05,
      "loss": 0.5369,
      "step": 1507
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6567796610169495e-05,
      "loss": 0.5064,
      "step": 1508
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.656544256120527e-05,
      "loss": 0.6483,
      "step": 1509
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6563088512241056e-05,
      "loss": 0.6141,
      "step": 1510
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.656073446327684e-05,
      "loss": 0.5604,
      "step": 1511
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.655838041431262e-05,
      "loss": 0.6031,
      "step": 1512
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.65560263653484e-05,
      "loss": 0.5185,
      "step": 1513
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6553672316384185e-05,
      "loss": 0.4911,
      "step": 1514
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.655131826741996e-05,
      "loss": 0.5682,
      "step": 1515
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6548964218455746e-05,
      "loss": 0.5867,
      "step": 1516
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.654661016949153e-05,
      "loss": 0.5916,
      "step": 1517
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.654425612052731e-05,
      "loss": 0.5794,
      "step": 1518
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.654190207156309e-05,
      "loss": 0.4399,
      "step": 1519
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6539548022598875e-05,
      "loss": 0.5883,
      "step": 1520
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.653719397363465e-05,
      "loss": 0.5694,
      "step": 1521
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6534839924670436e-05,
      "loss": 0.553,
      "step": 1522
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.653248587570622e-05,
      "loss": 0.5016,
      "step": 1523
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.6530131826742e-05,
      "loss": 0.5935,
      "step": 1524
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.652777777777778e-05,
      "loss": 0.6043,
      "step": 1525
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.652542372881356e-05,
      "loss": 0.4603,
      "step": 1526
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.652306967984934e-05,
      "loss": 0.5365,
      "step": 1527
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.6520715630885126e-05,
      "loss": 0.5654,
      "step": 1528
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.65183615819209e-05,
      "loss": 0.5401,
      "step": 1529
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.651600753295669e-05,
      "loss": 0.5475,
      "step": 1530
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.651365348399247e-05,
      "loss": 0.4467,
      "step": 1531
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.651129943502825e-05,
      "loss": 0.6001,
      "step": 1532
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.650894538606403e-05,
      "loss": 0.5331,
      "step": 1533
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.6506591337099816e-05,
      "loss": 0.4579,
      "step": 1534
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.650423728813559e-05,
      "loss": 0.5355,
      "step": 1535
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.650188323917138e-05,
      "loss": 0.5141,
      "step": 1536
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.649952919020716e-05,
      "loss": 0.5716,
      "step": 1537
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.649717514124294e-05,
      "loss": 0.6204,
      "step": 1538
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.649482109227872e-05,
      "loss": 0.5519,
      "step": 1539
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.6492467043314506e-05,
      "loss": 0.5591,
      "step": 1540
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.649011299435028e-05,
      "loss": 0.6689,
      "step": 1541
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.648775894538607e-05,
      "loss": 0.4384,
      "step": 1542
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.6485404896421844e-05,
      "loss": 0.5852,
      "step": 1543
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.6483050847457635e-05,
      "loss": 0.5256,
      "step": 1544
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.648069679849341e-05,
      "loss": 0.5894,
      "step": 1545
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.647834274952919e-05,
      "loss": 0.5893,
      "step": 1546
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.647598870056497e-05,
      "loss": 0.5797,
      "step": 1547
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.647363465160076e-05,
      "loss": 0.6019,
      "step": 1548
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.6471280602636534e-05,
      "loss": 0.6338,
      "step": 1549
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.646892655367232e-05,
      "loss": 0.6319,
      "step": 1550
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.64665725047081e-05,
      "loss": 0.5079,
      "step": 1551
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.646421845574388e-05,
      "loss": 0.5741,
      "step": 1552
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.646186440677966e-05,
      "loss": 0.5592,
      "step": 1553
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.645951035781545e-05,
      "loss": 0.5933,
      "step": 1554
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.6457156308851224e-05,
      "loss": 0.5668,
      "step": 1555
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.645480225988701e-05,
      "loss": 0.6471,
      "step": 1556
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.645244821092279e-05,
      "loss": 0.6411,
      "step": 1557
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.645009416195857e-05,
      "loss": 0.647,
      "step": 1558
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.644774011299435e-05,
      "loss": 0.578,
      "step": 1559
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.644538606403013e-05,
      "loss": 0.507,
      "step": 1560
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.644303201506592e-05,
      "loss": 0.5166,
      "step": 1561
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.64406779661017e-05,
      "loss": 0.6174,
      "step": 1562
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.6438323917137475e-05,
      "loss": 0.5634,
      "step": 1563
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.6435969868173266e-05,
      "loss": 0.5291,
      "step": 1564
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.643361581920904e-05,
      "loss": 0.5993,
      "step": 1565
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.643126177024482e-05,
      "loss": 0.562,
      "step": 1566
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6428907721280604e-05,
      "loss": 0.5551,
      "step": 1567
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.642655367231639e-05,
      "loss": 0.517,
      "step": 1568
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6424199623352165e-05,
      "loss": 0.5867,
      "step": 1569
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.642184557438795e-05,
      "loss": 0.4842,
      "step": 1570
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.641949152542373e-05,
      "loss": 0.487,
      "step": 1571
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.641713747645951e-05,
      "loss": 0.4189,
      "step": 1572
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6414783427495294e-05,
      "loss": 0.5627,
      "step": 1573
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.641242937853108e-05,
      "loss": 0.5018,
      "step": 1574
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6410075329566855e-05,
      "loss": 0.4888,
      "step": 1575
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.640772128060264e-05,
      "loss": 0.6392,
      "step": 1576
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6405367231638416e-05,
      "loss": 0.6451,
      "step": 1577
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.640301318267421e-05,
      "loss": 0.5864,
      "step": 1578
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6400659133709984e-05,
      "loss": 0.446,
      "step": 1579
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.639830508474576e-05,
      "loss": 0.5327,
      "step": 1580
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.639595103578155e-05,
      "loss": 0.5327,
      "step": 1581
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.639359698681733e-05,
      "loss": 0.6918,
      "step": 1582
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6391242937853106e-05,
      "loss": 0.5549,
      "step": 1583
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.638888888888889e-05,
      "loss": 0.5079,
      "step": 1584
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.6386534839924674e-05,
      "loss": 0.5911,
      "step": 1585
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.638418079096045e-05,
      "loss": 0.5842,
      "step": 1586
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6381826741996235e-05,
      "loss": 0.5399,
      "step": 1587
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.637947269303202e-05,
      "loss": 0.6261,
      "step": 1588
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6377118644067796e-05,
      "loss": 0.4261,
      "step": 1589
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.637476459510358e-05,
      "loss": 0.5201,
      "step": 1590
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6372410546139364e-05,
      "loss": 0.3936,
      "step": 1591
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.637005649717515e-05,
      "loss": 0.6301,
      "step": 1592
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6367702448210925e-05,
      "loss": 0.6117,
      "step": 1593
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.63653483992467e-05,
      "loss": 0.5137,
      "step": 1594
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.636299435028249e-05,
      "loss": 0.535,
      "step": 1595
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.636064030131827e-05,
      "loss": 0.44,
      "step": 1596
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.635828625235405e-05,
      "loss": 0.5099,
      "step": 1597
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.635593220338984e-05,
      "loss": 0.4961,
      "step": 1598
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6353578154425615e-05,
      "loss": 0.5484,
      "step": 1599
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.635122410546139e-05,
      "loss": 0.5525,
      "step": 1600
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6348870056497176e-05,
      "loss": 0.6866,
      "step": 1601
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.634651600753296e-05,
      "loss": 0.5344,
      "step": 1602
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.634416195856874e-05,
      "loss": 0.5865,
      "step": 1603
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.634180790960452e-05,
      "loss": 0.5989,
      "step": 1604
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6339453860640305e-05,
      "loss": 0.5942,
      "step": 1605
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.633709981167608e-05,
      "loss": 0.5786,
      "step": 1606
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.6334745762711866e-05,
      "loss": 0.4678,
      "step": 1607
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.633239171374765e-05,
      "loss": 0.5259,
      "step": 1608
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6330037664783434e-05,
      "loss": 0.46,
      "step": 1609
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.632768361581921e-05,
      "loss": 0.5637,
      "step": 1610
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.632532956685499e-05,
      "loss": 0.6315,
      "step": 1611
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.632297551789078e-05,
      "loss": 0.6605,
      "step": 1612
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6320621468926556e-05,
      "loss": 0.5671,
      "step": 1613
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.631826741996233e-05,
      "loss": 0.6173,
      "step": 1614
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6315913370998124e-05,
      "loss": 0.6155,
      "step": 1615
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.63135593220339e-05,
      "loss": 0.5156,
      "step": 1616
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.631120527306968e-05,
      "loss": 0.3922,
      "step": 1617
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.630885122410546e-05,
      "loss": 0.4846,
      "step": 1618
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6306497175141246e-05,
      "loss": 0.6476,
      "step": 1619
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.630414312617702e-05,
      "loss": 0.5539,
      "step": 1620
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.630178907721281e-05,
      "loss": 0.5212,
      "step": 1621
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.629943502824859e-05,
      "loss": 0.5484,
      "step": 1622
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6297080979284375e-05,
      "loss": 0.5172,
      "step": 1623
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.629472693032015e-05,
      "loss": 0.6693,
      "step": 1624
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6292372881355936e-05,
      "loss": 0.5385,
      "step": 1625
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.629001883239172e-05,
      "loss": 0.6476,
      "step": 1626
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.62876647834275e-05,
      "loss": 0.6714,
      "step": 1627
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.6285310734463274e-05,
      "loss": 0.4936,
      "step": 1628
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6282956685499065e-05,
      "loss": 0.5419,
      "step": 1629
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.628060263653484e-05,
      "loss": 0.4984,
      "step": 1630
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.627824858757062e-05,
      "loss": 0.5665,
      "step": 1631
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.627589453860641e-05,
      "loss": 0.4681,
      "step": 1632
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.627354048964219e-05,
      "loss": 0.6222,
      "step": 1633
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6271186440677964e-05,
      "loss": 0.6264,
      "step": 1634
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.626883239171375e-05,
      "loss": 0.6449,
      "step": 1635
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.626647834274953e-05,
      "loss": 0.5385,
      "step": 1636
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.626412429378531e-05,
      "loss": 0.6181,
      "step": 1637
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.626177024482109e-05,
      "loss": 0.7311,
      "step": 1638
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.625941619585688e-05,
      "loss": 0.5376,
      "step": 1639
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.625706214689266e-05,
      "loss": 0.5836,
      "step": 1640
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.625470809792844e-05,
      "loss": 0.5525,
      "step": 1641
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.625235404896422e-05,
      "loss": 0.5046,
      "step": 1642
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6250000000000006e-05,
      "loss": 0.5452,
      "step": 1643
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.624764595103578e-05,
      "loss": 0.5311,
      "step": 1644
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.624529190207156e-05,
      "loss": 0.5508,
      "step": 1645
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.624293785310735e-05,
      "loss": 0.5678,
      "step": 1646
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.624058380414313e-05,
      "loss": 0.4838,
      "step": 1647
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6238229755178905e-05,
      "loss": 0.6451,
      "step": 1648
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6235875706214696e-05,
      "loss": 0.5771,
      "step": 1649
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.623352165725047e-05,
      "loss": 0.5062,
      "step": 1650
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.623116760828625e-05,
      "loss": 0.5191,
      "step": 1651
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6228813559322034e-05,
      "loss": 0.4248,
      "step": 1652
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.622645951035782e-05,
      "loss": 0.5277,
      "step": 1653
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6224105461393595e-05,
      "loss": 0.492,
      "step": 1654
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.622175141242938e-05,
      "loss": 0.5218,
      "step": 1655
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.621939736346516e-05,
      "loss": 0.4534,
      "step": 1656
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.621704331450095e-05,
      "loss": 0.5334,
      "step": 1657
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6214689265536724e-05,
      "loss": 0.4917,
      "step": 1658
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.621233521657251e-05,
      "loss": 0.4903,
      "step": 1659
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.620998116760829e-05,
      "loss": 0.5469,
      "step": 1660
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.620762711864407e-05,
      "loss": 0.5628,
      "step": 1661
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6205273069679846e-05,
      "loss": 0.6198,
      "step": 1662
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.620291902071564e-05,
      "loss": 0.6251,
      "step": 1663
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6200564971751414e-05,
      "loss": 0.4735,
      "step": 1664
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.619821092278719e-05,
      "loss": 0.525,
      "step": 1665
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.619585687382298e-05,
      "loss": 0.5486,
      "step": 1666
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.619350282485876e-05,
      "loss": 0.6699,
      "step": 1667
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6191148775894536e-05,
      "loss": 0.4605,
      "step": 1668
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.618879472693032e-05,
      "loss": 0.3896,
      "step": 1669
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.6186440677966104e-05,
      "loss": 0.4928,
      "step": 1670
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.618408662900189e-05,
      "loss": 0.5022,
      "step": 1671
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6181732580037665e-05,
      "loss": 0.5689,
      "step": 1672
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.617937853107345e-05,
      "loss": 0.5116,
      "step": 1673
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.617702448210923e-05,
      "loss": 0.5958,
      "step": 1674
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.617467043314501e-05,
      "loss": 0.5797,
      "step": 1675
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6172316384180794e-05,
      "loss": 0.4507,
      "step": 1676
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.616996233521658e-05,
      "loss": 0.5781,
      "step": 1677
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6167608286252355e-05,
      "loss": 0.6061,
      "step": 1678
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.616525423728813e-05,
      "loss": 0.5463,
      "step": 1679
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.616290018832392e-05,
      "loss": 0.4648,
      "step": 1680
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.61605461393597e-05,
      "loss": 0.5288,
      "step": 1681
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.615819209039548e-05,
      "loss": 0.5572,
      "step": 1682
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.615583804143127e-05,
      "loss": 0.5575,
      "step": 1683
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6153483992467045e-05,
      "loss": 0.6124,
      "step": 1684
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.615112994350282e-05,
      "loss": 0.4943,
      "step": 1685
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6148775894538606e-05,
      "loss": 0.5201,
      "step": 1686
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.614642184557439e-05,
      "loss": 0.5574,
      "step": 1687
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6144067796610174e-05,
      "loss": 0.4204,
      "step": 1688
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.614171374764595e-05,
      "loss": 0.5169,
      "step": 1689
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6139359698681735e-05,
      "loss": 0.5216,
      "step": 1690
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.613700564971752e-05,
      "loss": 0.4594,
      "step": 1691
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.6134651600753296e-05,
      "loss": 0.6046,
      "step": 1692
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.613229755178908e-05,
      "loss": 0.5131,
      "step": 1693
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6129943502824864e-05,
      "loss": 0.6139,
      "step": 1694
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.612758945386064e-05,
      "loss": 0.4761,
      "step": 1695
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.612523540489642e-05,
      "loss": 0.5202,
      "step": 1696
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.612288135593221e-05,
      "loss": 0.5413,
      "step": 1697
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6120527306967986e-05,
      "loss": 0.5485,
      "step": 1698
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6118173258003764e-05,
      "loss": 0.601,
      "step": 1699
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6115819209039554e-05,
      "loss": 0.5966,
      "step": 1700
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.611346516007533e-05,
      "loss": 0.4134,
      "step": 1701
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6111111111111115e-05,
      "loss": 0.4647,
      "step": 1702
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.610875706214689e-05,
      "loss": 0.587,
      "step": 1703
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6106403013182676e-05,
      "loss": 0.5461,
      "step": 1704
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.610404896421846e-05,
      "loss": 0.6569,
      "step": 1705
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.610169491525424e-05,
      "loss": 0.5029,
      "step": 1706
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.609934086629002e-05,
      "loss": 0.59,
      "step": 1707
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6096986817325805e-05,
      "loss": 0.5142,
      "step": 1708
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.609463276836158e-05,
      "loss": 0.6051,
      "step": 1709
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.6092278719397366e-05,
      "loss": 0.5426,
      "step": 1710
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.608992467043315e-05,
      "loss": 0.5218,
      "step": 1711
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.608757062146893e-05,
      "loss": 0.5492,
      "step": 1712
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.608521657250471e-05,
      "loss": 0.5389,
      "step": 1713
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6082862523540495e-05,
      "loss": 0.5264,
      "step": 1714
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.608050847457627e-05,
      "loss": 0.5434,
      "step": 1715
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.607815442561205e-05,
      "loss": 0.5645,
      "step": 1716
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.607580037664784e-05,
      "loss": 0.5727,
      "step": 1717
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.607344632768362e-05,
      "loss": 0.515,
      "step": 1718
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.60710922787194e-05,
      "loss": 0.5492,
      "step": 1719
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.606873822975518e-05,
      "loss": 0.4492,
      "step": 1720
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.606638418079096e-05,
      "loss": 0.5724,
      "step": 1721
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6064030131826746e-05,
      "loss": 0.5112,
      "step": 1722
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6061676082862524e-05,
      "loss": 0.5428,
      "step": 1723
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.605932203389831e-05,
      "loss": 0.5246,
      "step": 1724
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.605696798493409e-05,
      "loss": 0.5202,
      "step": 1725
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.605461393596987e-05,
      "loss": 0.4891,
      "step": 1726
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.605225988700565e-05,
      "loss": 0.6684,
      "step": 1727
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6049905838041436e-05,
      "loss": 0.5898,
      "step": 1728
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6047551789077214e-05,
      "loss": 0.5655,
      "step": 1729
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6045197740113e-05,
      "loss": 0.5678,
      "step": 1730
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.604284369114878e-05,
      "loss": 0.5412,
      "step": 1731
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.604048964218456e-05,
      "loss": 0.503,
      "step": 1732
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6038135593220336e-05,
      "loss": 0.5223,
      "step": 1733
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6035781544256126e-05,
      "loss": 0.4668,
      "step": 1734
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.6033427495291904e-05,
      "loss": 0.5438,
      "step": 1735
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.603107344632769e-05,
      "loss": 0.5109,
      "step": 1736
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.6028719397363465e-05,
      "loss": 0.4781,
      "step": 1737
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.602636534839925e-05,
      "loss": 0.5329,
      "step": 1738
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.602401129943503e-05,
      "loss": 0.5132,
      "step": 1739
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.602165725047081e-05,
      "loss": 0.608,
      "step": 1740
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.6019303201506594e-05,
      "loss": 0.5412,
      "step": 1741
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.601694915254238e-05,
      "loss": 0.6759,
      "step": 1742
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.6014595103578155e-05,
      "loss": 0.5446,
      "step": 1743
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.601224105461394e-05,
      "loss": 0.5842,
      "step": 1744
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.600988700564972e-05,
      "loss": 0.4977,
      "step": 1745
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.60075329566855e-05,
      "loss": 0.4777,
      "step": 1746
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.6005178907721284e-05,
      "loss": 0.5347,
      "step": 1747
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.600282485875707e-05,
      "loss": 0.6622,
      "step": 1748
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.6000470809792845e-05,
      "loss": 0.5595,
      "step": 1749
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.599811676082863e-05,
      "loss": 0.5638,
      "step": 1750
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.599576271186441e-05,
      "loss": 0.5622,
      "step": 1751
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.599340866290019e-05,
      "loss": 0.5366,
      "step": 1752
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.5991054613935974e-05,
      "loss": 0.4313,
      "step": 1753
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.598870056497175e-05,
      "loss": 0.4962,
      "step": 1754
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.5986346516007535e-05,
      "loss": 0.6005,
      "step": 1755
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.598399246704332e-05,
      "loss": 0.6076,
      "step": 1756
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5981638418079096e-05,
      "loss": 0.4718,
      "step": 1757
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.597928436911488e-05,
      "loss": 0.6435,
      "step": 1758
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5976930320150664e-05,
      "loss": 0.5365,
      "step": 1759
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.597457627118644e-05,
      "loss": 0.4749,
      "step": 1760
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5972222222222225e-05,
      "loss": 0.5583,
      "step": 1761
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.596986817325801e-05,
      "loss": 0.573,
      "step": 1762
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5967514124293786e-05,
      "loss": 0.5344,
      "step": 1763
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.596516007532957e-05,
      "loss": 0.6334,
      "step": 1764
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5962806026365354e-05,
      "loss": 0.6358,
      "step": 1765
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.596045197740113e-05,
      "loss": 0.5728,
      "step": 1766
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5958097928436915e-05,
      "loss": 0.5968,
      "step": 1767
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.59557438794727e-05,
      "loss": 0.4984,
      "step": 1768
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5953389830508476e-05,
      "loss": 0.609,
      "step": 1769
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.595103578154426e-05,
      "loss": 0.4984,
      "step": 1770
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.594868173258004e-05,
      "loss": 0.6784,
      "step": 1771
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.594632768361582e-05,
      "loss": 0.5458,
      "step": 1772
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5943973634651605e-05,
      "loss": 0.569,
      "step": 1773
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.594161958568738e-05,
      "loss": 0.6591,
      "step": 1774
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5939265536723166e-05,
      "loss": 0.5717,
      "step": 1775
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.593691148775895e-05,
      "loss": 0.6267,
      "step": 1776
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.593455743879473e-05,
      "loss": 0.5508,
      "step": 1777
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.593220338983051e-05,
      "loss": 0.5777,
      "step": 1778
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.5929849340866295e-05,
      "loss": 0.4851,
      "step": 1779
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.592749529190207e-05,
      "loss": 0.4335,
      "step": 1780
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.5925141242937856e-05,
      "loss": 0.5605,
      "step": 1781
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.592278719397364e-05,
      "loss": 0.461,
      "step": 1782
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.592043314500942e-05,
      "loss": 0.5216,
      "step": 1783
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.59180790960452e-05,
      "loss": 0.6091,
      "step": 1784
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.5915725047080985e-05,
      "loss": 0.4632,
      "step": 1785
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.591337099811676e-05,
      "loss": 0.5109,
      "step": 1786
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.5911016949152546e-05,
      "loss": 0.4464,
      "step": 1787
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.590866290018832e-05,
      "loss": 0.5445,
      "step": 1788
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.590630885122411e-05,
      "loss": 0.5632,
      "step": 1789
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.590395480225989e-05,
      "loss": 0.5349,
      "step": 1790
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.590160075329567e-05,
      "loss": 0.5623,
      "step": 1791
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.589924670433145e-05,
      "loss": 0.4843,
      "step": 1792
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.5896892655367236e-05,
      "loss": 0.4473,
      "step": 1793
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.589453860640301e-05,
      "loss": 0.5419,
      "step": 1794
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.58921845574388e-05,
      "loss": 0.525,
      "step": 1795
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.588983050847458e-05,
      "loss": 0.5238,
      "step": 1796
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.588747645951036e-05,
      "loss": 0.4099,
      "step": 1797
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.588512241054614e-05,
      "loss": 0.3777,
      "step": 1798
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.5882768361581926e-05,
      "loss": 0.6051,
      "step": 1799
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.58804143126177e-05,
      "loss": 0.528,
      "step": 1800
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.587806026365349e-05,
      "loss": 0.6409,
      "step": 1801
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.587570621468927e-05,
      "loss": 0.4732,
      "step": 1802
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.587335216572505e-05,
      "loss": 0.5386,
      "step": 1803
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.587099811676083e-05,
      "loss": 0.5124,
      "step": 1804
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.5868644067796616e-05,
      "loss": 0.5123,
      "step": 1805
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.586629001883239e-05,
      "loss": 0.6335,
      "step": 1806
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.586393596986818e-05,
      "loss": 0.4779,
      "step": 1807
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.5861581920903954e-05,
      "loss": 0.4725,
      "step": 1808
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.585922787193974e-05,
      "loss": 0.5324,
      "step": 1809
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.585687382297552e-05,
      "loss": 0.4714,
      "step": 1810
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.58545197740113e-05,
      "loss": 0.489,
      "step": 1811
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.585216572504708e-05,
      "loss": 0.5309,
      "step": 1812
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.584981167608287e-05,
      "loss": 0.4759,
      "step": 1813
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.5847457627118644e-05,
      "loss": 0.7659,
      "step": 1814
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.584510357815443e-05,
      "loss": 0.5423,
      "step": 1815
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.584274952919021e-05,
      "loss": 0.6168,
      "step": 1816
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.584039548022599e-05,
      "loss": 0.4681,
      "step": 1817
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.583804143126177e-05,
      "loss": 0.6491,
      "step": 1818
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.583568738229756e-05,
      "loss": 0.5342,
      "step": 1819
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.5833333333333334e-05,
      "loss": 0.473,
      "step": 1820
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.583097928436912e-05,
      "loss": 0.565,
      "step": 1821
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.58286252354049e-05,
      "loss": 0.5584,
      "step": 1822
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.582627118644068e-05,
      "loss": 0.507,
      "step": 1823
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.582391713747646e-05,
      "loss": 0.5233,
      "step": 1824
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.582156308851224e-05,
      "loss": 0.5417,
      "step": 1825
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.5819209039548024e-05,
      "loss": 0.5781,
      "step": 1826
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.581685499058381e-05,
      "loss": 0.5609,
      "step": 1827
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.5814500941619585e-05,
      "loss": 0.4724,
      "step": 1828
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.581214689265537e-05,
      "loss": 0.5981,
      "step": 1829
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.580979284369115e-05,
      "loss": 0.5593,
      "step": 1830
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.580743879472693e-05,
      "loss": 0.5256,
      "step": 1831
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.5805084745762714e-05,
      "loss": 0.5226,
      "step": 1832
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.58027306967985e-05,
      "loss": 0.5426,
      "step": 1833
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.5800376647834275e-05,
      "loss": 0.5132,
      "step": 1834
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.579802259887006e-05,
      "loss": 0.5585,
      "step": 1835
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.579566854990584e-05,
      "loss": 0.5211,
      "step": 1836
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.579331450094162e-05,
      "loss": 0.555,
      "step": 1837
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.5790960451977404e-05,
      "loss": 0.4953,
      "step": 1838
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.578860640301319e-05,
      "loss": 0.5224,
      "step": 1839
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.5786252354048965e-05,
      "loss": 0.4644,
      "step": 1840
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.578389830508475e-05,
      "loss": 0.4857,
      "step": 1841
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5781544256120526e-05,
      "loss": 0.6456,
      "step": 1842
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.577919020715631e-05,
      "loss": 0.4836,
      "step": 1843
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5776836158192094e-05,
      "loss": 0.5852,
      "step": 1844
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.577448210922787e-05,
      "loss": 0.5206,
      "step": 1845
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5772128060263655e-05,
      "loss": 0.5786,
      "step": 1846
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.576977401129944e-05,
      "loss": 0.5667,
      "step": 1847
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5767419962335216e-05,
      "loss": 0.507,
      "step": 1848
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5765065913371e-05,
      "loss": 0.5826,
      "step": 1849
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5762711864406784e-05,
      "loss": 0.4978,
      "step": 1850
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.576035781544256e-05,
      "loss": 0.6572,
      "step": 1851
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5758003766478345e-05,
      "loss": 0.4455,
      "step": 1852
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.575564971751413e-05,
      "loss": 0.5588,
      "step": 1853
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5753295668549906e-05,
      "loss": 0.4231,
      "step": 1854
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.575094161958569e-05,
      "loss": 0.4761,
      "step": 1855
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5748587570621474e-05,
      "loss": 0.6295,
      "step": 1856
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.574623352165725e-05,
      "loss": 0.5048,
      "step": 1857
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5743879472693035e-05,
      "loss": 0.5222,
      "step": 1858
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.574152542372881e-05,
      "loss": 0.4613,
      "step": 1859
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.57391713747646e-05,
      "loss": 0.4883,
      "step": 1860
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.573681732580038e-05,
      "loss": 0.479,
      "step": 1861
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.573446327683616e-05,
      "loss": 0.5448,
      "step": 1862
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.573210922787194e-05,
      "loss": 0.5835,
      "step": 1863
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5729755178907725e-05,
      "loss": 0.4204,
      "step": 1864
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.57274011299435e-05,
      "loss": 0.5746,
      "step": 1865
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5725047080979286e-05,
      "loss": 0.602,
      "step": 1866
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.572269303201507e-05,
      "loss": 0.6809,
      "step": 1867
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.572033898305085e-05,
      "loss": 0.5444,
      "step": 1868
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.571798493408663e-05,
      "loss": 0.4741,
      "step": 1869
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5715630885122415e-05,
      "loss": 0.3841,
      "step": 1870
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.571327683615819e-05,
      "loss": 0.5508,
      "step": 1871
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5710922787193976e-05,
      "loss": 0.4713,
      "step": 1872
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.570856873822976e-05,
      "loss": 0.4536,
      "step": 1873
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.570621468926554e-05,
      "loss": 0.4478,
      "step": 1874
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.570386064030132e-05,
      "loss": 0.4871,
      "step": 1875
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.57015065913371e-05,
      "loss": 0.5409,
      "step": 1876
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.569915254237289e-05,
      "loss": 0.6287,
      "step": 1877
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5696798493408666e-05,
      "loss": 0.7449,
      "step": 1878
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.569444444444444e-05,
      "loss": 0.4984,
      "step": 1879
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5692090395480234e-05,
      "loss": 0.4605,
      "step": 1880
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.568973634651601e-05,
      "loss": 0.5489,
      "step": 1881
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.568738229755179e-05,
      "loss": 0.6178,
      "step": 1882
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.568502824858757e-05,
      "loss": 0.4879,
      "step": 1883
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.5682674199623356e-05,
      "loss": 0.4676,
      "step": 1884
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.568032015065913e-05,
      "loss": 0.6087,
      "step": 1885
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.567796610169492e-05,
      "loss": 0.5174,
      "step": 1886
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.56756120527307e-05,
      "loss": 0.5617,
      "step": 1887
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.567325800376648e-05,
      "loss": 0.5014,
      "step": 1888
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.567090395480226e-05,
      "loss": 0.4638,
      "step": 1889
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.5668549905838046e-05,
      "loss": 0.6897,
      "step": 1890
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.566619585687382e-05,
      "loss": 0.6515,
      "step": 1891
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.566384180790961e-05,
      "loss": 0.5315,
      "step": 1892
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.5661487758945384e-05,
      "loss": 0.5721,
      "step": 1893
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.5659133709981175e-05,
      "loss": 0.4255,
      "step": 1894
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.565677966101695e-05,
      "loss": 0.5368,
      "step": 1895
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.565442561205273e-05,
      "loss": 0.5889,
      "step": 1896
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.565207156308852e-05,
      "loss": 0.6,
      "step": 1897
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.56497175141243e-05,
      "loss": 0.5081,
      "step": 1898
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.5647363465160074e-05,
      "loss": 0.5107,
      "step": 1899
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.564500941619586e-05,
      "loss": 0.5609,
      "step": 1900
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.564265536723164e-05,
      "loss": 0.4925,
      "step": 1901
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.564030131826742e-05,
      "loss": 0.5374,
      "step": 1902
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.56379472693032e-05,
      "loss": 0.6085,
      "step": 1903
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.563559322033899e-05,
      "loss": 0.4408,
      "step": 1904
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.5633239171374764e-05,
      "loss": 0.4743,
      "step": 1905
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.563088512241055e-05,
      "loss": 0.6174,
      "step": 1906
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.562853107344633e-05,
      "loss": 0.5594,
      "step": 1907
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5626177024482116e-05,
      "loss": 0.5383,
      "step": 1908
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.562382297551789e-05,
      "loss": 0.4631,
      "step": 1909
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.562146892655367e-05,
      "loss": 0.4721,
      "step": 1910
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.561911487758946e-05,
      "loss": 0.4373,
      "step": 1911
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.561676082862524e-05,
      "loss": 0.6091,
      "step": 1912
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5614406779661015e-05,
      "loss": 0.579,
      "step": 1913
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5612052730696806e-05,
      "loss": 0.5497,
      "step": 1914
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.560969868173258e-05,
      "loss": 0.5186,
      "step": 1915
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.560734463276836e-05,
      "loss": 0.4943,
      "step": 1916
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5604990583804144e-05,
      "loss": 0.5114,
      "step": 1917
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.560263653483993e-05,
      "loss": 0.5721,
      "step": 1918
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5600282485875705e-05,
      "loss": 0.6207,
      "step": 1919
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.559792843691149e-05,
      "loss": 0.5621,
      "step": 1920
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.559557438794727e-05,
      "loss": 0.5615,
      "step": 1921
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.559322033898305e-05,
      "loss": 0.6848,
      "step": 1922
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5590866290018834e-05,
      "loss": 0.5786,
      "step": 1923
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.558851224105462e-05,
      "loss": 0.4635,
      "step": 1924
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.55861581920904e-05,
      "loss": 0.5264,
      "step": 1925
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.558380414312618e-05,
      "loss": 0.6074,
      "step": 1926
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5581450094161956e-05,
      "loss": 0.533,
      "step": 1927
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.557909604519775e-05,
      "loss": 0.6225,
      "step": 1928
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5576741996233524e-05,
      "loss": 0.6496,
      "step": 1929
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.55743879472693e-05,
      "loss": 0.5514,
      "step": 1930
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.557203389830509e-05,
      "loss": 0.42,
      "step": 1931
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.556967984934087e-05,
      "loss": 0.5256,
      "step": 1932
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5567325800376646e-05,
      "loss": 0.5584,
      "step": 1933
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.556497175141243e-05,
      "loss": 0.5683,
      "step": 1934
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5562617702448214e-05,
      "loss": 0.4415,
      "step": 1935
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.556026365348399e-05,
      "loss": 0.5844,
      "step": 1936
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5557909604519775e-05,
      "loss": 0.5842,
      "step": 1937
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.555555555555556e-05,
      "loss": 0.465,
      "step": 1938
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.555320150659134e-05,
      "loss": 0.4962,
      "step": 1939
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.555084745762712e-05,
      "loss": 0.5059,
      "step": 1940
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5548493408662904e-05,
      "loss": 0.5529,
      "step": 1941
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.554613935969869e-05,
      "loss": 0.6004,
      "step": 1942
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5543785310734465e-05,
      "loss": 0.4674,
      "step": 1943
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.554143126177024e-05,
      "loss": 0.4757,
      "step": 1944
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.553907721280603e-05,
      "loss": 0.591,
      "step": 1945
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.553672316384181e-05,
      "loss": 0.531,
      "step": 1946
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.553436911487759e-05,
      "loss": 0.4563,
      "step": 1947
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.553201506591338e-05,
      "loss": 0.4935,
      "step": 1948
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5529661016949155e-05,
      "loss": 0.4833,
      "step": 1949
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.552730696798493e-05,
      "loss": 0.4974,
      "step": 1950
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5524952919020716e-05,
      "loss": 0.4191,
      "step": 1951
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.55225988700565e-05,
      "loss": 0.4672,
      "step": 1952
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.552024482109228e-05,
      "loss": 0.5292,
      "step": 1953
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.551789077212806e-05,
      "loss": 0.4765,
      "step": 1954
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5515536723163845e-05,
      "loss": 0.5042,
      "step": 1955
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.551318267419963e-05,
      "loss": 0.4545,
      "step": 1956
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5510828625235406e-05,
      "loss": 0.4725,
      "step": 1957
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.550847457627119e-05,
      "loss": 0.4512,
      "step": 1958
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5506120527306974e-05,
      "loss": 0.456,
      "step": 1959
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.550376647834275e-05,
      "loss": 0.4093,
      "step": 1960
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.550141242937853e-05,
      "loss": 0.3909,
      "step": 1961
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.549905838041432e-05,
      "loss": 0.4886,
      "step": 1962
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5496704331450096e-05,
      "loss": 0.5867,
      "step": 1963
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.549435028248587e-05,
      "loss": 0.4676,
      "step": 1964
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5491996233521664e-05,
      "loss": 0.6013,
      "step": 1965
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.548964218455744e-05,
      "loss": 0.4536,
      "step": 1966
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.548728813559322e-05,
      "loss": 0.4021,
      "step": 1967
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5484934086629e-05,
      "loss": 0.6665,
      "step": 1968
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.5482580037664786e-05,
      "loss": 0.6261,
      "step": 1969
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.548022598870056e-05,
      "loss": 0.5359,
      "step": 1970
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.547787193973635e-05,
      "loss": 0.5207,
      "step": 1971
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.547551789077213e-05,
      "loss": 0.4952,
      "step": 1972
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.5473163841807915e-05,
      "loss": 0.4852,
      "step": 1973
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.547080979284369e-05,
      "loss": 0.5115,
      "step": 1974
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.5468455743879476e-05,
      "loss": 0.505,
      "step": 1975
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.546610169491526e-05,
      "loss": 0.4887,
      "step": 1976
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.546374764595104e-05,
      "loss": 0.5663,
      "step": 1977
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.5461393596986814e-05,
      "loss": 0.4554,
      "step": 1978
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.5459039548022605e-05,
      "loss": 0.5218,
      "step": 1979
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.545668549905838e-05,
      "loss": 0.5417,
      "step": 1980
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.545433145009416e-05,
      "loss": 0.4331,
      "step": 1981
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.545197740112995e-05,
      "loss": 0.626,
      "step": 1982
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.544962335216573e-05,
      "loss": 0.4845,
      "step": 1983
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.5447269303201504e-05,
      "loss": 0.5004,
      "step": 1984
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.544491525423729e-05,
      "loss": 0.5016,
      "step": 1985
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.544256120527307e-05,
      "loss": 0.5263,
      "step": 1986
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.5440207156308856e-05,
      "loss": 0.5312,
      "step": 1987
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.543785310734463e-05,
      "loss": 0.4283,
      "step": 1988
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.543549905838042e-05,
      "loss": 0.4779,
      "step": 1989
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.54331450094162e-05,
      "loss": 0.5588,
      "step": 1990
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.543079096045198e-05,
      "loss": 0.6557,
      "step": 1991
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.542843691148776e-05,
      "loss": 0.6026,
      "step": 1992
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.5426082862523546e-05,
      "loss": 0.5201,
      "step": 1993
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.542372881355932e-05,
      "loss": 0.5253,
      "step": 1994
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.54213747645951e-05,
      "loss": 0.4726,
      "step": 1995
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.541902071563089e-05,
      "loss": 0.6189,
      "step": 1996
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.541666666666667e-05,
      "loss": 0.5074,
      "step": 1997
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.5414312617702445e-05,
      "loss": 0.4144,
      "step": 1998
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.5411958568738236e-05,
      "loss": 0.484,
      "step": 1999
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.540960451977401e-05,
      "loss": 0.4458,
      "step": 2000
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.540725047080979e-05,
      "loss": 0.4966,
      "step": 2001
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.5404896421845574e-05,
      "loss": 0.5718,
      "step": 2002
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.540254237288136e-05,
      "loss": 0.416,
      "step": 2003
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.540018832391714e-05,
      "loss": 0.4647,
      "step": 2004
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.539783427495292e-05,
      "loss": 0.5042,
      "step": 2005
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.53954802259887e-05,
      "loss": 0.5624,
      "step": 2006
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.539312617702449e-05,
      "loss": 0.4273,
      "step": 2007
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.5390772128060264e-05,
      "loss": 0.5558,
      "step": 2008
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.538841807909605e-05,
      "loss": 0.5369,
      "step": 2009
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.538606403013183e-05,
      "loss": 0.5198,
      "step": 2010
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.538370998116761e-05,
      "loss": 0.6098,
      "step": 2011
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.5381355932203387e-05,
      "loss": 0.4313,
      "step": 2012
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.537900188323918e-05,
      "loss": 0.4835,
      "step": 2013
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.5376647834274954e-05,
      "loss": 0.4478,
      "step": 2014
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.537429378531073e-05,
      "loss": 0.5029,
      "step": 2015
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.537193973634652e-05,
      "loss": 0.5363,
      "step": 2016
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.53695856873823e-05,
      "loss": 0.4664,
      "step": 2017
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.536723163841808e-05,
      "loss": 0.6602,
      "step": 2018
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.536487758945386e-05,
      "loss": 0.578,
      "step": 2019
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.5362523540489644e-05,
      "loss": 0.6891,
      "step": 2020
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.536016949152543e-05,
      "loss": 0.4884,
      "step": 2021
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.5357815442561205e-05,
      "loss": 0.5316,
      "step": 2022
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.535546139359699e-05,
      "loss": 0.6828,
      "step": 2023
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.535310734463277e-05,
      "loss": 0.4627,
      "step": 2024
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.535075329566855e-05,
      "loss": 0.4302,
      "step": 2025
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.5348399246704334e-05,
      "loss": 0.4515,
      "step": 2026
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.534604519774012e-05,
      "loss": 0.6073,
      "step": 2027
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.5343691148775895e-05,
      "loss": 0.5593,
      "step": 2028
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.534133709981167e-05,
      "loss": 0.5023,
      "step": 2029
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.533898305084746e-05,
      "loss": 0.5828,
      "step": 2030
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.533662900188324e-05,
      "loss": 0.5086,
      "step": 2031
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.533427495291902e-05,
      "loss": 0.5007,
      "step": 2032
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.533192090395481e-05,
      "loss": 0.5566,
      "step": 2033
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5329566854990585e-05,
      "loss": 0.5712,
      "step": 2034
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.532721280602637e-05,
      "loss": 0.578,
      "step": 2035
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5324858757062147e-05,
      "loss": 0.5718,
      "step": 2036
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.532250470809793e-05,
      "loss": 0.665,
      "step": 2037
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5320150659133714e-05,
      "loss": 0.5046,
      "step": 2038
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.531779661016949e-05,
      "loss": 0.5559,
      "step": 2039
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5315442561205275e-05,
      "loss": 0.4482,
      "step": 2040
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.531308851224106e-05,
      "loss": 0.5763,
      "step": 2041
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5310734463276837e-05,
      "loss": 0.5886,
      "step": 2042
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.530838041431262e-05,
      "loss": 0.6628,
      "step": 2043
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5306026365348404e-05,
      "loss": 0.4722,
      "step": 2044
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.530367231638418e-05,
      "loss": 0.486,
      "step": 2045
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5301318267419965e-05,
      "loss": 0.4965,
      "step": 2046
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.529896421845575e-05,
      "loss": 0.5371,
      "step": 2047
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5296610169491527e-05,
      "loss": 0.6226,
      "step": 2048
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5294256120527304e-05,
      "loss": 0.5445,
      "step": 2049
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5291902071563094e-05,
      "loss": 0.4366,
      "step": 2050
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.528954802259887e-05,
      "loss": 0.6041,
      "step": 2051
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5287193973634655e-05,
      "loss": 0.5117,
      "step": 2052
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.528483992467043e-05,
      "loss": 0.5056,
      "step": 2053
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.5282485875706217e-05,
      "loss": 0.5244,
      "step": 2054
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5280131826742e-05,
      "loss": 0.552,
      "step": 2055
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.527777777777778e-05,
      "loss": 0.4876,
      "step": 2056
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.527542372881356e-05,
      "loss": 0.5854,
      "step": 2057
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5273069679849345e-05,
      "loss": 0.4941,
      "step": 2058
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.527071563088512e-05,
      "loss": 0.6248,
      "step": 2059
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5268361581920906e-05,
      "loss": 0.535,
      "step": 2060
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.526600753295669e-05,
      "loss": 0.51,
      "step": 2061
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.526365348399247e-05,
      "loss": 0.4018,
      "step": 2062
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.526129943502825e-05,
      "loss": 0.587,
      "step": 2063
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5258945386064035e-05,
      "loss": 0.5394,
      "step": 2064
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.525659133709981e-05,
      "loss": 0.4587,
      "step": 2065
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5254237288135596e-05,
      "loss": 0.4922,
      "step": 2066
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.525188323917138e-05,
      "loss": 0.466,
      "step": 2067
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.524952919020716e-05,
      "loss": 0.4787,
      "step": 2068
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.524717514124294e-05,
      "loss": 0.48,
      "step": 2069
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.524482109227872e-05,
      "loss": 0.4885,
      "step": 2070
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.52424670433145e-05,
      "loss": 0.636,
      "step": 2071
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5240112994350286e-05,
      "loss": 0.5474,
      "step": 2072
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5237758945386064e-05,
      "loss": 0.6022,
      "step": 2073
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.523540489642185e-05,
      "loss": 0.3503,
      "step": 2074
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.523305084745763e-05,
      "loss": 0.52,
      "step": 2075
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.523069679849341e-05,
      "loss": 0.7183,
      "step": 2076
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.522834274952919e-05,
      "loss": 0.5031,
      "step": 2077
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5225988700564976e-05,
      "loss": 0.525,
      "step": 2078
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5223634651600754e-05,
      "loss": 0.426,
      "step": 2079
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.522128060263654e-05,
      "loss": 0.4216,
      "step": 2080
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.521892655367232e-05,
      "loss": 0.5084,
      "step": 2081
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.52165725047081e-05,
      "loss": 0.6203,
      "step": 2082
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.521421845574388e-05,
      "loss": 0.6199,
      "step": 2083
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5211864406779666e-05,
      "loss": 0.57,
      "step": 2084
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5209510357815444e-05,
      "loss": 0.4302,
      "step": 2085
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.520715630885123e-05,
      "loss": 0.4752,
      "step": 2086
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5204802259887005e-05,
      "loss": 0.5117,
      "step": 2087
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.520244821092279e-05,
      "loss": 0.5092,
      "step": 2088
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.520009416195857e-05,
      "loss": 0.5619,
      "step": 2089
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.519774011299435e-05,
      "loss": 0.6229,
      "step": 2090
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5195386064030134e-05,
      "loss": 0.4522,
      "step": 2091
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.519303201506592e-05,
      "loss": 0.6451,
      "step": 2092
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5190677966101695e-05,
      "loss": 0.5271,
      "step": 2093
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.518832391713748e-05,
      "loss": 0.5021,
      "step": 2094
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.518596986817326e-05,
      "loss": 0.4549,
      "step": 2095
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.518361581920904e-05,
      "loss": 0.5435,
      "step": 2096
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.5181261770244824e-05,
      "loss": 0.4428,
      "step": 2097
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.517890772128061e-05,
      "loss": 0.4882,
      "step": 2098
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5176553672316385e-05,
      "loss": 0.5213,
      "step": 2099
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.517419962335217e-05,
      "loss": 0.4972,
      "step": 2100
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.517184557438795e-05,
      "loss": 0.5481,
      "step": 2101
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.516949152542373e-05,
      "loss": 0.4411,
      "step": 2102
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5167137476459514e-05,
      "loss": 0.5125,
      "step": 2103
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.516478342749529e-05,
      "loss": 0.5153,
      "step": 2104
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5162429378531075e-05,
      "loss": 0.4678,
      "step": 2105
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.516007532956686e-05,
      "loss": 0.4839,
      "step": 2106
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5157721280602636e-05,
      "loss": 0.5742,
      "step": 2107
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.515536723163842e-05,
      "loss": 0.5021,
      "step": 2108
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5153013182674204e-05,
      "loss": 0.4891,
      "step": 2109
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.515065913370998e-05,
      "loss": 0.5039,
      "step": 2110
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5148305084745765e-05,
      "loss": 0.433,
      "step": 2111
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.514595103578155e-05,
      "loss": 0.59,
      "step": 2112
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5143596986817326e-05,
      "loss": 0.5469,
      "step": 2113
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.514124293785311e-05,
      "loss": 0.5549,
      "step": 2114
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5138888888888894e-05,
      "loss": 0.4061,
      "step": 2115
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.513653483992467e-05,
      "loss": 0.5515,
      "step": 2116
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.5134180790960455e-05,
      "loss": 0.4851,
      "step": 2117
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.513182674199624e-05,
      "loss": 0.453,
      "step": 2118
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5129472693032016e-05,
      "loss": 0.4965,
      "step": 2119
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.51271186440678e-05,
      "loss": 0.3845,
      "step": 2120
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5124764595103584e-05,
      "loss": 0.5538,
      "step": 2121
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.512241054613936e-05,
      "loss": 0.5994,
      "step": 2122
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5120056497175145e-05,
      "loss": 0.515,
      "step": 2123
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.511770244821092e-05,
      "loss": 0.4969,
      "step": 2124
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5115348399246706e-05,
      "loss": 0.5056,
      "step": 2125
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.511299435028249e-05,
      "loss": 0.4451,
      "step": 2126
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.511064030131827e-05,
      "loss": 0.4218,
      "step": 2127
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.510828625235405e-05,
      "loss": 0.569,
      "step": 2128
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5105932203389835e-05,
      "loss": 0.453,
      "step": 2129
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7692714298299709,
      "eval_loss": 0.48256269097328186,
      "eval_runtime": 23.104,
      "eval_samples_per_second": 1473.902,
      "eval_steps_per_second": 23.07,
      "step": 2129
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.510357815442561e-05,
      "loss": 0.5826,
      "step": 2130
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5101224105461396e-05,
      "loss": 0.4543,
      "step": 2131
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.509887005649718e-05,
      "loss": 0.4577,
      "step": 2132
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.509651600753296e-05,
      "loss": 0.4586,
      "step": 2133
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.509416195856874e-05,
      "loss": 0.3438,
      "step": 2134
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5091807909604525e-05,
      "loss": 0.4342,
      "step": 2135
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.50894538606403e-05,
      "loss": 0.3943,
      "step": 2136
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.5087099811676086e-05,
      "loss": 0.4815,
      "step": 2137
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.508474576271187e-05,
      "loss": 0.3903,
      "step": 2138
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.508239171374765e-05,
      "loss": 0.5026,
      "step": 2139
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.508003766478343e-05,
      "loss": 0.3979,
      "step": 2140
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.507768361581921e-05,
      "loss": 0.4111,
      "step": 2141
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.507532956685499e-05,
      "loss": 0.3912,
      "step": 2142
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.5072975517890776e-05,
      "loss": 0.4245,
      "step": 2143
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.507062146892655e-05,
      "loss": 0.416,
      "step": 2144
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.506826741996234e-05,
      "loss": 0.3512,
      "step": 2145
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.506591337099812e-05,
      "loss": 0.3775,
      "step": 2146
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.50635593220339e-05,
      "loss": 0.4773,
      "step": 2147
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.506120527306968e-05,
      "loss": 0.4094,
      "step": 2148
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.5058851224105466e-05,
      "loss": 0.5193,
      "step": 2149
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.505649717514124e-05,
      "loss": 0.5734,
      "step": 2150
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.505414312617703e-05,
      "loss": 0.4035,
      "step": 2151
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.505178907721281e-05,
      "loss": 0.4633,
      "step": 2152
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.504943502824859e-05,
      "loss": 0.3238,
      "step": 2153
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.504708097928437e-05,
      "loss": 0.3561,
      "step": 2154
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.5044726930320156e-05,
      "loss": 0.4366,
      "step": 2155
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.504237288135593e-05,
      "loss": 0.4711,
      "step": 2156
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.504001883239172e-05,
      "loss": 0.4084,
      "step": 2157
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.5037664783427494e-05,
      "loss": 0.3337,
      "step": 2158
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.503531073446328e-05,
      "loss": 0.3021,
      "step": 2159
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.503295668549906e-05,
      "loss": 0.3732,
      "step": 2160
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.503060263653484e-05,
      "loss": 0.4668,
      "step": 2161
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.502824858757062e-05,
      "loss": 0.5012,
      "step": 2162
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.502589453860641e-05,
      "loss": 0.3236,
      "step": 2163
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.5023540489642184e-05,
      "loss": 0.4898,
      "step": 2164
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.502118644067797e-05,
      "loss": 0.4922,
      "step": 2165
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.501883239171375e-05,
      "loss": 0.3428,
      "step": 2166
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.501647834274953e-05,
      "loss": 0.4079,
      "step": 2167
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.501412429378531e-05,
      "loss": 0.329,
      "step": 2168
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.50117702448211e-05,
      "loss": 0.3665,
      "step": 2169
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.5009416195856874e-05,
      "loss": 0.5566,
      "step": 2170
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.500706214689266e-05,
      "loss": 0.4378,
      "step": 2171
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.500470809792844e-05,
      "loss": 0.3564,
      "step": 2172
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.500235404896422e-05,
      "loss": 0.4779,
      "step": 2173
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.5e-05,
      "loss": 0.5447,
      "step": 2174
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.499764595103578e-05,
      "loss": 0.3601,
      "step": 2175
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.499529190207157e-05,
      "loss": 0.4555,
      "step": 2176
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.499293785310735e-05,
      "loss": 0.3108,
      "step": 2177
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.4990583804143125e-05,
      "loss": 0.4777,
      "step": 2178
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.498822975517891e-05,
      "loss": 0.3227,
      "step": 2179
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.498587570621469e-05,
      "loss": 0.46,
      "step": 2180
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.498352165725047e-05,
      "loss": 0.271,
      "step": 2181
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.4981167608286254e-05,
      "loss": 0.4699,
      "step": 2182
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.497881355932204e-05,
      "loss": 0.3364,
      "step": 2183
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4976459510357815e-05,
      "loss": 0.4911,
      "step": 2184
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.49741054613936e-05,
      "loss": 0.5416,
      "step": 2185
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.497175141242938e-05,
      "loss": 0.4678,
      "step": 2186
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.496939736346516e-05,
      "loss": 0.3486,
      "step": 2187
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4967043314500944e-05,
      "loss": 0.4157,
      "step": 2188
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.496468926553673e-05,
      "loss": 0.3899,
      "step": 2189
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4962335216572505e-05,
      "loss": 0.4132,
      "step": 2190
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.495998116760829e-05,
      "loss": 0.4897,
      "step": 2191
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4957627118644066e-05,
      "loss": 0.4184,
      "step": 2192
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.495527306967986e-05,
      "loss": 0.4542,
      "step": 2193
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4952919020715634e-05,
      "loss": 0.4383,
      "step": 2194
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.495056497175141e-05,
      "loss": 0.2769,
      "step": 2195
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.49482109227872e-05,
      "loss": 0.3544,
      "step": 2196
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.494585687382298e-05,
      "loss": 0.3046,
      "step": 2197
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4943502824858756e-05,
      "loss": 0.4755,
      "step": 2198
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.494114877589454e-05,
      "loss": 0.3866,
      "step": 2199
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4938794726930324e-05,
      "loss": 0.3913,
      "step": 2200
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.49364406779661e-05,
      "loss": 0.3526,
      "step": 2201
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4934086629001885e-05,
      "loss": 0.4299,
      "step": 2202
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.493173258003767e-05,
      "loss": 0.3865,
      "step": 2203
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4929378531073446e-05,
      "loss": 0.3614,
      "step": 2204
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.492702448210923e-05,
      "loss": 0.4417,
      "step": 2205
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4924670433145014e-05,
      "loss": 0.3917,
      "step": 2206
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.492231638418079e-05,
      "loss": 0.4419,
      "step": 2207
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4919962335216575e-05,
      "loss": 0.2903,
      "step": 2208
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.491760828625235e-05,
      "loss": 0.3632,
      "step": 2209
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.491525423728814e-05,
      "loss": 0.4353,
      "step": 2210
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.491290018832392e-05,
      "loss": 0.4589,
      "step": 2211
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.49105461393597e-05,
      "loss": 0.2857,
      "step": 2212
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.490819209039549e-05,
      "loss": 0.3801,
      "step": 2213
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4905838041431265e-05,
      "loss": 0.2525,
      "step": 2214
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.490348399246704e-05,
      "loss": 0.4163,
      "step": 2215
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4901129943502826e-05,
      "loss": 0.5561,
      "step": 2216
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.489877589453861e-05,
      "loss": 0.4344,
      "step": 2217
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.489642184557439e-05,
      "loss": 0.4921,
      "step": 2218
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.489406779661017e-05,
      "loss": 0.5735,
      "step": 2219
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4891713747645955e-05,
      "loss": 0.369,
      "step": 2220
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.488935969868173e-05,
      "loss": 0.4306,
      "step": 2221
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4887005649717516e-05,
      "loss": 0.4158,
      "step": 2222
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.48846516007533e-05,
      "loss": 0.4718,
      "step": 2223
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.4882297551789084e-05,
      "loss": 0.3204,
      "step": 2224
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.487994350282486e-05,
      "loss": 0.3638,
      "step": 2225
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.487758945386064e-05,
      "loss": 0.3554,
      "step": 2226
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.487523540489643e-05,
      "loss": 0.3619,
      "step": 2227
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.4872881355932206e-05,
      "loss": 0.3658,
      "step": 2228
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.487052730696798e-05,
      "loss": 0.3726,
      "step": 2229
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.4868173258003774e-05,
      "loss": 0.3754,
      "step": 2230
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.486581920903955e-05,
      "loss": 0.4727,
      "step": 2231
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.486346516007533e-05,
      "loss": 0.4763,
      "step": 2232
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.486111111111111e-05,
      "loss": 0.294,
      "step": 2233
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.4858757062146896e-05,
      "loss": 0.5063,
      "step": 2234
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.485640301318267e-05,
      "loss": 0.4684,
      "step": 2235
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.485404896421846e-05,
      "loss": 0.5026,
      "step": 2236
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.485169491525424e-05,
      "loss": 0.2644,
      "step": 2237
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.484934086629002e-05,
      "loss": 0.4787,
      "step": 2238
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.48469868173258e-05,
      "loss": 0.4466,
      "step": 2239
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.4844632768361586e-05,
      "loss": 0.3801,
      "step": 2240
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.484227871939737e-05,
      "loss": 0.3906,
      "step": 2241
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.483992467043315e-05,
      "loss": 0.3427,
      "step": 2242
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.4837570621468924e-05,
      "loss": 0.4057,
      "step": 2243
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.4835216572504715e-05,
      "loss": 0.3435,
      "step": 2244
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.483286252354049e-05,
      "loss": 0.3852,
      "step": 2245
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.483050847457627e-05,
      "loss": 0.3455,
      "step": 2246
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.482815442561206e-05,
      "loss": 0.3675,
      "step": 2247
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.482580037664784e-05,
      "loss": 0.4146,
      "step": 2248
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.4823446327683614e-05,
      "loss": 0.3326,
      "step": 2249
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.48210922787194e-05,
      "loss": 0.3578,
      "step": 2250
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.481873822975518e-05,
      "loss": 0.3375,
      "step": 2251
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.481638418079096e-05,
      "loss": 0.6612,
      "step": 2252
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.481403013182674e-05,
      "loss": 0.4411,
      "step": 2253
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.481167608286253e-05,
      "loss": 0.3547,
      "step": 2254
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.480932203389831e-05,
      "loss": 0.2948,
      "step": 2255
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.480696798493409e-05,
      "loss": 0.3389,
      "step": 2256
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.480461393596987e-05,
      "loss": 0.3237,
      "step": 2257
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.4802259887005656e-05,
      "loss": 0.3854,
      "step": 2258
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.479990583804143e-05,
      "loss": 0.4024,
      "step": 2259
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.479755178907721e-05,
      "loss": 0.4036,
      "step": 2260
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.4795197740113e-05,
      "loss": 0.5163,
      "step": 2261
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.479284369114878e-05,
      "loss": 0.3311,
      "step": 2262
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.4790489642184555e-05,
      "loss": 0.3554,
      "step": 2263
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.4788135593220346e-05,
      "loss": 0.4476,
      "step": 2264
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.478578154425612e-05,
      "loss": 0.2699,
      "step": 2265
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.47834274952919e-05,
      "loss": 0.3356,
      "step": 2266
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.4781073446327684e-05,
      "loss": 0.4194,
      "step": 2267
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.477871939736347e-05,
      "loss": 0.4427,
      "step": 2268
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4776365348399245e-05,
      "loss": 0.3943,
      "step": 2269
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.477401129943503e-05,
      "loss": 0.5866,
      "step": 2270
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.477165725047081e-05,
      "loss": 0.4156,
      "step": 2271
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.47693032015066e-05,
      "loss": 0.4239,
      "step": 2272
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4766949152542374e-05,
      "loss": 0.4312,
      "step": 2273
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.476459510357816e-05,
      "loss": 0.4215,
      "step": 2274
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.476224105461394e-05,
      "loss": 0.3943,
      "step": 2275
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.475988700564972e-05,
      "loss": 0.4238,
      "step": 2276
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4757532956685496e-05,
      "loss": 0.3821,
      "step": 2277
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.475517890772129e-05,
      "loss": 0.4069,
      "step": 2278
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4752824858757064e-05,
      "loss": 0.5432,
      "step": 2279
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.475047080979284e-05,
      "loss": 0.3796,
      "step": 2280
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.474811676082863e-05,
      "loss": 0.4567,
      "step": 2281
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.474576271186441e-05,
      "loss": 0.4138,
      "step": 2282
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4743408662900186e-05,
      "loss": 0.4334,
      "step": 2283
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.474105461393597e-05,
      "loss": 0.34,
      "step": 2284
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4738700564971754e-05,
      "loss": 0.5009,
      "step": 2285
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.473634651600753e-05,
      "loss": 0.4199,
      "step": 2286
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4733992467043315e-05,
      "loss": 0.3497,
      "step": 2287
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.47316384180791e-05,
      "loss": 0.4368,
      "step": 2288
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.472928436911488e-05,
      "loss": 0.3431,
      "step": 2289
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.472693032015066e-05,
      "loss": 0.3833,
      "step": 2290
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4724576271186444e-05,
      "loss": 0.4254,
      "step": 2291
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.472222222222223e-05,
      "loss": 0.2665,
      "step": 2292
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4719868173258005e-05,
      "loss": 0.4305,
      "step": 2293
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.471751412429378e-05,
      "loss": 0.3942,
      "step": 2294
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.471516007532957e-05,
      "loss": 0.539,
      "step": 2295
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.471280602636535e-05,
      "loss": 0.3592,
      "step": 2296
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.471045197740113e-05,
      "loss": 0.4309,
      "step": 2297
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.470809792843692e-05,
      "loss": 0.4302,
      "step": 2298
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4705743879472695e-05,
      "loss": 0.3685,
      "step": 2299
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.470338983050847e-05,
      "loss": 0.2704,
      "step": 2300
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4701035781544256e-05,
      "loss": 0.4266,
      "step": 2301
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.469868173258004e-05,
      "loss": 0.4691,
      "step": 2302
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4696327683615824e-05,
      "loss": 0.3019,
      "step": 2303
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.46939736346516e-05,
      "loss": 0.39,
      "step": 2304
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4691619585687385e-05,
      "loss": 0.4118,
      "step": 2305
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.468926553672317e-05,
      "loss": 0.3876,
      "step": 2306
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4686911487758946e-05,
      "loss": 0.5173,
      "step": 2307
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.468455743879473e-05,
      "loss": 0.5315,
      "step": 2308
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.4682203389830514e-05,
      "loss": 0.4454,
      "step": 2309
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.467984934086629e-05,
      "loss": 0.418,
      "step": 2310
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.467749529190207e-05,
      "loss": 0.3873,
      "step": 2311
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.467514124293786e-05,
      "loss": 0.4443,
      "step": 2312
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4672787193973636e-05,
      "loss": 0.3838,
      "step": 2313
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4670433145009413e-05,
      "loss": 0.3304,
      "step": 2314
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4668079096045204e-05,
      "loss": 0.3948,
      "step": 2315
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.466572504708098e-05,
      "loss": 0.4018,
      "step": 2316
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.466337099811676e-05,
      "loss": 0.4434,
      "step": 2317
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.466101694915254e-05,
      "loss": 0.3805,
      "step": 2318
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4658662900188326e-05,
      "loss": 0.4395,
      "step": 2319
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.465630885122411e-05,
      "loss": 0.3402,
      "step": 2320
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.465395480225989e-05,
      "loss": 0.445,
      "step": 2321
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.465160075329567e-05,
      "loss": 0.4563,
      "step": 2322
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4649246704331455e-05,
      "loss": 0.3507,
      "step": 2323
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.464689265536723e-05,
      "loss": 0.3348,
      "step": 2324
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4644538606403016e-05,
      "loss": 0.5451,
      "step": 2325
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.46421845574388e-05,
      "loss": 0.3423,
      "step": 2326
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.463983050847458e-05,
      "loss": 0.4681,
      "step": 2327
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4637476459510355e-05,
      "loss": 0.3566,
      "step": 2328
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.4635122410546145e-05,
      "loss": 0.3408,
      "step": 2329
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.463276836158192e-05,
      "loss": 0.4714,
      "step": 2330
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.46304143126177e-05,
      "loss": 0.3679,
      "step": 2331
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.462806026365349e-05,
      "loss": 0.302,
      "step": 2332
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.462570621468927e-05,
      "loss": 0.4951,
      "step": 2333
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.462335216572505e-05,
      "loss": 0.3243,
      "step": 2334
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.462099811676083e-05,
      "loss": 0.4063,
      "step": 2335
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.461864406779661e-05,
      "loss": 0.4899,
      "step": 2336
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4616290018832396e-05,
      "loss": 0.5007,
      "step": 2337
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4613935969868173e-05,
      "loss": 0.3407,
      "step": 2338
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.461158192090396e-05,
      "loss": 0.4984,
      "step": 2339
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.460922787193974e-05,
      "loss": 0.4042,
      "step": 2340
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.460687382297552e-05,
      "loss": 0.3517,
      "step": 2341
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.46045197740113e-05,
      "loss": 0.2831,
      "step": 2342
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4602165725047086e-05,
      "loss": 0.4884,
      "step": 2343
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4599811676082863e-05,
      "loss": 0.2459,
      "step": 2344
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.459745762711864e-05,
      "loss": 0.401,
      "step": 2345
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.459510357815443e-05,
      "loss": 0.4205,
      "step": 2346
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.459274952919021e-05,
      "loss": 0.3397,
      "step": 2347
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4590395480225986e-05,
      "loss": 0.4352,
      "step": 2348
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4588041431261776e-05,
      "loss": 0.4169,
      "step": 2349
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4585687382297553e-05,
      "loss": 0.4515,
      "step": 2350
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.458333333333334e-05,
      "loss": 0.458,
      "step": 2351
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.4580979284369114e-05,
      "loss": 0.4275,
      "step": 2352
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.45786252354049e-05,
      "loss": 0.3808,
      "step": 2353
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.457627118644068e-05,
      "loss": 0.2753,
      "step": 2354
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.457391713747646e-05,
      "loss": 0.413,
      "step": 2355
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4571563088512243e-05,
      "loss": 0.4837,
      "step": 2356
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.456920903954803e-05,
      "loss": 0.4081,
      "step": 2357
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4566854990583804e-05,
      "loss": 0.3041,
      "step": 2358
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.456450094161959e-05,
      "loss": 0.481,
      "step": 2359
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.456214689265537e-05,
      "loss": 0.2671,
      "step": 2360
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.455979284369115e-05,
      "loss": 0.3126,
      "step": 2361
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4557438794726933e-05,
      "loss": 0.4714,
      "step": 2362
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.455508474576272e-05,
      "loss": 0.2895,
      "step": 2363
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4552730696798494e-05,
      "loss": 0.3553,
      "step": 2364
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.455037664783427e-05,
      "loss": 0.2952,
      "step": 2365
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.454802259887006e-05,
      "loss": 0.4427,
      "step": 2366
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.454566854990584e-05,
      "loss": 0.4471,
      "step": 2367
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4543314500941623e-05,
      "loss": 0.4843,
      "step": 2368
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.45409604519774e-05,
      "loss": 0.3733,
      "step": 2369
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4538606403013184e-05,
      "loss": 0.4941,
      "step": 2370
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.453625235404897e-05,
      "loss": 0.4262,
      "step": 2371
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.4533898305084746e-05,
      "loss": 0.5048,
      "step": 2372
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.453154425612053e-05,
      "loss": 0.5404,
      "step": 2373
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.452919020715631e-05,
      "loss": 0.3465,
      "step": 2374
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.452683615819209e-05,
      "loss": 0.2917,
      "step": 2375
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4524482109227874e-05,
      "loss": 0.4748,
      "step": 2376
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.452212806026366e-05,
      "loss": 0.5302,
      "step": 2377
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4519774011299436e-05,
      "loss": 0.4079,
      "step": 2378
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.451741996233522e-05,
      "loss": 0.2451,
      "step": 2379
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4515065913371e-05,
      "loss": 0.405,
      "step": 2380
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.451271186440678e-05,
      "loss": 0.2847,
      "step": 2381
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4510357815442564e-05,
      "loss": 0.5197,
      "step": 2382
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.450800376647835e-05,
      "loss": 0.4251,
      "step": 2383
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4505649717514126e-05,
      "loss": 0.4104,
      "step": 2384
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.450329566854991e-05,
      "loss": 0.5035,
      "step": 2385
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4500941619585687e-05,
      "loss": 0.455,
      "step": 2386
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.449858757062147e-05,
      "loss": 0.467,
      "step": 2387
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4496233521657254e-05,
      "loss": 0.5135,
      "step": 2388
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.449387947269303e-05,
      "loss": 0.4125,
      "step": 2389
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4491525423728816e-05,
      "loss": 0.7081,
      "step": 2390
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.44891713747646e-05,
      "loss": 0.3717,
      "step": 2391
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4486817325800377e-05,
      "loss": 0.4643,
      "step": 2392
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.448446327683616e-05,
      "loss": 0.3344,
      "step": 2393
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.4482109227871944e-05,
      "loss": 0.4275,
      "step": 2394
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.447975517890772e-05,
      "loss": 0.3479,
      "step": 2395
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4477401129943506e-05,
      "loss": 0.329,
      "step": 2396
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.447504708097929e-05,
      "loss": 0.3437,
      "step": 2397
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4472693032015067e-05,
      "loss": 0.5861,
      "step": 2398
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.447033898305085e-05,
      "loss": 0.4432,
      "step": 2399
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4467984934086634e-05,
      "loss": 0.4509,
      "step": 2400
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.446563088512241e-05,
      "loss": 0.3798,
      "step": 2401
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4463276836158196e-05,
      "loss": 0.3453,
      "step": 2402
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.446092278719397e-05,
      "loss": 0.5449,
      "step": 2403
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4458568738229757e-05,
      "loss": 0.4543,
      "step": 2404
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.445621468926554e-05,
      "loss": 0.4095,
      "step": 2405
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.445386064030132e-05,
      "loss": 0.535,
      "step": 2406
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.44515065913371e-05,
      "loss": 0.3873,
      "step": 2407
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4449152542372886e-05,
      "loss": 0.246,
      "step": 2408
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.444679849340866e-05,
      "loss": 0.5748,
      "step": 2409
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4444444444444447e-05,
      "loss": 0.3782,
      "step": 2410
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.444209039548023e-05,
      "loss": 0.2427,
      "step": 2411
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.443973634651601e-05,
      "loss": 0.4131,
      "step": 2412
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.443738229755179e-05,
      "loss": 0.4256,
      "step": 2413
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4435028248587575e-05,
      "loss": 0.4035,
      "step": 2414
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.443267419962335e-05,
      "loss": 0.3472,
      "step": 2415
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.4430320150659137e-05,
      "loss": 0.3293,
      "step": 2416
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.442796610169492e-05,
      "loss": 0.4341,
      "step": 2417
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.44256120527307e-05,
      "loss": 0.3846,
      "step": 2418
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.442325800376648e-05,
      "loss": 0.2845,
      "step": 2419
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.442090395480226e-05,
      "loss": 0.423,
      "step": 2420
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.441854990583804e-05,
      "loss": 0.311,
      "step": 2421
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.4416195856873827e-05,
      "loss": 0.4163,
      "step": 2422
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.4413841807909604e-05,
      "loss": 0.4626,
      "step": 2423
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.441148775894539e-05,
      "loss": 0.3255,
      "step": 2424
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.440913370998117e-05,
      "loss": 0.2891,
      "step": 2425
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.440677966101695e-05,
      "loss": 0.3549,
      "step": 2426
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.440442561205273e-05,
      "loss": 0.284,
      "step": 2427
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.4402071563088517e-05,
      "loss": 0.4299,
      "step": 2428
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.4399717514124294e-05,
      "loss": 0.3897,
      "step": 2429
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.439736346516008e-05,
      "loss": 0.4119,
      "step": 2430
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.439500941619586e-05,
      "loss": 0.4698,
      "step": 2431
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.439265536723164e-05,
      "loss": 0.3204,
      "step": 2432
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.439030131826742e-05,
      "loss": 0.4324,
      "step": 2433
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.4387947269303207e-05,
      "loss": 0.5178,
      "step": 2434
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.4385593220338984e-05,
      "loss": 0.4161,
      "step": 2435
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.438323917137477e-05,
      "loss": 0.433,
      "step": 2436
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.438088512241055e-05,
      "loss": 0.2371,
      "step": 2437
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.437853107344633e-05,
      "loss": 0.4247,
      "step": 2438
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.437617702448211e-05,
      "loss": 0.534,
      "step": 2439
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.437382297551789e-05,
      "loss": 0.4061,
      "step": 2440
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4371468926553674e-05,
      "loss": 0.3188,
      "step": 2441
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.436911487758946e-05,
      "loss": 0.4397,
      "step": 2442
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4366760828625235e-05,
      "loss": 0.5101,
      "step": 2443
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.436440677966102e-05,
      "loss": 0.3132,
      "step": 2444
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.43620527306968e-05,
      "loss": 0.2986,
      "step": 2445
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.435969868173258e-05,
      "loss": 0.421,
      "step": 2446
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4357344632768364e-05,
      "loss": 0.4035,
      "step": 2447
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.435499058380415e-05,
      "loss": 0.4246,
      "step": 2448
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4352636534839925e-05,
      "loss": 0.301,
      "step": 2449
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.435028248587571e-05,
      "loss": 0.4043,
      "step": 2450
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.434792843691149e-05,
      "loss": 0.4709,
      "step": 2451
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.434557438794727e-05,
      "loss": 0.3639,
      "step": 2452
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4343220338983054e-05,
      "loss": 0.4448,
      "step": 2453
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.434086629001884e-05,
      "loss": 0.3158,
      "step": 2454
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4338512241054615e-05,
      "loss": 0.5286,
      "step": 2455
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.43361581920904e-05,
      "loss": 0.406,
      "step": 2456
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.4333804143126176e-05,
      "loss": 0.5293,
      "step": 2457
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.433145009416196e-05,
      "loss": 0.4521,
      "step": 2458
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4329096045197744e-05,
      "loss": 0.4247,
      "step": 2459
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.432674199623352e-05,
      "loss": 0.3678,
      "step": 2460
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4324387947269305e-05,
      "loss": 0.4607,
      "step": 2461
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.432203389830509e-05,
      "loss": 0.3729,
      "step": 2462
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4319679849340866e-05,
      "loss": 0.368,
      "step": 2463
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.431732580037665e-05,
      "loss": 0.3747,
      "step": 2464
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4314971751412434e-05,
      "loss": 0.4019,
      "step": 2465
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.431261770244821e-05,
      "loss": 0.3681,
      "step": 2466
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4310263653483995e-05,
      "loss": 0.4473,
      "step": 2467
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.430790960451978e-05,
      "loss": 0.3396,
      "step": 2468
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4305555555555556e-05,
      "loss": 0.348,
      "step": 2469
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.430320150659134e-05,
      "loss": 0.4891,
      "step": 2470
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4300847457627124e-05,
      "loss": 0.3164,
      "step": 2471
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.42984934086629e-05,
      "loss": 0.3628,
      "step": 2472
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4296139359698685e-05,
      "loss": 0.324,
      "step": 2473
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.429378531073446e-05,
      "loss": 0.3313,
      "step": 2474
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4291431261770246e-05,
      "loss": 0.3113,
      "step": 2475
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.428907721280603e-05,
      "loss": 0.3337,
      "step": 2476
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.428672316384181e-05,
      "loss": 0.4394,
      "step": 2477
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.428436911487759e-05,
      "loss": 0.516,
      "step": 2478
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.4282015065913375e-05,
      "loss": 0.5217,
      "step": 2479
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.427966101694915e-05,
      "loss": 0.3838,
      "step": 2480
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.4277306967984936e-05,
      "loss": 0.3996,
      "step": 2481
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.427495291902072e-05,
      "loss": 0.2792,
      "step": 2482
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.42725988700565e-05,
      "loss": 0.4346,
      "step": 2483
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.427024482109228e-05,
      "loss": 0.5248,
      "step": 2484
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.4267890772128065e-05,
      "loss": 0.3685,
      "step": 2485
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.426553672316384e-05,
      "loss": 0.4237,
      "step": 2486
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.4263182674199626e-05,
      "loss": 0.3983,
      "step": 2487
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.426082862523541e-05,
      "loss": 0.4248,
      "step": 2488
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.425847457627119e-05,
      "loss": 0.3387,
      "step": 2489
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.425612052730697e-05,
      "loss": 0.4235,
      "step": 2490
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.425376647834275e-05,
      "loss": 0.3432,
      "step": 2491
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.425141242937854e-05,
      "loss": 0.3675,
      "step": 2492
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.4249058380414316e-05,
      "loss": 0.2681,
      "step": 2493
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.424670433145009e-05,
      "loss": 0.4392,
      "step": 2494
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.424435028248588e-05,
      "loss": 0.5015,
      "step": 2495
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.424199623352166e-05,
      "loss": 0.3107,
      "step": 2496
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.423964218455744e-05,
      "loss": 0.299,
      "step": 2497
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.423728813559322e-05,
      "loss": 0.4478,
      "step": 2498
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.4234934086629006e-05,
      "loss": 0.3652,
      "step": 2499
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.423258003766478e-05,
      "loss": 0.4581,
      "step": 2500
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.423022598870057e-05,
      "loss": 0.4598,
      "step": 2501
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.422787193973635e-05,
      "loss": 0.4576,
      "step": 2502
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.422551789077213e-05,
      "loss": 0.5318,
      "step": 2503
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.422316384180791e-05,
      "loss": 0.3469,
      "step": 2504
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.4220809792843696e-05,
      "loss": 0.2829,
      "step": 2505
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.421845574387947e-05,
      "loss": 0.4041,
      "step": 2506
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.421610169491526e-05,
      "loss": 0.3993,
      "step": 2507
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.4213747645951034e-05,
      "loss": 0.4652,
      "step": 2508
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.4211393596986825e-05,
      "loss": 0.4373,
      "step": 2509
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.42090395480226e-05,
      "loss": 0.3948,
      "step": 2510
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.420668549905838e-05,
      "loss": 0.3879,
      "step": 2511
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.420433145009417e-05,
      "loss": 0.412,
      "step": 2512
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.420197740112995e-05,
      "loss": 0.2715,
      "step": 2513
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.4199623352165724e-05,
      "loss": 0.3326,
      "step": 2514
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.419726930320151e-05,
      "loss": 0.3444,
      "step": 2515
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.419491525423729e-05,
      "loss": 0.4734,
      "step": 2516
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.419256120527307e-05,
      "loss": 0.3192,
      "step": 2517
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.419020715630885e-05,
      "loss": 0.2933,
      "step": 2518
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.418785310734464e-05,
      "loss": 0.3131,
      "step": 2519
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.4185499058380414e-05,
      "loss": 0.4216,
      "step": 2520
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.41831450094162e-05,
      "loss": 0.4102,
      "step": 2521
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.418079096045198e-05,
      "loss": 0.4322,
      "step": 2522
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.417843691148776e-05,
      "loss": 0.4647,
      "step": 2523
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.417608286252354e-05,
      "loss": 0.4186,
      "step": 2524
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.417372881355932e-05,
      "loss": 0.5414,
      "step": 2525
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.417137476459511e-05,
      "loss": 0.3412,
      "step": 2526
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.416902071563089e-05,
      "loss": 0.6269,
      "step": 2527
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4166666666666665e-05,
      "loss": 0.3888,
      "step": 2528
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4164312617702456e-05,
      "loss": 0.3221,
      "step": 2529
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.416195856873823e-05,
      "loss": 0.337,
      "step": 2530
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.415960451977401e-05,
      "loss": 0.4342,
      "step": 2531
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4157250470809794e-05,
      "loss": 0.3287,
      "step": 2532
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.415489642184558e-05,
      "loss": 0.268,
      "step": 2533
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4152542372881355e-05,
      "loss": 0.4698,
      "step": 2534
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.415018832391714e-05,
      "loss": 0.3656,
      "step": 2535
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.414783427495292e-05,
      "loss": 0.4759,
      "step": 2536
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.41454802259887e-05,
      "loss": 0.4369,
      "step": 2537
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4143126177024484e-05,
      "loss": 0.3339,
      "step": 2538
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.414077212806027e-05,
      "loss": 0.3001,
      "step": 2539
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.413841807909605e-05,
      "loss": 0.4628,
      "step": 2540
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.413606403013183e-05,
      "loss": 0.5554,
      "step": 2541
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4133709981167606e-05,
      "loss": 0.3327,
      "step": 2542
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.41313559322034e-05,
      "loss": 0.4117,
      "step": 2543
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.4129001883239174e-05,
      "loss": 0.3088,
      "step": 2544
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.412664783427495e-05,
      "loss": 0.4405,
      "step": 2545
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.412429378531074e-05,
      "loss": 0.429,
      "step": 2546
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.412193973634652e-05,
      "loss": 0.3285,
      "step": 2547
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4119585687382296e-05,
      "loss": 0.4089,
      "step": 2548
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.411723163841808e-05,
      "loss": 0.3651,
      "step": 2549
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4114877589453864e-05,
      "loss": 0.4012,
      "step": 2550
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.411252354048964e-05,
      "loss": 0.4519,
      "step": 2551
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4110169491525425e-05,
      "loss": 0.3381,
      "step": 2552
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.410781544256121e-05,
      "loss": 0.3207,
      "step": 2553
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4105461393596986e-05,
      "loss": 0.4094,
      "step": 2554
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.410310734463277e-05,
      "loss": 0.2846,
      "step": 2555
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4100753295668554e-05,
      "loss": 0.2985,
      "step": 2556
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.409839924670434e-05,
      "loss": 0.2752,
      "step": 2557
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4096045197740115e-05,
      "loss": 0.3632,
      "step": 2558
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.409369114877589e-05,
      "loss": 0.38,
      "step": 2559
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.409133709981168e-05,
      "loss": 0.3435,
      "step": 2560
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.408898305084746e-05,
      "loss": 0.3932,
      "step": 2561
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.408662900188324e-05,
      "loss": 0.5075,
      "step": 2562
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.408427495291903e-05,
      "loss": 0.3473,
      "step": 2563
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.4081920903954805e-05,
      "loss": 0.4203,
      "step": 2564
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.407956685499058e-05,
      "loss": 0.3346,
      "step": 2565
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4077212806026366e-05,
      "loss": 0.3819,
      "step": 2566
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.407485875706215e-05,
      "loss": 0.4239,
      "step": 2567
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.407250470809793e-05,
      "loss": 0.3473,
      "step": 2568
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.407015065913371e-05,
      "loss": 0.3831,
      "step": 2569
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4067796610169495e-05,
      "loss": 0.4311,
      "step": 2570
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.406544256120528e-05,
      "loss": 0.4454,
      "step": 2571
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4063088512241056e-05,
      "loss": 0.3769,
      "step": 2572
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.406073446327684e-05,
      "loss": 0.4352,
      "step": 2573
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4058380414312624e-05,
      "loss": 0.307,
      "step": 2574
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.40560263653484e-05,
      "loss": 0.3328,
      "step": 2575
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.405367231638418e-05,
      "loss": 0.384,
      "step": 2576
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.405131826741997e-05,
      "loss": 0.3771,
      "step": 2577
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4048964218455746e-05,
      "loss": 0.3955,
      "step": 2578
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.404661016949152e-05,
      "loss": 0.4304,
      "step": 2579
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4044256120527314e-05,
      "loss": 0.3568,
      "step": 2580
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.404190207156309e-05,
      "loss": 0.3081,
      "step": 2581
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.403954802259887e-05,
      "loss": 0.3816,
      "step": 2582
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.403719397363465e-05,
      "loss": 0.4728,
      "step": 2583
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4034839924670436e-05,
      "loss": 0.3971,
      "step": 2584
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.403248587570621e-05,
      "loss": 0.3173,
      "step": 2585
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.4030131826742e-05,
      "loss": 0.3217,
      "step": 2586
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.402777777777778e-05,
      "loss": 0.5156,
      "step": 2587
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.4025423728813565e-05,
      "loss": 0.4125,
      "step": 2588
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.402306967984934e-05,
      "loss": 0.3521,
      "step": 2589
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.4020715630885126e-05,
      "loss": 0.3301,
      "step": 2590
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.401836158192091e-05,
      "loss": 0.3294,
      "step": 2591
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.401600753295669e-05,
      "loss": 0.4171,
      "step": 2592
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.4013653483992464e-05,
      "loss": 0.4895,
      "step": 2593
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.4011299435028255e-05,
      "loss": 0.4266,
      "step": 2594
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.400894538606403e-05,
      "loss": 0.4507,
      "step": 2595
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.400659133709981e-05,
      "loss": 0.4167,
      "step": 2596
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.40042372881356e-05,
      "loss": 0.4652,
      "step": 2597
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.400188323917138e-05,
      "loss": 0.5024,
      "step": 2598
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.3999529190207154e-05,
      "loss": 0.4148,
      "step": 2599
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.399717514124294e-05,
      "loss": 0.2943,
      "step": 2600
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.399482109227872e-05,
      "loss": 0.3585,
      "step": 2601
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.39924670433145e-05,
      "loss": 0.457,
      "step": 2602
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.399011299435028e-05,
      "loss": 0.4661,
      "step": 2603
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.398775894538607e-05,
      "loss": 0.3609,
      "step": 2604
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.398540489642185e-05,
      "loss": 0.219,
      "step": 2605
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.398305084745763e-05,
      "loss": 0.3835,
      "step": 2606
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.398069679849341e-05,
      "loss": 0.3086,
      "step": 2607
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.3978342749529196e-05,
      "loss": 0.3565,
      "step": 2608
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.397598870056497e-05,
      "loss": 0.4523,
      "step": 2609
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.397363465160075e-05,
      "loss": 0.3384,
      "step": 2610
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.397128060263654e-05,
      "loss": 0.501,
      "step": 2611
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.396892655367232e-05,
      "loss": 0.3844,
      "step": 2612
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.3966572504708095e-05,
      "loss": 0.389,
      "step": 2613
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.3964218455743886e-05,
      "loss": 0.3833,
      "step": 2614
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.396186440677966e-05,
      "loss": 0.4833,
      "step": 2615
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.395951035781544e-05,
      "loss": 0.4174,
      "step": 2616
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.3957156308851224e-05,
      "loss": 0.4949,
      "step": 2617
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.395480225988701e-05,
      "loss": 0.2707,
      "step": 2618
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.395244821092279e-05,
      "loss": 0.3634,
      "step": 2619
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.395009416195857e-05,
      "loss": 0.4254,
      "step": 2620
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.394774011299435e-05,
      "loss": 0.4675,
      "step": 2621
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.394538606403014e-05,
      "loss": 0.2153,
      "step": 2622
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.3943032015065914e-05,
      "loss": 0.3898,
      "step": 2623
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.39406779661017e-05,
      "loss": 0.457,
      "step": 2624
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.393832391713748e-05,
      "loss": 0.2907,
      "step": 2625
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.393596986817326e-05,
      "loss": 0.3925,
      "step": 2626
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.3933615819209036e-05,
      "loss": 0.3768,
      "step": 2627
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.393126177024483e-05,
      "loss": 0.4449,
      "step": 2628
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.3928907721280604e-05,
      "loss": 0.4715,
      "step": 2629
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.392655367231638e-05,
      "loss": 0.4454,
      "step": 2630
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.392419962335217e-05,
      "loss": 0.3686,
      "step": 2631
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.392184557438795e-05,
      "loss": 0.4875,
      "step": 2632
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.3919491525423726e-05,
      "loss": 0.4128,
      "step": 2633
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.391713747645951e-05,
      "loss": 0.4199,
      "step": 2634
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.3914783427495294e-05,
      "loss": 0.3641,
      "step": 2635
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.391242937853108e-05,
      "loss": 0.3634,
      "step": 2636
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.3910075329566855e-05,
      "loss": 0.4509,
      "step": 2637
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.390772128060264e-05,
      "loss": 0.4421,
      "step": 2638
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.390536723163842e-05,
      "loss": 0.3638,
      "step": 2639
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.39030131826742e-05,
      "loss": 0.4259,
      "step": 2640
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.3900659133709984e-05,
      "loss": 0.4361,
      "step": 2641
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.389830508474577e-05,
      "loss": 0.4168,
      "step": 2642
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.3895951035781545e-05,
      "loss": 0.3654,
      "step": 2643
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.389359698681732e-05,
      "loss": 0.4685,
      "step": 2644
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.389124293785311e-05,
      "loss": 0.3676,
      "step": 2645
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.388888888888889e-05,
      "loss": 0.3963,
      "step": 2646
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.388653483992467e-05,
      "loss": 0.3835,
      "step": 2647
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.388418079096046e-05,
      "loss": 0.3287,
      "step": 2648
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.3881826741996235e-05,
      "loss": 0.2891,
      "step": 2649
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.387947269303202e-05,
      "loss": 0.2727,
      "step": 2650
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3877118644067796e-05,
      "loss": 0.3165,
      "step": 2651
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.387476459510358e-05,
      "loss": 0.2671,
      "step": 2652
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3872410546139364e-05,
      "loss": 0.3213,
      "step": 2653
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.387005649717514e-05,
      "loss": 0.3573,
      "step": 2654
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3867702448210925e-05,
      "loss": 0.3546,
      "step": 2655
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.386534839924671e-05,
      "loss": 0.4806,
      "step": 2656
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3862994350282486e-05,
      "loss": 0.3997,
      "step": 2657
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.386064030131827e-05,
      "loss": 0.235,
      "step": 2658
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3858286252354054e-05,
      "loss": 0.4329,
      "step": 2659
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.385593220338983e-05,
      "loss": 0.4057,
      "step": 2660
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.385357815442561e-05,
      "loss": 0.2987,
      "step": 2661
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.38512241054614e-05,
      "loss": 0.3087,
      "step": 2662
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3848870056497176e-05,
      "loss": 0.3073,
      "step": 2663
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3846516007532954e-05,
      "loss": 0.4368,
      "step": 2664
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3844161958568744e-05,
      "loss": 0.4395,
      "step": 2665
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.384180790960452e-05,
      "loss": 0.288,
      "step": 2666
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3839453860640305e-05,
      "loss": 0.3921,
      "step": 2667
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.383709981167608e-05,
      "loss": 0.4078,
      "step": 2668
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.3834745762711866e-05,
      "loss": 0.2928,
      "step": 2669
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.383239171374765e-05,
      "loss": 0.3847,
      "step": 2670
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.383003766478343e-05,
      "loss": 0.2643,
      "step": 2671
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.382768361581921e-05,
      "loss": 0.5336,
      "step": 2672
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.3825329566854995e-05,
      "loss": 0.3854,
      "step": 2673
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.382297551789077e-05,
      "loss": 0.4182,
      "step": 2674
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.3820621468926556e-05,
      "loss": 0.4575,
      "step": 2675
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.381826741996234e-05,
      "loss": 0.2906,
      "step": 2676
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.381591337099812e-05,
      "loss": 0.1891,
      "step": 2677
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.38135593220339e-05,
      "loss": 0.2454,
      "step": 2678
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.3811205273069685e-05,
      "loss": 0.3394,
      "step": 2679
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.380885122410546e-05,
      "loss": 0.4188,
      "step": 2680
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.380649717514124e-05,
      "loss": 0.4615,
      "step": 2681
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.380414312617703e-05,
      "loss": 0.3675,
      "step": 2682
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.380178907721281e-05,
      "loss": 0.3691,
      "step": 2683
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.379943502824859e-05,
      "loss": 0.3709,
      "step": 2684
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.379708097928437e-05,
      "loss": 0.6007,
      "step": 2685
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.379472693032015e-05,
      "loss": 0.4134,
      "step": 2686
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.3792372881355936e-05,
      "loss": 0.361,
      "step": 2687
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.3790018832391713e-05,
      "loss": 0.4143,
      "step": 2688
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.37876647834275e-05,
      "loss": 0.2538,
      "step": 2689
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.378531073446328e-05,
      "loss": 0.3965,
      "step": 2690
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.378295668549906e-05,
      "loss": 0.4714,
      "step": 2691
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.378060263653484e-05,
      "loss": 0.3849,
      "step": 2692
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.3778248587570626e-05,
      "loss": 0.4317,
      "step": 2693
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3775894538606403e-05,
      "loss": 0.232,
      "step": 2694
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.377354048964219e-05,
      "loss": 0.3833,
      "step": 2695
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.377118644067797e-05,
      "loss": 0.3842,
      "step": 2696
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.376883239171375e-05,
      "loss": 0.3634,
      "step": 2697
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.376647834274953e-05,
      "loss": 0.362,
      "step": 2698
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3764124293785316e-05,
      "loss": 0.326,
      "step": 2699
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3761770244821093e-05,
      "loss": 0.4608,
      "step": 2700
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.375941619585688e-05,
      "loss": 0.4009,
      "step": 2701
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3757062146892655e-05,
      "loss": 0.448,
      "step": 2702
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.375470809792844e-05,
      "loss": 0.3205,
      "step": 2703
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.375235404896422e-05,
      "loss": 0.437,
      "step": 2704
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.375e-05,
      "loss": 0.4203,
      "step": 2705
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3747645951035783e-05,
      "loss": 0.4791,
      "step": 2706
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.374529190207157e-05,
      "loss": 0.4506,
      "step": 2707
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3742937853107345e-05,
      "loss": 0.269,
      "step": 2708
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.374058380414313e-05,
      "loss": 0.3523,
      "step": 2709
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.373822975517891e-05,
      "loss": 0.3804,
      "step": 2710
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.373587570621469e-05,
      "loss": 0.4727,
      "step": 2711
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3733521657250473e-05,
      "loss": 0.3734,
      "step": 2712
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.373116760828626e-05,
      "loss": 0.6498,
      "step": 2713
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.3728813559322035e-05,
      "loss": 0.4472,
      "step": 2714
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.372645951035782e-05,
      "loss": 0.6315,
      "step": 2715
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.37241054613936e-05,
      "loss": 0.3465,
      "step": 2716
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.372175141242938e-05,
      "loss": 0.5125,
      "step": 2717
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3719397363465163e-05,
      "loss": 0.4905,
      "step": 2718
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.371704331450094e-05,
      "loss": 0.3502,
      "step": 2719
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3714689265536725e-05,
      "loss": 0.5483,
      "step": 2720
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.371233521657251e-05,
      "loss": 0.3425,
      "step": 2721
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3709981167608286e-05,
      "loss": 0.3669,
      "step": 2722
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.370762711864407e-05,
      "loss": 0.3955,
      "step": 2723
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3705273069679853e-05,
      "loss": 0.3503,
      "step": 2724
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.370291902071563e-05,
      "loss": 0.3342,
      "step": 2725
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3700564971751415e-05,
      "loss": 0.3505,
      "step": 2726
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.36982109227872e-05,
      "loss": 0.3601,
      "step": 2727
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3695856873822976e-05,
      "loss": 0.347,
      "step": 2728
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.369350282485876e-05,
      "loss": 0.3941,
      "step": 2729
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3691148775894543e-05,
      "loss": 0.423,
      "step": 2730
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.368879472693032e-05,
      "loss": 0.3504,
      "step": 2731
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3686440677966105e-05,
      "loss": 0.2657,
      "step": 2732
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.368408662900189e-05,
      "loss": 0.3023,
      "step": 2733
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.3681732580037666e-05,
      "loss": 0.4181,
      "step": 2734
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.367937853107345e-05,
      "loss": 0.4667,
      "step": 2735
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.367702448210923e-05,
      "loss": 0.4623,
      "step": 2736
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.367467043314501e-05,
      "loss": 0.339,
      "step": 2737
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.3672316384180795e-05,
      "loss": 0.4737,
      "step": 2738
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.366996233521657e-05,
      "loss": 0.3666,
      "step": 2739
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.3667608286252356e-05,
      "loss": 0.4441,
      "step": 2740
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.366525423728814e-05,
      "loss": 0.2703,
      "step": 2741
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.366290018832392e-05,
      "loss": 0.4418,
      "step": 2742
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.36605461393597e-05,
      "loss": 0.37,
      "step": 2743
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.3658192090395485e-05,
      "loss": 0.3745,
      "step": 2744
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.365583804143126e-05,
      "loss": 0.4703,
      "step": 2745
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.3653483992467046e-05,
      "loss": 0.3782,
      "step": 2746
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.365112994350283e-05,
      "loss": 0.2435,
      "step": 2747
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.364877589453861e-05,
      "loss": 0.3581,
      "step": 2748
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.364642184557439e-05,
      "loss": 0.4568,
      "step": 2749
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.3644067796610175e-05,
      "loss": 0.4073,
      "step": 2750
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.364171374764595e-05,
      "loss": 0.3944,
      "step": 2751
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.3639359698681736e-05,
      "loss": 0.3408,
      "step": 2752
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.363700564971752e-05,
      "loss": 0.4593,
      "step": 2753
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.36346516007533e-05,
      "loss": 0.3584,
      "step": 2754
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.363229755178908e-05,
      "loss": 0.4258,
      "step": 2755
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.362994350282486e-05,
      "loss": 0.3994,
      "step": 2756
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.362758945386064e-05,
      "loss": 0.4523,
      "step": 2757
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.3625235404896426e-05,
      "loss": 0.4487,
      "step": 2758
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.36228813559322e-05,
      "loss": 0.4654,
      "step": 2759
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.362052730696799e-05,
      "loss": 0.2992,
      "step": 2760
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.361817325800377e-05,
      "loss": 0.2245,
      "step": 2761
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.361581920903955e-05,
      "loss": 0.3528,
      "step": 2762
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.361346516007533e-05,
      "loss": 0.3912,
      "step": 2763
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.3611111111111116e-05,
      "loss": 0.3978,
      "step": 2764
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.360875706214689e-05,
      "loss": 0.3248,
      "step": 2765
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.360640301318268e-05,
      "loss": 0.3101,
      "step": 2766
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.360404896421846e-05,
      "loss": 0.3338,
      "step": 2767
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.360169491525424e-05,
      "loss": 0.3384,
      "step": 2768
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.359934086629002e-05,
      "loss": 0.4611,
      "step": 2769
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.3596986817325806e-05,
      "loss": 0.3814,
      "step": 2770
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.359463276836158e-05,
      "loss": 0.4229,
      "step": 2771
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.359227871939737e-05,
      "loss": 0.3707,
      "step": 2772
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.3589924670433144e-05,
      "loss": 0.4887,
      "step": 2773
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.358757062146893e-05,
      "loss": 0.437,
      "step": 2774
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.358521657250471e-05,
      "loss": 0.4191,
      "step": 2775
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.358286252354049e-05,
      "loss": 0.4409,
      "step": 2776
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.358050847457627e-05,
      "loss": 0.5313,
      "step": 2777
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.357815442561206e-05,
      "loss": 0.3413,
      "step": 2778
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3575800376647834e-05,
      "loss": 0.364,
      "step": 2779
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.357344632768362e-05,
      "loss": 0.2957,
      "step": 2780
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.35710922787194e-05,
      "loss": 0.4519,
      "step": 2781
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.356873822975518e-05,
      "loss": 0.33,
      "step": 2782
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.356638418079096e-05,
      "loss": 0.4713,
      "step": 2783
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.356403013182675e-05,
      "loss": 0.3882,
      "step": 2784
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3561676082862524e-05,
      "loss": 0.3619,
      "step": 2785
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.355932203389831e-05,
      "loss": 0.33,
      "step": 2786
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.355696798493409e-05,
      "loss": 0.351,
      "step": 2787
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.355461393596987e-05,
      "loss": 0.4753,
      "step": 2788
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.355225988700565e-05,
      "loss": 0.3451,
      "step": 2789
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.354990583804143e-05,
      "loss": 0.2758,
      "step": 2790
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3547551789077214e-05,
      "loss": 0.4755,
      "step": 2791
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3545197740113e-05,
      "loss": 0.3446,
      "step": 2792
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3542843691148775e-05,
      "loss": 0.3439,
      "step": 2793
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.354048964218456e-05,
      "loss": 0.384,
      "step": 2794
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.353813559322034e-05,
      "loss": 0.4188,
      "step": 2795
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.353578154425612e-05,
      "loss": 0.3835,
      "step": 2796
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3533427495291904e-05,
      "loss": 0.3159,
      "step": 2797
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.353107344632769e-05,
      "loss": 0.4199,
      "step": 2798
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.3528719397363465e-05,
      "loss": 0.3017,
      "step": 2799
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.352636534839925e-05,
      "loss": 0.443,
      "step": 2800
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.352401129943503e-05,
      "loss": 0.4078,
      "step": 2801
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.352165725047081e-05,
      "loss": 0.4473,
      "step": 2802
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3519303201506594e-05,
      "loss": 0.3763,
      "step": 2803
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.351694915254238e-05,
      "loss": 0.3308,
      "step": 2804
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3514595103578155e-05,
      "loss": 0.4383,
      "step": 2805
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.351224105461394e-05,
      "loss": 0.3675,
      "step": 2806
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3509887005649716e-05,
      "loss": 0.4266,
      "step": 2807
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3507532956685507e-05,
      "loss": 0.3707,
      "step": 2808
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3505178907721284e-05,
      "loss": 0.4003,
      "step": 2809
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.350282485875706e-05,
      "loss": 0.5227,
      "step": 2810
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3500470809792845e-05,
      "loss": 0.3981,
      "step": 2811
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.349811676082863e-05,
      "loss": 0.3579,
      "step": 2812
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3495762711864406e-05,
      "loss": 0.339,
      "step": 2813
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.349340866290019e-05,
      "loss": 0.3777,
      "step": 2814
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3491054613935974e-05,
      "loss": 0.3531,
      "step": 2815
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.348870056497175e-05,
      "loss": 0.4414,
      "step": 2816
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3486346516007535e-05,
      "loss": 0.3231,
      "step": 2817
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.348399246704332e-05,
      "loss": 0.3201,
      "step": 2818
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.3481638418079096e-05,
      "loss": 0.2447,
      "step": 2819
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.347928436911488e-05,
      "loss": 0.3243,
      "step": 2820
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3476930320150664e-05,
      "loss": 0.3811,
      "step": 2821
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.347457627118644e-05,
      "loss": 0.3177,
      "step": 2822
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3472222222222225e-05,
      "loss": 0.4294,
      "step": 2823
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3469868173258e-05,
      "loss": 0.6724,
      "step": 2824
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.346751412429379e-05,
      "loss": 0.5909,
      "step": 2825
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.346516007532957e-05,
      "loss": 0.4208,
      "step": 2826
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.346280602636535e-05,
      "loss": 0.3509,
      "step": 2827
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.346045197740114e-05,
      "loss": 0.2323,
      "step": 2828
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3458097928436915e-05,
      "loss": 0.3195,
      "step": 2829
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.345574387947269e-05,
      "loss": 0.2976,
      "step": 2830
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3453389830508476e-05,
      "loss": 0.2969,
      "step": 2831
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.345103578154426e-05,
      "loss": 0.3182,
      "step": 2832
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.344868173258004e-05,
      "loss": 0.4547,
      "step": 2833
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.344632768361582e-05,
      "loss": 0.3712,
      "step": 2834
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3443973634651605e-05,
      "loss": 0.4508,
      "step": 2835
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.344161958568738e-05,
      "loss": 0.5325,
      "step": 2836
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3439265536723166e-05,
      "loss": 0.4184,
      "step": 2837
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.343691148775895e-05,
      "loss": 0.4019,
      "step": 2838
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.343455743879473e-05,
      "loss": 0.4675,
      "step": 2839
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.343220338983051e-05,
      "loss": 0.4616,
      "step": 2840
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.342984934086629e-05,
      "loss": 0.3186,
      "step": 2841
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.342749529190208e-05,
      "loss": 0.4555,
      "step": 2842
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.3425141242937856e-05,
      "loss": 0.4151,
      "step": 2843
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.342278719397363e-05,
      "loss": 0.3747,
      "step": 2844
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.3420433145009424e-05,
      "loss": 0.4874,
      "step": 2845
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.34180790960452e-05,
      "loss": 0.4061,
      "step": 2846
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.341572504708098e-05,
      "loss": 0.4052,
      "step": 2847
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.341337099811676e-05,
      "loss": 0.4152,
      "step": 2848
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.3411016949152546e-05,
      "loss": 0.3949,
      "step": 2849
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.340866290018832e-05,
      "loss": 0.3969,
      "step": 2850
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.340630885122411e-05,
      "loss": 0.4013,
      "step": 2851
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.340395480225989e-05,
      "loss": 0.4207,
      "step": 2852
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.340160075329567e-05,
      "loss": 0.4492,
      "step": 2853
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.339924670433145e-05,
      "loss": 0.356,
      "step": 2854
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.3396892655367236e-05,
      "loss": 0.3917,
      "step": 2855
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.339453860640302e-05,
      "loss": 0.2806,
      "step": 2856
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.33921845574388e-05,
      "loss": 0.4238,
      "step": 2857
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.3389830508474574e-05,
      "loss": 0.2793,
      "step": 2858
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.3387476459510365e-05,
      "loss": 0.3003,
      "step": 2859
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.338512241054614e-05,
      "loss": 0.3824,
      "step": 2860
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.338276836158192e-05,
      "loss": 0.416,
      "step": 2861
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.338041431261771e-05,
      "loss": 0.3982,
      "step": 2862
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.337806026365349e-05,
      "loss": 0.4288,
      "step": 2863
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3375706214689264e-05,
      "loss": 0.4661,
      "step": 2864
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.337335216572505e-05,
      "loss": 0.4377,
      "step": 2865
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.337099811676083e-05,
      "loss": 0.2414,
      "step": 2866
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.336864406779661e-05,
      "loss": 0.3788,
      "step": 2867
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.336629001883239e-05,
      "loss": 0.2985,
      "step": 2868
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.336393596986818e-05,
      "loss": 0.4266,
      "step": 2869
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3361581920903954e-05,
      "loss": 0.3323,
      "step": 2870
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.335922787193974e-05,
      "loss": 0.4119,
      "step": 2871
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.335687382297552e-05,
      "loss": 0.4305,
      "step": 2872
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3354519774011306e-05,
      "loss": 0.4047,
      "step": 2873
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.335216572504708e-05,
      "loss": 0.3039,
      "step": 2874
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.334981167608286e-05,
      "loss": 0.3664,
      "step": 2875
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.334745762711865e-05,
      "loss": 0.4417,
      "step": 2876
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.334510357815443e-05,
      "loss": 0.3368,
      "step": 2877
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3342749529190205e-05,
      "loss": 0.5073,
      "step": 2878
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3340395480225996e-05,
      "loss": 0.4984,
      "step": 2879
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.333804143126177e-05,
      "loss": 0.4926,
      "step": 2880
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.333568738229755e-05,
      "loss": 0.5094,
      "step": 2881
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3333333333333334e-05,
      "loss": 0.3616,
      "step": 2882
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.333097928436912e-05,
      "loss": 0.4341,
      "step": 2883
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.3328625235404895e-05,
      "loss": 0.3329,
      "step": 2884
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.332627118644068e-05,
      "loss": 0.3458,
      "step": 2885
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.332391713747646e-05,
      "loss": 0.3199,
      "step": 2886
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.332156308851225e-05,
      "loss": 0.4483,
      "step": 2887
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3319209039548024e-05,
      "loss": 0.5369,
      "step": 2888
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.331685499058381e-05,
      "loss": 0.3723,
      "step": 2889
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.331450094161959e-05,
      "loss": 0.4155,
      "step": 2890
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.331214689265537e-05,
      "loss": 0.4595,
      "step": 2891
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3309792843691146e-05,
      "loss": 0.3393,
      "step": 2892
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.330743879472694e-05,
      "loss": 0.3387,
      "step": 2893
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3305084745762714e-05,
      "loss": 0.3632,
      "step": 2894
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.330273069679849e-05,
      "loss": 0.3668,
      "step": 2895
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.330037664783428e-05,
      "loss": 0.3314,
      "step": 2896
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.329802259887006e-05,
      "loss": 0.4337,
      "step": 2897
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3295668549905836e-05,
      "loss": 0.3536,
      "step": 2898
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.329331450094162e-05,
      "loss": 0.3671,
      "step": 2899
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3290960451977404e-05,
      "loss": 0.4133,
      "step": 2900
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.328860640301318e-05,
      "loss": 0.3728,
      "step": 2901
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3286252354048965e-05,
      "loss": 0.3031,
      "step": 2902
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.328389830508475e-05,
      "loss": 0.3042,
      "step": 2903
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.328154425612053e-05,
      "loss": 0.3295,
      "step": 2904
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.327919020715631e-05,
      "loss": 0.328,
      "step": 2905
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.3276836158192094e-05,
      "loss": 0.3713,
      "step": 2906
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.327448210922788e-05,
      "loss": 0.2174,
      "step": 2907
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3272128060263655e-05,
      "loss": 0.3165,
      "step": 2908
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.326977401129943e-05,
      "loss": 0.4014,
      "step": 2909
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.326741996233522e-05,
      "loss": 0.54,
      "step": 2910
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3265065913371e-05,
      "loss": 0.3928,
      "step": 2911
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.326271186440678e-05,
      "loss": 0.3109,
      "step": 2912
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.326035781544257e-05,
      "loss": 0.3118,
      "step": 2913
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3258003766478345e-05,
      "loss": 0.2205,
      "step": 2914
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.325564971751412e-05,
      "loss": 0.4318,
      "step": 2915
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3253295668549906e-05,
      "loss": 0.2334,
      "step": 2916
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.325094161958569e-05,
      "loss": 0.5417,
      "step": 2917
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.324858757062147e-05,
      "loss": 0.3744,
      "step": 2918
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.324623352165725e-05,
      "loss": 0.5058,
      "step": 2919
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3243879472693035e-05,
      "loss": 0.4753,
      "step": 2920
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.324152542372882e-05,
      "loss": 0.5681,
      "step": 2921
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3239171374764596e-05,
      "loss": 0.4117,
      "step": 2922
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.323681732580038e-05,
      "loss": 0.3476,
      "step": 2923
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3234463276836164e-05,
      "loss": 0.3977,
      "step": 2924
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.323210922787194e-05,
      "loss": 0.2429,
      "step": 2925
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.322975517890772e-05,
      "loss": 0.3139,
      "step": 2926
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.322740112994351e-05,
      "loss": 0.4687,
      "step": 2927
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3225047080979286e-05,
      "loss": 0.2964,
      "step": 2928
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.322269303201506e-05,
      "loss": 0.399,
      "step": 2929
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3220338983050854e-05,
      "loss": 0.5041,
      "step": 2930
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.321798493408663e-05,
      "loss": 0.32,
      "step": 2931
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.321563088512241e-05,
      "loss": 0.3986,
      "step": 2932
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.321327683615819e-05,
      "loss": 0.3871,
      "step": 2933
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3210922787193976e-05,
      "loss": 0.3367,
      "step": 2934
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.320856873822976e-05,
      "loss": 0.4201,
      "step": 2935
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.320621468926554e-05,
      "loss": 0.3991,
      "step": 2936
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.320386064030132e-05,
      "loss": 0.4082,
      "step": 2937
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3201506591337105e-05,
      "loss": 0.4342,
      "step": 2938
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.319915254237288e-05,
      "loss": 0.4364,
      "step": 2939
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3196798493408666e-05,
      "loss": 0.3281,
      "step": 2940
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.319444444444445e-05,
      "loss": 0.361,
      "step": 2941
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.319209039548023e-05,
      "loss": 0.3721,
      "step": 2942
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3189736346516004e-05,
      "loss": 0.4278,
      "step": 2943
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.3187382297551795e-05,
      "loss": 0.3693,
      "step": 2944
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.318502824858757e-05,
      "loss": 0.3412,
      "step": 2945
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.318267419962335e-05,
      "loss": 0.354,
      "step": 2946
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.318032015065914e-05,
      "loss": 0.3952,
      "step": 2947
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.317796610169492e-05,
      "loss": 0.5121,
      "step": 2948
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.3175612052730694e-05,
      "loss": 0.3728,
      "step": 2949
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.317325800376648e-05,
      "loss": 0.393,
      "step": 2950
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.317090395480226e-05,
      "loss": 0.2488,
      "step": 2951
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.3168549905838046e-05,
      "loss": 0.3888,
      "step": 2952
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.316619585687382e-05,
      "loss": 0.3708,
      "step": 2953
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.316384180790961e-05,
      "loss": 0.3309,
      "step": 2954
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.316148775894539e-05,
      "loss": 0.4505,
      "step": 2955
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.315913370998117e-05,
      "loss": 0.3483,
      "step": 2956
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.315677966101695e-05,
      "loss": 0.2707,
      "step": 2957
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.3154425612052736e-05,
      "loss": 0.3345,
      "step": 2958
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.315207156308851e-05,
      "loss": 0.4333,
      "step": 2959
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.314971751412429e-05,
      "loss": 0.3956,
      "step": 2960
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.314736346516008e-05,
      "loss": 0.3941,
      "step": 2961
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.314500941619586e-05,
      "loss": 0.3502,
      "step": 2962
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.3142655367231635e-05,
      "loss": 0.3262,
      "step": 2963
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.3140301318267426e-05,
      "loss": 0.2927,
      "step": 2964
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.31379472693032e-05,
      "loss": 0.3721,
      "step": 2965
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.313559322033899e-05,
      "loss": 0.4059,
      "step": 2966
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.3133239171374764e-05,
      "loss": 0.4469,
      "step": 2967
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.313088512241055e-05,
      "loss": 0.322,
      "step": 2968
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.312853107344633e-05,
      "loss": 0.4238,
      "step": 2969
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.312617702448211e-05,
      "loss": 0.3898,
      "step": 2970
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.312382297551789e-05,
      "loss": 0.3889,
      "step": 2971
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.312146892655368e-05,
      "loss": 0.3694,
      "step": 2972
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3119114877589454e-05,
      "loss": 0.5344,
      "step": 2973
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.311676082862524e-05,
      "loss": 0.3574,
      "step": 2974
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.311440677966102e-05,
      "loss": 0.3842,
      "step": 2975
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.31120527306968e-05,
      "loss": 0.3554,
      "step": 2976
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3109698681732576e-05,
      "loss": 0.4321,
      "step": 2977
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.310734463276837e-05,
      "loss": 0.3328,
      "step": 2978
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3104990583804144e-05,
      "loss": 0.2957,
      "step": 2979
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.310263653483992e-05,
      "loss": 0.4516,
      "step": 2980
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.310028248587571e-05,
      "loss": 0.3525,
      "step": 2981
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.309792843691149e-05,
      "loss": 0.4227,
      "step": 2982
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.309557438794727e-05,
      "loss": 0.3991,
      "step": 2983
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.309322033898305e-05,
      "loss": 0.5274,
      "step": 2984
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3090866290018834e-05,
      "loss": 0.406,
      "step": 2985
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.308851224105462e-05,
      "loss": 0.3351,
      "step": 2986
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3086158192090395e-05,
      "loss": 0.4066,
      "step": 2987
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.308380414312618e-05,
      "loss": 0.3484,
      "step": 2988
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.308145009416196e-05,
      "loss": 0.3788,
      "step": 2989
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.307909604519774e-05,
      "loss": 0.4059,
      "step": 2990
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3076741996233524e-05,
      "loss": 0.3833,
      "step": 2991
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.307438794726931e-05,
      "loss": 0.4194,
      "step": 2992
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3072033898305085e-05,
      "loss": 0.4089,
      "step": 2993
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.306967984934087e-05,
      "loss": 0.3486,
      "step": 2994
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.306732580037665e-05,
      "loss": 0.5285,
      "step": 2995
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.306497175141243e-05,
      "loss": 0.465,
      "step": 2996
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.306261770244821e-05,
      "loss": 0.3208,
      "step": 2997
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3060263653484e-05,
      "loss": 0.4546,
      "step": 2998
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3057909604519775e-05,
      "loss": 0.3511,
      "step": 2999
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.305555555555556e-05,
      "loss": 0.3189,
      "step": 3000
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3053201506591336e-05,
      "loss": 0.3582,
      "step": 3001
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.305084745762712e-05,
      "loss": 0.4247,
      "step": 3002
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3048493408662904e-05,
      "loss": 0.3524,
      "step": 3003
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.304613935969868e-05,
      "loss": 0.3298,
      "step": 3004
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3043785310734465e-05,
      "loss": 0.4326,
      "step": 3005
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.304143126177025e-05,
      "loss": 0.5143,
      "step": 3006
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3039077212806026e-05,
      "loss": 0.4599,
      "step": 3007
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.303672316384181e-05,
      "loss": 0.3859,
      "step": 3008
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3034369114877594e-05,
      "loss": 0.3689,
      "step": 3009
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.303201506591337e-05,
      "loss": 0.386,
      "step": 3010
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.3029661016949155e-05,
      "loss": 0.3319,
      "step": 3011
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.302730696798494e-05,
      "loss": 0.4706,
      "step": 3012
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.3024952919020716e-05,
      "loss": 0.379,
      "step": 3013
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.30225988700565e-05,
      "loss": 0.3475,
      "step": 3014
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.3020244821092284e-05,
      "loss": 0.4698,
      "step": 3015
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.301789077212806e-05,
      "loss": 0.2794,
      "step": 3016
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.3015536723163845e-05,
      "loss": 0.2982,
      "step": 3017
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.301318267419962e-05,
      "loss": 0.4764,
      "step": 3018
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.3010828625235406e-05,
      "loss": 0.32,
      "step": 3019
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.300847457627119e-05,
      "loss": 0.4469,
      "step": 3020
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.300612052730697e-05,
      "loss": 0.3125,
      "step": 3021
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.300376647834275e-05,
      "loss": 0.2336,
      "step": 3022
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.3001412429378535e-05,
      "loss": 0.274,
      "step": 3023
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.299905838041431e-05,
      "loss": 0.331,
      "step": 3024
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.2996704331450096e-05,
      "loss": 0.3001,
      "step": 3025
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.299435028248588e-05,
      "loss": 0.353,
      "step": 3026
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.299199623352166e-05,
      "loss": 0.4307,
      "step": 3027
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.298964218455744e-05,
      "loss": 0.3507,
      "step": 3028
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.2987288135593225e-05,
      "loss": 0.3548,
      "step": 3029
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.2984934086629e-05,
      "loss": 0.397,
      "step": 3030
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.2982580037664786e-05,
      "loss": 0.3303,
      "step": 3031
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.298022598870057e-05,
      "loss": 0.3268,
      "step": 3032
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.297787193973635e-05,
      "loss": 0.5329,
      "step": 3033
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.297551789077213e-05,
      "loss": 0.3059,
      "step": 3034
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.297316384180791e-05,
      "loss": 0.3185,
      "step": 3035
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.297080979284369e-05,
      "loss": 0.4519,
      "step": 3036
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2968455743879476e-05,
      "loss": 0.271,
      "step": 3037
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2966101694915254e-05,
      "loss": 0.4813,
      "step": 3038
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.296374764595104e-05,
      "loss": 0.456,
      "step": 3039
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.296139359698682e-05,
      "loss": 0.3136,
      "step": 3040
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.29590395480226e-05,
      "loss": 0.3373,
      "step": 3041
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.295668549905838e-05,
      "loss": 0.3301,
      "step": 3042
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2954331450094166e-05,
      "loss": 0.3102,
      "step": 3043
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2951977401129944e-05,
      "loss": 0.3176,
      "step": 3044
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.294962335216573e-05,
      "loss": 0.3778,
      "step": 3045
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.294726930320151e-05,
      "loss": 0.3175,
      "step": 3046
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.294491525423729e-05,
      "loss": 0.2687,
      "step": 3047
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.294256120527307e-05,
      "loss": 0.3953,
      "step": 3048
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2940207156308856e-05,
      "loss": 0.2509,
      "step": 3049
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2937853107344634e-05,
      "loss": 0.482,
      "step": 3050
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.293549905838042e-05,
      "loss": 0.3088,
      "step": 3051
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2933145009416195e-05,
      "loss": 0.4305,
      "step": 3052
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.293079096045198e-05,
      "loss": 0.33,
      "step": 3053
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.292843691148776e-05,
      "loss": 0.5258,
      "step": 3054
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.292608286252354e-05,
      "loss": 0.3442,
      "step": 3055
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2923728813559324e-05,
      "loss": 0.3206,
      "step": 3056
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.292137476459511e-05,
      "loss": 0.2041,
      "step": 3057
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2919020715630885e-05,
      "loss": 0.2876,
      "step": 3058
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.291666666666667e-05,
      "loss": 0.3355,
      "step": 3059
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.291431261770245e-05,
      "loss": 0.3411,
      "step": 3060
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.291195856873823e-05,
      "loss": 0.3511,
      "step": 3061
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2909604519774014e-05,
      "loss": 0.3459,
      "step": 3062
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.29072504708098e-05,
      "loss": 0.4088,
      "step": 3063
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2904896421845575e-05,
      "loss": 0.3658,
      "step": 3064
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.290254237288136e-05,
      "loss": 0.404,
      "step": 3065
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.290018832391714e-05,
      "loss": 0.2589,
      "step": 3066
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.289783427495292e-05,
      "loss": 0.3315,
      "step": 3067
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2895480225988704e-05,
      "loss": 0.4433,
      "step": 3068
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.289312617702449e-05,
      "loss": 0.4018,
      "step": 3069
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2890772128060265e-05,
      "loss": 0.2432,
      "step": 3070
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.288841807909605e-05,
      "loss": 0.3892,
      "step": 3071
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2886064030131826e-05,
      "loss": 0.3804,
      "step": 3072
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.288370998116761e-05,
      "loss": 0.4063,
      "step": 3073
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2881355932203394e-05,
      "loss": 0.3925,
      "step": 3074
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.287900188323917e-05,
      "loss": 0.4649,
      "step": 3075
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.2876647834274955e-05,
      "loss": 0.2829,
      "step": 3076
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.287429378531074e-05,
      "loss": 0.3508,
      "step": 3077
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2871939736346516e-05,
      "loss": 0.3055,
      "step": 3078
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.28695856873823e-05,
      "loss": 0.3725,
      "step": 3079
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2867231638418084e-05,
      "loss": 0.4033,
      "step": 3080
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.286487758945386e-05,
      "loss": 0.2621,
      "step": 3081
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2862523540489645e-05,
      "loss": 0.364,
      "step": 3082
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.286016949152543e-05,
      "loss": 0.3372,
      "step": 3083
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2857815442561206e-05,
      "loss": 0.3313,
      "step": 3084
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.285546139359699e-05,
      "loss": 0.2142,
      "step": 3085
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2853107344632774e-05,
      "loss": 0.313,
      "step": 3086
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.285075329566855e-05,
      "loss": 0.4062,
      "step": 3087
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2848399246704335e-05,
      "loss": 0.4272,
      "step": 3088
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.284604519774011e-05,
      "loss": 0.3644,
      "step": 3089
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2843691148775896e-05,
      "loss": 0.4755,
      "step": 3090
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.284133709981168e-05,
      "loss": 0.3009,
      "step": 3091
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.283898305084746e-05,
      "loss": 0.32,
      "step": 3092
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.283662900188324e-05,
      "loss": 0.3327,
      "step": 3093
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2834274952919025e-05,
      "loss": 0.3836,
      "step": 3094
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.28319209039548e-05,
      "loss": 0.5186,
      "step": 3095
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.2829566854990586e-05,
      "loss": 0.4268,
      "step": 3096
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.282721280602637e-05,
      "loss": 0.3328,
      "step": 3097
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.282485875706215e-05,
      "loss": 0.3774,
      "step": 3098
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.282250470809793e-05,
      "loss": 0.4055,
      "step": 3099
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.2820150659133715e-05,
      "loss": 0.3739,
      "step": 3100
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.281779661016949e-05,
      "loss": 0.3447,
      "step": 3101
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.2815442561205276e-05,
      "loss": 0.2697,
      "step": 3102
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.281308851224106e-05,
      "loss": 0.2621,
      "step": 3103
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.281073446327684e-05,
      "loss": 0.3755,
      "step": 3104
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.280838041431262e-05,
      "loss": 0.3555,
      "step": 3105
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.28060263653484e-05,
      "loss": 0.3126,
      "step": 3106
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.280367231638418e-05,
      "loss": 0.3742,
      "step": 3107
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.2801318267419966e-05,
      "loss": 0.3558,
      "step": 3108
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.279896421845574e-05,
      "loss": 0.4095,
      "step": 3109
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.279661016949153e-05,
      "loss": 0.3707,
      "step": 3110
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.279425612052731e-05,
      "loss": 0.3236,
      "step": 3111
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.279190207156309e-05,
      "loss": 0.3311,
      "step": 3112
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.278954802259887e-05,
      "loss": 0.3742,
      "step": 3113
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.2787193973634656e-05,
      "loss": 0.3563,
      "step": 3114
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.278483992467043e-05,
      "loss": 0.357,
      "step": 3115
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.278248587570622e-05,
      "loss": 0.3223,
      "step": 3116
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.2780131826742e-05,
      "loss": 0.2803,
      "step": 3117
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.277777777777778e-05,
      "loss": 0.4271,
      "step": 3118
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.277542372881356e-05,
      "loss": 0.3031,
      "step": 3119
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.2773069679849346e-05,
      "loss": 0.517,
      "step": 3120
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.277071563088512e-05,
      "loss": 0.4653,
      "step": 3121
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.276836158192091e-05,
      "loss": 0.3251,
      "step": 3122
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.2766007532956684e-05,
      "loss": 0.2874,
      "step": 3123
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.2763653483992475e-05,
      "loss": 0.2918,
      "step": 3124
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.276129943502825e-05,
      "loss": 0.4091,
      "step": 3125
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.275894538606403e-05,
      "loss": 0.2631,
      "step": 3126
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.275659133709981e-05,
      "loss": 0.3513,
      "step": 3127
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.27542372881356e-05,
      "loss": 0.3402,
      "step": 3128
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.2751883239171374e-05,
      "loss": 0.2918,
      "step": 3129
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.274952919020716e-05,
      "loss": 0.3449,
      "step": 3130
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.274717514124294e-05,
      "loss": 0.2857,
      "step": 3131
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.274482109227872e-05,
      "loss": 0.4675,
      "step": 3132
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.27424670433145e-05,
      "loss": 0.2987,
      "step": 3133
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.274011299435029e-05,
      "loss": 0.418,
      "step": 3134
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.2737758945386064e-05,
      "loss": 0.3257,
      "step": 3135
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.273540489642185e-05,
      "loss": 0.4955,
      "step": 3136
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.273305084745763e-05,
      "loss": 0.471,
      "step": 3137
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.273069679849341e-05,
      "loss": 0.375,
      "step": 3138
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.272834274952919e-05,
      "loss": 0.4287,
      "step": 3139
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.272598870056497e-05,
      "loss": 0.3568,
      "step": 3140
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.272363465160076e-05,
      "loss": 0.4804,
      "step": 3141
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.272128060263654e-05,
      "loss": 0.3262,
      "step": 3142
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2718926553672315e-05,
      "loss": 0.3214,
      "step": 3143
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2716572504708106e-05,
      "loss": 0.4031,
      "step": 3144
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.271421845574388e-05,
      "loss": 0.263,
      "step": 3145
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.271186440677966e-05,
      "loss": 0.3872,
      "step": 3146
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2709510357815444e-05,
      "loss": 0.2715,
      "step": 3147
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.270715630885123e-05,
      "loss": 0.401,
      "step": 3148
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2704802259887005e-05,
      "loss": 0.2682,
      "step": 3149
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.270244821092279e-05,
      "loss": 0.3939,
      "step": 3150
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.270009416195857e-05,
      "loss": 0.4507,
      "step": 3151
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.269774011299435e-05,
      "loss": 0.4887,
      "step": 3152
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2695386064030134e-05,
      "loss": 0.3405,
      "step": 3153
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.269303201506592e-05,
      "loss": 0.3536,
      "step": 3154
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2690677966101695e-05,
      "loss": 0.3464,
      "step": 3155
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.268832391713748e-05,
      "loss": 0.2618,
      "step": 3156
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2685969868173256e-05,
      "loss": 0.2772,
      "step": 3157
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.268361581920905e-05,
      "loss": 0.511,
      "step": 3158
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.2681261770244824e-05,
      "loss": 0.2972,
      "step": 3159
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.26789077212806e-05,
      "loss": 0.261,
      "step": 3160
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.267655367231639e-05,
      "loss": 0.3146,
      "step": 3161
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.267419962335217e-05,
      "loss": 0.3121,
      "step": 3162
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2671845574387946e-05,
      "loss": 0.4454,
      "step": 3163
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.266949152542373e-05,
      "loss": 0.5027,
      "step": 3164
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2667137476459514e-05,
      "loss": 0.4263,
      "step": 3165
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.266478342749529e-05,
      "loss": 0.377,
      "step": 3166
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2662429378531075e-05,
      "loss": 0.3492,
      "step": 3167
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.266007532956686e-05,
      "loss": 0.3601,
      "step": 3168
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2657721280602636e-05,
      "loss": 0.4336,
      "step": 3169
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.265536723163842e-05,
      "loss": 0.2876,
      "step": 3170
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2653013182674204e-05,
      "loss": 0.4479,
      "step": 3171
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.265065913370999e-05,
      "loss": 0.4789,
      "step": 3172
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2648305084745765e-05,
      "loss": 0.5306,
      "step": 3173
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.264595103578154e-05,
      "loss": 0.3822,
      "step": 3174
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.264359698681733e-05,
      "loss": 0.2929,
      "step": 3175
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.264124293785311e-05,
      "loss": 0.3629,
      "step": 3176
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.263888888888889e-05,
      "loss": 0.3189,
      "step": 3177
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.263653483992468e-05,
      "loss": 0.2494,
      "step": 3178
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2634180790960455e-05,
      "loss": 0.3647,
      "step": 3179
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.263182674199623e-05,
      "loss": 0.3084,
      "step": 3180
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.2629472693032016e-05,
      "loss": 0.3602,
      "step": 3181
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.26271186440678e-05,
      "loss": 0.2997,
      "step": 3182
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.262476459510358e-05,
      "loss": 0.3563,
      "step": 3183
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.262241054613936e-05,
      "loss": 0.2267,
      "step": 3184
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2620056497175145e-05,
      "loss": 0.4389,
      "step": 3185
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.261770244821092e-05,
      "loss": 0.4264,
      "step": 3186
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2615348399246706e-05,
      "loss": 0.3153,
      "step": 3187
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.261299435028249e-05,
      "loss": 0.3622,
      "step": 3188
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2610640301318274e-05,
      "loss": 0.384,
      "step": 3189
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.260828625235405e-05,
      "loss": 0.4389,
      "step": 3190
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.260593220338983e-05,
      "loss": 0.2393,
      "step": 3191
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.260357815442562e-05,
      "loss": 0.4048,
      "step": 3192
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2601224105461396e-05,
      "loss": 0.3434,
      "step": 3193
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.259887005649717e-05,
      "loss": 0.2772,
      "step": 3194
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2596516007532964e-05,
      "loss": 0.3518,
      "step": 3195
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.259416195856874e-05,
      "loss": 0.4057,
      "step": 3196
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.259180790960452e-05,
      "loss": 0.2676,
      "step": 3197
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.25894538606403e-05,
      "loss": 0.593,
      "step": 3198
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2587099811676086e-05,
      "loss": 0.3141,
      "step": 3199
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.258474576271186e-05,
      "loss": 0.4511,
      "step": 3200
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.258239171374765e-05,
      "loss": 0.2833,
      "step": 3201
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.258003766478343e-05,
      "loss": 0.4353,
      "step": 3202
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2577683615819215e-05,
      "loss": 0.3256,
      "step": 3203
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.257532956685499e-05,
      "loss": 0.3297,
      "step": 3204
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.2572975517890776e-05,
      "loss": 0.3062,
      "step": 3205
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.257062146892656e-05,
      "loss": 0.3118,
      "step": 3206
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.256826741996234e-05,
      "loss": 0.3298,
      "step": 3207
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.2565913370998114e-05,
      "loss": 0.2962,
      "step": 3208
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.2563559322033905e-05,
      "loss": 0.2894,
      "step": 3209
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.256120527306968e-05,
      "loss": 0.4105,
      "step": 3210
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.255885122410546e-05,
      "loss": 0.4774,
      "step": 3211
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.255649717514125e-05,
      "loss": 0.4046,
      "step": 3212
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.255414312617703e-05,
      "loss": 0.674,
      "step": 3213
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.2551789077212804e-05,
      "loss": 0.4614,
      "step": 3214
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.254943502824859e-05,
      "loss": 0.4137,
      "step": 3215
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.254708097928437e-05,
      "loss": 0.4459,
      "step": 3216
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.254472693032015e-05,
      "loss": 0.5755,
      "step": 3217
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.254237288135593e-05,
      "loss": 0.347,
      "step": 3218
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.254001883239172e-05,
      "loss": 0.3965,
      "step": 3219
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.25376647834275e-05,
      "loss": 0.4126,
      "step": 3220
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.253531073446328e-05,
      "loss": 0.3681,
      "step": 3221
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.253295668549906e-05,
      "loss": 0.3765,
      "step": 3222
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.2530602636534846e-05,
      "loss": 0.3049,
      "step": 3223
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.252824858757062e-05,
      "loss": 0.3141,
      "step": 3224
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.25258945386064e-05,
      "loss": 0.4961,
      "step": 3225
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.252354048964219e-05,
      "loss": 0.3519,
      "step": 3226
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.252118644067797e-05,
      "loss": 0.287,
      "step": 3227
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2518832391713745e-05,
      "loss": 0.2471,
      "step": 3228
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2516478342749536e-05,
      "loss": 0.3591,
      "step": 3229
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.251412429378531e-05,
      "loss": 0.4715,
      "step": 3230
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.251177024482109e-05,
      "loss": 0.4504,
      "step": 3231
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2509416195856874e-05,
      "loss": 0.3054,
      "step": 3232
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.250706214689266e-05,
      "loss": 0.3547,
      "step": 3233
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2504708097928435e-05,
      "loss": 0.3656,
      "step": 3234
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.250235404896422e-05,
      "loss": 0.4668,
      "step": 3235
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.25e-05,
      "loss": 0.3038,
      "step": 3236
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.249764595103579e-05,
      "loss": 0.4148,
      "step": 3237
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2495291902071564e-05,
      "loss": 0.4396,
      "step": 3238
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.249293785310735e-05,
      "loss": 0.4057,
      "step": 3239
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.249058380414313e-05,
      "loss": 0.2989,
      "step": 3240
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.248822975517891e-05,
      "loss": 0.3261,
      "step": 3241
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2485875706214686e-05,
      "loss": 0.4206,
      "step": 3242
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.248352165725048e-05,
      "loss": 0.3372,
      "step": 3243
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.2481167608286254e-05,
      "loss": 0.3621,
      "step": 3244
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.247881355932203e-05,
      "loss": 0.5025,
      "step": 3245
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.247645951035782e-05,
      "loss": 0.3611,
      "step": 3246
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.24741054613936e-05,
      "loss": 0.4914,
      "step": 3247
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2471751412429376e-05,
      "loss": 0.4629,
      "step": 3248
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.246939736346516e-05,
      "loss": 0.3509,
      "step": 3249
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2467043314500944e-05,
      "loss": 0.2644,
      "step": 3250
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.246468926553673e-05,
      "loss": 0.3423,
      "step": 3251
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2462335216572505e-05,
      "loss": 0.3505,
      "step": 3252
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.245998116760829e-05,
      "loss": 0.3892,
      "step": 3253
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.245762711864407e-05,
      "loss": 0.3185,
      "step": 3254
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.245527306967985e-05,
      "loss": 0.3714,
      "step": 3255
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2452919020715634e-05,
      "loss": 0.3998,
      "step": 3256
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.245056497175142e-05,
      "loss": 0.5582,
      "step": 3257
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2448210922787195e-05,
      "loss": 0.2464,
      "step": 3258
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.244585687382297e-05,
      "loss": 0.3095,
      "step": 3259
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.244350282485876e-05,
      "loss": 0.465,
      "step": 3260
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.244114877589454e-05,
      "loss": 0.2385,
      "step": 3261
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.243879472693032e-05,
      "loss": 0.3075,
      "step": 3262
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.243644067796611e-05,
      "loss": 0.3798,
      "step": 3263
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2434086629001885e-05,
      "loss": 0.4469,
      "step": 3264
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.243173258003766e-05,
      "loss": 0.4539,
      "step": 3265
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2429378531073446e-05,
      "loss": 0.249,
      "step": 3266
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.242702448210923e-05,
      "loss": 0.4465,
      "step": 3267
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2424670433145014e-05,
      "loss": 0.2952,
      "step": 3268
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.242231638418079e-05,
      "loss": 0.3636,
      "step": 3269
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2419962335216575e-05,
      "loss": 0.2925,
      "step": 3270
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.241760828625236e-05,
      "loss": 0.4174,
      "step": 3271
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2415254237288136e-05,
      "loss": 0.3463,
      "step": 3272
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.241290018832392e-05,
      "loss": 0.3204,
      "step": 3273
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2410546139359704e-05,
      "loss": 0.3153,
      "step": 3274
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.240819209039548e-05,
      "loss": 0.3249,
      "step": 3275
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.240583804143126e-05,
      "loss": 0.4146,
      "step": 3276
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.240348399246705e-05,
      "loss": 0.3472,
      "step": 3277
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2401129943502826e-05,
      "loss": 0.3696,
      "step": 3278
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.23987758945386e-05,
      "loss": 0.4529,
      "step": 3279
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2396421845574394e-05,
      "loss": 0.2662,
      "step": 3280
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.239406779661017e-05,
      "loss": 0.3155,
      "step": 3281
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2391713747645955e-05,
      "loss": 0.4728,
      "step": 3282
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.238935969868173e-05,
      "loss": 0.2821,
      "step": 3283
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2387005649717516e-05,
      "loss": 0.248,
      "step": 3284
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.23846516007533e-05,
      "loss": 0.3099,
      "step": 3285
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.238229755178908e-05,
      "loss": 0.4452,
      "step": 3286
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.237994350282486e-05,
      "loss": 0.3915,
      "step": 3287
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.2377589453860645e-05,
      "loss": 0.5636,
      "step": 3288
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.237523540489642e-05,
      "loss": 0.2301,
      "step": 3289
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.2372881355932206e-05,
      "loss": 0.3036,
      "step": 3290
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.237052730696799e-05,
      "loss": 0.3961,
      "step": 3291
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.236817325800377e-05,
      "loss": 0.2977,
      "step": 3292
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.2365819209039544e-05,
      "loss": 0.3239,
      "step": 3293
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.2363465160075335e-05,
      "loss": 0.3532,
      "step": 3294
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.236111111111111e-05,
      "loss": 0.3458,
      "step": 3295
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.235875706214689e-05,
      "loss": 0.3732,
      "step": 3296
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.235640301318268e-05,
      "loss": 0.3505,
      "step": 3297
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.235404896421846e-05,
      "loss": 0.2934,
      "step": 3298
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.235169491525424e-05,
      "loss": 0.49,
      "step": 3299
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.234934086629002e-05,
      "loss": 0.3293,
      "step": 3300
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.23469868173258e-05,
      "loss": 0.3554,
      "step": 3301
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.2344632768361586e-05,
      "loss": 0.2945,
      "step": 3302
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.234227871939736e-05,
      "loss": 0.2878,
      "step": 3303
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.233992467043315e-05,
      "loss": 0.3584,
      "step": 3304
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.233757062146893e-05,
      "loss": 0.3281,
      "step": 3305
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.233521657250471e-05,
      "loss": 0.3832,
      "step": 3306
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.233286252354049e-05,
      "loss": 0.404,
      "step": 3307
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.2330508474576276e-05,
      "loss": 0.4399,
      "step": 3308
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.232815442561205e-05,
      "loss": 0.3027,
      "step": 3309
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.232580037664784e-05,
      "loss": 0.4054,
      "step": 3310
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.232344632768362e-05,
      "loss": 0.2449,
      "step": 3311
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.23210922787194e-05,
      "loss": 0.3089,
      "step": 3312
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.2318738229755176e-05,
      "loss": 0.3358,
      "step": 3313
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.2316384180790966e-05,
      "loss": 0.4598,
      "step": 3314
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.231403013182674e-05,
      "loss": 0.3262,
      "step": 3315
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.231167608286253e-05,
      "loss": 0.3449,
      "step": 3316
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.2309322033898304e-05,
      "loss": 0.4095,
      "step": 3317
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.230696798493409e-05,
      "loss": 0.2389,
      "step": 3318
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.230461393596987e-05,
      "loss": 0.3606,
      "step": 3319
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.230225988700565e-05,
      "loss": 0.3959,
      "step": 3320
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.229990583804143e-05,
      "loss": 0.4018,
      "step": 3321
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.229755178907722e-05,
      "loss": 0.3394,
      "step": 3322
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.2295197740112994e-05,
      "loss": 0.4139,
      "step": 3323
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.229284369114878e-05,
      "loss": 0.2225,
      "step": 3324
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.229048964218456e-05,
      "loss": 0.3729,
      "step": 3325
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.228813559322034e-05,
      "loss": 0.3831,
      "step": 3326
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.228578154425612e-05,
      "loss": 0.4162,
      "step": 3327
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.228342749529191e-05,
      "loss": 0.3911,
      "step": 3328
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.2281073446327684e-05,
      "loss": 0.2972,
      "step": 3329
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.227871939736347e-05,
      "loss": 0.2615,
      "step": 3330
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.227636534839925e-05,
      "loss": 0.4228,
      "step": 3331
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.227401129943503e-05,
      "loss": 0.4646,
      "step": 3332
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.227165725047081e-05,
      "loss": 0.3643,
      "step": 3333
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.226930320150659e-05,
      "loss": 0.3246,
      "step": 3334
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2266949152542374e-05,
      "loss": 0.3838,
      "step": 3335
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.226459510357816e-05,
      "loss": 0.4333,
      "step": 3336
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2262241054613935e-05,
      "loss": 0.3168,
      "step": 3337
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.225988700564972e-05,
      "loss": 0.2584,
      "step": 3338
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.22575329566855e-05,
      "loss": 0.3375,
      "step": 3339
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.225517890772128e-05,
      "loss": 0.3049,
      "step": 3340
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2252824858757064e-05,
      "loss": 0.3487,
      "step": 3341
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.225047080979285e-05,
      "loss": 0.4672,
      "step": 3342
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2248116760828625e-05,
      "loss": 0.4066,
      "step": 3343
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.224576271186441e-05,
      "loss": 0.3666,
      "step": 3344
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.224340866290019e-05,
      "loss": 0.3773,
      "step": 3345
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.224105461393597e-05,
      "loss": 0.4029,
      "step": 3346
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2238700564971754e-05,
      "loss": 0.2735,
      "step": 3347
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.223634651600754e-05,
      "loss": 0.3094,
      "step": 3348
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2233992467043315e-05,
      "loss": 0.4059,
      "step": 3349
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.22316384180791e-05,
      "loss": 0.363,
      "step": 3350
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2229284369114877e-05,
      "loss": 0.2973,
      "step": 3351
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.222693032015066e-05,
      "loss": 0.3273,
      "step": 3352
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.2224576271186444e-05,
      "loss": 0.4019,
      "step": 3353
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.222222222222222e-05,
      "loss": 0.4563,
      "step": 3354
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2219868173258005e-05,
      "loss": 0.4432,
      "step": 3355
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.221751412429379e-05,
      "loss": 0.4725,
      "step": 3356
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2215160075329567e-05,
      "loss": 0.4327,
      "step": 3357
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.221280602636535e-05,
      "loss": 0.3052,
      "step": 3358
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2210451977401134e-05,
      "loss": 0.3011,
      "step": 3359
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.220809792843691e-05,
      "loss": 0.3116,
      "step": 3360
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2205743879472695e-05,
      "loss": 0.274,
      "step": 3361
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.220338983050848e-05,
      "loss": 0.3576,
      "step": 3362
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2201035781544257e-05,
      "loss": 0.2957,
      "step": 3363
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.219868173258004e-05,
      "loss": 0.2739,
      "step": 3364
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2196327683615824e-05,
      "loss": 0.2917,
      "step": 3365
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.21939736346516e-05,
      "loss": 0.3055,
      "step": 3366
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2191619585687385e-05,
      "loss": 0.5123,
      "step": 3367
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.218926553672316e-05,
      "loss": 0.5251,
      "step": 3368
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2186911487758947e-05,
      "loss": 0.317,
      "step": 3369
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.218455743879473e-05,
      "loss": 0.179,
      "step": 3370
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.218220338983051e-05,
      "loss": 0.4097,
      "step": 3371
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.217984934086629e-05,
      "loss": 0.3635,
      "step": 3372
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.2177495291902075e-05,
      "loss": 0.3032,
      "step": 3373
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.217514124293785e-05,
      "loss": 0.3497,
      "step": 3374
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2172787193973637e-05,
      "loss": 0.321,
      "step": 3375
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.217043314500942e-05,
      "loss": 0.3413,
      "step": 3376
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.21680790960452e-05,
      "loss": 0.3757,
      "step": 3377
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.216572504708098e-05,
      "loss": 0.2995,
      "step": 3378
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2163370998116765e-05,
      "loss": 0.2707,
      "step": 3379
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.216101694915254e-05,
      "loss": 0.3229,
      "step": 3380
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2158662900188327e-05,
      "loss": 0.3733,
      "step": 3381
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.215630885122411e-05,
      "loss": 0.441,
      "step": 3382
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.215395480225989e-05,
      "loss": 0.2912,
      "step": 3383
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.215160075329567e-05,
      "loss": 0.3885,
      "step": 3384
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2149246704331455e-05,
      "loss": 0.2715,
      "step": 3385
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.214689265536723e-05,
      "loss": 0.4382,
      "step": 3386
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2144538606403017e-05,
      "loss": 0.3133,
      "step": 3387
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2142184557438794e-05,
      "loss": 0.4074,
      "step": 3388
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.213983050847458e-05,
      "loss": 0.3254,
      "step": 3389
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.213747645951036e-05,
      "loss": 0.39,
      "step": 3390
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.213512241054614e-05,
      "loss": 0.3868,
      "step": 3391
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.213276836158192e-05,
      "loss": 0.3381,
      "step": 3392
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2130414312617706e-05,
      "loss": 0.6521,
      "step": 3393
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.2128060263653484e-05,
      "loss": 0.322,
      "step": 3394
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.212570621468927e-05,
      "loss": 0.4645,
      "step": 3395
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.212335216572505e-05,
      "loss": 0.2788,
      "step": 3396
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.212099811676083e-05,
      "loss": 0.2932,
      "step": 3397
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.211864406779661e-05,
      "loss": 0.3572,
      "step": 3398
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.2116290018832396e-05,
      "loss": 0.3381,
      "step": 3399
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.2113935969868174e-05,
      "loss": 0.344,
      "step": 3400
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.211158192090396e-05,
      "loss": 0.2509,
      "step": 3401
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.210922787193974e-05,
      "loss": 0.295,
      "step": 3402
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.210687382297552e-05,
      "loss": 0.2841,
      "step": 3403
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.21045197740113e-05,
      "loss": 0.3192,
      "step": 3404
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.210216572504708e-05,
      "loss": 0.2735,
      "step": 3405
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.2099811676082864e-05,
      "loss": 0.5026,
      "step": 3406
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.209745762711865e-05,
      "loss": 0.2736,
      "step": 3407
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.2095103578154425e-05,
      "loss": 0.2409,
      "step": 3408
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.209274952919021e-05,
      "loss": 0.3775,
      "step": 3409
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.209039548022599e-05,
      "loss": 0.3507,
      "step": 3410
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.208804143126177e-05,
      "loss": 0.3945,
      "step": 3411
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.2085687382297554e-05,
      "loss": 0.3723,
      "step": 3412
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.208333333333334e-05,
      "loss": 0.3434,
      "step": 3413
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.2080979284369115e-05,
      "loss": 0.4315,
      "step": 3414
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.20786252354049e-05,
      "loss": 0.4015,
      "step": 3415
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.207627118644068e-05,
      "loss": 0.4905,
      "step": 3416
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.207391713747646e-05,
      "loss": 0.2975,
      "step": 3417
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2071563088512244e-05,
      "loss": 0.4449,
      "step": 3418
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.206920903954803e-05,
      "loss": 0.2685,
      "step": 3419
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2066854990583805e-05,
      "loss": 0.4715,
      "step": 3420
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.206450094161959e-05,
      "loss": 0.3588,
      "step": 3421
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2062146892655366e-05,
      "loss": 0.3287,
      "step": 3422
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.205979284369115e-05,
      "loss": 0.39,
      "step": 3423
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2057438794726934e-05,
      "loss": 0.4097,
      "step": 3424
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.205508474576271e-05,
      "loss": 0.2905,
      "step": 3425
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2052730696798495e-05,
      "loss": 0.2147,
      "step": 3426
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.205037664783428e-05,
      "loss": 0.2873,
      "step": 3427
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2048022598870056e-05,
      "loss": 0.4103,
      "step": 3428
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.204566854990584e-05,
      "loss": 0.4219,
      "step": 3429
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2043314500941624e-05,
      "loss": 0.2975,
      "step": 3430
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.20409604519774e-05,
      "loss": 0.3542,
      "step": 3431
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2038606403013185e-05,
      "loss": 0.4903,
      "step": 3432
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.203625235404897e-05,
      "loss": 0.2717,
      "step": 3433
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2033898305084746e-05,
      "loss": 0.3295,
      "step": 3434
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.203154425612053e-05,
      "loss": 0.2828,
      "step": 3435
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2029190207156314e-05,
      "loss": 0.3796,
      "step": 3436
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.202683615819209e-05,
      "loss": 0.4208,
      "step": 3437
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.2024482109227875e-05,
      "loss": 0.3022,
      "step": 3438
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.202212806026365e-05,
      "loss": 0.4546,
      "step": 3439
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.201977401129944e-05,
      "loss": 0.445,
      "step": 3440
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.201741996233522e-05,
      "loss": 0.4128,
      "step": 3441
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.2015065913371e-05,
      "loss": 0.3269,
      "step": 3442
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.201271186440678e-05,
      "loss": 0.3603,
      "step": 3443
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.2010357815442565e-05,
      "loss": 0.4121,
      "step": 3444
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.200800376647834e-05,
      "loss": 0.4098,
      "step": 3445
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.2005649717514126e-05,
      "loss": 0.4213,
      "step": 3446
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.200329566854991e-05,
      "loss": 0.2908,
      "step": 3447
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.200094161958569e-05,
      "loss": 0.3033,
      "step": 3448
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.199858757062147e-05,
      "loss": 0.3265,
      "step": 3449
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.1996233521657255e-05,
      "loss": 0.3366,
      "step": 3450
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.199387947269303e-05,
      "loss": 0.3231,
      "step": 3451
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.1991525423728816e-05,
      "loss": 0.3705,
      "step": 3452
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.19891713747646e-05,
      "loss": 0.3081,
      "step": 3453
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.198681732580038e-05,
      "loss": 0.3322,
      "step": 3454
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.198446327683616e-05,
      "loss": 0.2886,
      "step": 3455
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.198210922787194e-05,
      "loss": 0.3926,
      "step": 3456
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.197975517890773e-05,
      "loss": 0.2733,
      "step": 3457
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.1977401129943506e-05,
      "loss": 0.358,
      "step": 3458
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.197504708097928e-05,
      "loss": 0.336,
      "step": 3459
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.1972693032015074e-05,
      "loss": 0.3106,
      "step": 3460
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.197033898305085e-05,
      "loss": 0.3416,
      "step": 3461
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.196798493408663e-05,
      "loss": 0.3755,
      "step": 3462
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.196563088512241e-05,
      "loss": 0.2685,
      "step": 3463
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.1963276836158196e-05,
      "loss": 0.4633,
      "step": 3464
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.196092278719397e-05,
      "loss": 0.2442,
      "step": 3465
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.195856873822976e-05,
      "loss": 0.2893,
      "step": 3466
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.195621468926554e-05,
      "loss": 0.4047,
      "step": 3467
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.195386064030132e-05,
      "loss": 0.2612,
      "step": 3468
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.19515065913371e-05,
      "loss": 0.379,
      "step": 3469
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.1949152542372886e-05,
      "loss": 0.3659,
      "step": 3470
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.194679849340866e-05,
      "loss": 0.254,
      "step": 3471
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.194444444444445e-05,
      "loss": 0.4041,
      "step": 3472
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.1942090395480224e-05,
      "loss": 0.5208,
      "step": 3473
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.1939736346516015e-05,
      "loss": 0.4372,
      "step": 3474
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.193738229755179e-05,
      "loss": 0.3101,
      "step": 3475
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.193502824858757e-05,
      "loss": 0.4078,
      "step": 3476
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.193267419962336e-05,
      "loss": 0.3296,
      "step": 3477
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.193032015065914e-05,
      "loss": 0.4229,
      "step": 3478
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.1927966101694914e-05,
      "loss": 0.3967,
      "step": 3479
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.19256120527307e-05,
      "loss": 0.3783,
      "step": 3480
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.192325800376648e-05,
      "loss": 0.3343,
      "step": 3481
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.192090395480226e-05,
      "loss": 0.3151,
      "step": 3482
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.191854990583804e-05,
      "loss": 0.4266,
      "step": 3483
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.191619585687383e-05,
      "loss": 0.3678,
      "step": 3484
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.1913841807909604e-05,
      "loss": 0.3794,
      "step": 3485
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.191148775894539e-05,
      "loss": 0.3685,
      "step": 3486
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.190913370998117e-05,
      "loss": 0.3165,
      "step": 3487
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.1906779661016956e-05,
      "loss": 0.2414,
      "step": 3488
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.190442561205273e-05,
      "loss": 0.3447,
      "step": 3489
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.190207156308851e-05,
      "loss": 0.3056,
      "step": 3490
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.18997175141243e-05,
      "loss": 0.31,
      "step": 3491
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.189736346516008e-05,
      "loss": 0.1767,
      "step": 3492
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.1895009416195855e-05,
      "loss": 0.2968,
      "step": 3493
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.1892655367231646e-05,
      "loss": 0.3141,
      "step": 3494
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.189030131826742e-05,
      "loss": 0.4384,
      "step": 3495
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.18879472693032e-05,
      "loss": 0.2893,
      "step": 3496
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.1885593220338984e-05,
      "loss": 0.3286,
      "step": 3497
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.188323917137477e-05,
      "loss": 0.3504,
      "step": 3498
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.1880885122410545e-05,
      "loss": 0.3075,
      "step": 3499
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.187853107344633e-05,
      "loss": 0.3887,
      "step": 3500
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.187617702448211e-05,
      "loss": 0.2818,
      "step": 3501
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.187382297551789e-05,
      "loss": 0.3308,
      "step": 3502
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1871468926553674e-05,
      "loss": 0.4911,
      "step": 3503
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.186911487758946e-05,
      "loss": 0.3996,
      "step": 3504
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.186676082862524e-05,
      "loss": 0.4467,
      "step": 3505
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.186440677966102e-05,
      "loss": 0.2695,
      "step": 3506
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1862052730696796e-05,
      "loss": 0.3275,
      "step": 3507
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.185969868173259e-05,
      "loss": 0.3268,
      "step": 3508
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1857344632768364e-05,
      "loss": 0.3121,
      "step": 3509
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.185499058380414e-05,
      "loss": 0.3408,
      "step": 3510
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.185263653483993e-05,
      "loss": 0.2405,
      "step": 3511
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.185028248587571e-05,
      "loss": 0.2813,
      "step": 3512
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1847928436911486e-05,
      "loss": 0.3204,
      "step": 3513
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.184557438794727e-05,
      "loss": 0.5245,
      "step": 3514
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1843220338983054e-05,
      "loss": 0.4202,
      "step": 3515
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.184086629001883e-05,
      "loss": 0.2608,
      "step": 3516
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1838512241054615e-05,
      "loss": 0.3965,
      "step": 3517
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.18361581920904e-05,
      "loss": 0.3328,
      "step": 3518
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.183380414312618e-05,
      "loss": 0.2978,
      "step": 3519
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.183145009416196e-05,
      "loss": 0.3512,
      "step": 3520
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1829096045197744e-05,
      "loss": 0.2921,
      "step": 3521
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.182674199623353e-05,
      "loss": 0.3241,
      "step": 3522
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.1824387947269305e-05,
      "loss": 0.4209,
      "step": 3523
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.182203389830508e-05,
      "loss": 0.3281,
      "step": 3524
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.181967984934087e-05,
      "loss": 0.2821,
      "step": 3525
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.181732580037665e-05,
      "loss": 0.4379,
      "step": 3526
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.181497175141243e-05,
      "loss": 0.4275,
      "step": 3527
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.181261770244822e-05,
      "loss": 0.4434,
      "step": 3528
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.1810263653483995e-05,
      "loss": 0.4084,
      "step": 3529
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.180790960451977e-05,
      "loss": 0.5475,
      "step": 3530
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.1805555555555556e-05,
      "loss": 0.2645,
      "step": 3531
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.180320150659134e-05,
      "loss": 0.3739,
      "step": 3532
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.180084745762712e-05,
      "loss": 0.3264,
      "step": 3533
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.17984934086629e-05,
      "loss": 0.3516,
      "step": 3534
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.1796139359698685e-05,
      "loss": 0.3307,
      "step": 3535
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.179378531073447e-05,
      "loss": 0.3057,
      "step": 3536
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.1791431261770246e-05,
      "loss": 0.2987,
      "step": 3537
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.178907721280603e-05,
      "loss": 0.3329,
      "step": 3538
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.1786723163841814e-05,
      "loss": 0.412,
      "step": 3539
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.178436911487759e-05,
      "loss": 0.4517,
      "step": 3540
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.178201506591337e-05,
      "loss": 0.366,
      "step": 3541
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.177966101694916e-05,
      "loss": 0.4419,
      "step": 3542
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.1777306967984936e-05,
      "loss": 0.2259,
      "step": 3543
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.177495291902071e-05,
      "loss": 0.26,
      "step": 3544
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1772598870056504e-05,
      "loss": 0.3708,
      "step": 3545
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.177024482109228e-05,
      "loss": 0.4396,
      "step": 3546
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.176789077212806e-05,
      "loss": 0.4213,
      "step": 3547
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.176553672316384e-05,
      "loss": 0.3226,
      "step": 3548
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1763182674199626e-05,
      "loss": 0.256,
      "step": 3549
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.17608286252354e-05,
      "loss": 0.3604,
      "step": 3550
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.175847457627119e-05,
      "loss": 0.2419,
      "step": 3551
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.175612052730697e-05,
      "loss": 0.4155,
      "step": 3552
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1753766478342755e-05,
      "loss": 0.3894,
      "step": 3553
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.175141242937853e-05,
      "loss": 0.3773,
      "step": 3554
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1749058380414316e-05,
      "loss": 0.251,
      "step": 3555
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.17467043314501e-05,
      "loss": 0.4264,
      "step": 3556
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.174435028248588e-05,
      "loss": 0.3409,
      "step": 3557
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1741996233521654e-05,
      "loss": 0.3461,
      "step": 3558
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1739642184557445e-05,
      "loss": 0.2608,
      "step": 3559
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.173728813559322e-05,
      "loss": 0.4705,
      "step": 3560
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1734934086629e-05,
      "loss": 0.3048,
      "step": 3561
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.173258003766479e-05,
      "loss": 0.3204,
      "step": 3562
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.173022598870057e-05,
      "loss": 0.3213,
      "step": 3563
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.1727871939736344e-05,
      "loss": 0.4579,
      "step": 3564
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.172551789077213e-05,
      "loss": 0.3893,
      "step": 3565
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.172316384180791e-05,
      "loss": 0.3343,
      "step": 3566
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.1720809792843696e-05,
      "loss": 0.3733,
      "step": 3567
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.171845574387947e-05,
      "loss": 0.3359,
      "step": 3568
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.171610169491526e-05,
      "loss": 0.3532,
      "step": 3569
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.171374764595104e-05,
      "loss": 0.3578,
      "step": 3570
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.171139359698682e-05,
      "loss": 0.3985,
      "step": 3571
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.17090395480226e-05,
      "loss": 0.3282,
      "step": 3572
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.1706685499058386e-05,
      "loss": 0.4852,
      "step": 3573
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.170433145009416e-05,
      "loss": 0.3116,
      "step": 3574
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.170197740112994e-05,
      "loss": 0.3296,
      "step": 3575
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.169962335216573e-05,
      "loss": 0.3398,
      "step": 3576
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.169726930320151e-05,
      "loss": 0.2752,
      "step": 3577
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.1694915254237285e-05,
      "loss": 0.3873,
      "step": 3578
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.1692561205273076e-05,
      "loss": 0.2099,
      "step": 3579
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.169020715630885e-05,
      "loss": 0.3954,
      "step": 3580
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.168785310734463e-05,
      "loss": 0.261,
      "step": 3581
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.1685499058380414e-05,
      "loss": 0.3071,
      "step": 3582
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.16831450094162e-05,
      "loss": 0.4652,
      "step": 3583
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.168079096045198e-05,
      "loss": 0.3058,
      "step": 3584
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.167843691148776e-05,
      "loss": 0.3226,
      "step": 3585
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.167608286252354e-05,
      "loss": 0.3871,
      "step": 3586
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.167372881355933e-05,
      "loss": 0.2255,
      "step": 3587
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1671374764595104e-05,
      "loss": 0.2454,
      "step": 3588
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.166902071563089e-05,
      "loss": 0.2206,
      "step": 3589
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.166666666666667e-05,
      "loss": 0.2466,
      "step": 3590
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.166431261770245e-05,
      "loss": 0.4421,
      "step": 3591
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1661958568738226e-05,
      "loss": 0.3376,
      "step": 3592
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.165960451977402e-05,
      "loss": 0.4525,
      "step": 3593
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1657250470809794e-05,
      "loss": 0.3043,
      "step": 3594
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.165489642184557e-05,
      "loss": 0.4159,
      "step": 3595
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.165254237288136e-05,
      "loss": 0.3296,
      "step": 3596
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.165018832391714e-05,
      "loss": 0.3585,
      "step": 3597
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.164783427495292e-05,
      "loss": 0.2253,
      "step": 3598
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.16454802259887e-05,
      "loss": 0.3405,
      "step": 3599
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1643126177024484e-05,
      "loss": 0.4103,
      "step": 3600
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.164077212806027e-05,
      "loss": 0.2941,
      "step": 3601
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1638418079096045e-05,
      "loss": 0.4313,
      "step": 3602
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.163606403013183e-05,
      "loss": 0.2885,
      "step": 3603
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.163370998116761e-05,
      "loss": 0.4736,
      "step": 3604
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.163135593220339e-05,
      "loss": 0.321,
      "step": 3605
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1629001883239174e-05,
      "loss": 0.2951,
      "step": 3606
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.162664783427496e-05,
      "loss": 0.2732,
      "step": 3607
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.1624293785310735e-05,
      "loss": 0.4192,
      "step": 3608
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.162193973634651e-05,
      "loss": 0.256,
      "step": 3609
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.16195856873823e-05,
      "loss": 0.385,
      "step": 3610
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.161723163841808e-05,
      "loss": 0.4116,
      "step": 3611
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.161487758945386e-05,
      "loss": 0.2088,
      "step": 3612
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.161252354048965e-05,
      "loss": 0.3683,
      "step": 3613
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1610169491525425e-05,
      "loss": 0.4134,
      "step": 3614
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.160781544256121e-05,
      "loss": 0.2857,
      "step": 3615
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1605461393596986e-05,
      "loss": 0.4794,
      "step": 3616
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.160310734463277e-05,
      "loss": 0.4198,
      "step": 3617
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1600753295668554e-05,
      "loss": 0.3698,
      "step": 3618
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.159839924670433e-05,
      "loss": 0.3397,
      "step": 3619
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1596045197740115e-05,
      "loss": 0.4524,
      "step": 3620
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.15936911487759e-05,
      "loss": 0.258,
      "step": 3621
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1591337099811676e-05,
      "loss": 0.3911,
      "step": 3622
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.158898305084746e-05,
      "loss": 0.3082,
      "step": 3623
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1586629001883244e-05,
      "loss": 0.3703,
      "step": 3624
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.158427495291902e-05,
      "loss": 0.3142,
      "step": 3625
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1581920903954805e-05,
      "loss": 0.3547,
      "step": 3626
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.157956685499059e-05,
      "loss": 0.3614,
      "step": 3627
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1577212806026366e-05,
      "loss": 0.3961,
      "step": 3628
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.1574858757062143e-05,
      "loss": 0.2354,
      "step": 3629
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1572504708097934e-05,
      "loss": 0.3501,
      "step": 3630
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.157015065913371e-05,
      "loss": 0.3103,
      "step": 3631
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1567796610169495e-05,
      "loss": 0.3736,
      "step": 3632
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.156544256120527e-05,
      "loss": 0.4262,
      "step": 3633
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1563088512241056e-05,
      "loss": 0.3776,
      "step": 3634
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.156073446327684e-05,
      "loss": 0.3423,
      "step": 3635
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.155838041431262e-05,
      "loss": 0.3781,
      "step": 3636
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.15560263653484e-05,
      "loss": 0.3143,
      "step": 3637
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1553672316384185e-05,
      "loss": 0.2675,
      "step": 3638
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.155131826741996e-05,
      "loss": 0.2784,
      "step": 3639
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1548964218455746e-05,
      "loss": 0.3319,
      "step": 3640
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.154661016949153e-05,
      "loss": 0.2917,
      "step": 3641
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.154425612052731e-05,
      "loss": 0.3627,
      "step": 3642
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.154190207156309e-05,
      "loss": 0.3191,
      "step": 3643
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1539548022598875e-05,
      "loss": 0.4893,
      "step": 3644
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.153719397363465e-05,
      "loss": 0.4374,
      "step": 3645
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1534839924670436e-05,
      "loss": 0.4624,
      "step": 3646
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.153248587570622e-05,
      "loss": 0.2273,
      "step": 3647
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.1530131826742e-05,
      "loss": 0.3877,
      "step": 3648
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.152777777777778e-05,
      "loss": 0.2587,
      "step": 3649
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.152542372881356e-05,
      "loss": 0.2417,
      "step": 3650
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.152306967984934e-05,
      "loss": 0.4524,
      "step": 3651
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1520715630885126e-05,
      "loss": 0.4265,
      "step": 3652
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1518361581920903e-05,
      "loss": 0.305,
      "step": 3653
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.151600753295669e-05,
      "loss": 0.4563,
      "step": 3654
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.151365348399247e-05,
      "loss": 0.385,
      "step": 3655
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.151129943502825e-05,
      "loss": 0.303,
      "step": 3656
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.150894538606403e-05,
      "loss": 0.3497,
      "step": 3657
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1506591337099816e-05,
      "loss": 0.2474,
      "step": 3658
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1504237288135593e-05,
      "loss": 0.2029,
      "step": 3659
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.150188323917138e-05,
      "loss": 0.249,
      "step": 3660
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.149952919020716e-05,
      "loss": 0.4338,
      "step": 3661
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.149717514124294e-05,
      "loss": 0.273,
      "step": 3662
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.149482109227872e-05,
      "loss": 0.2318,
      "step": 3663
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1492467043314506e-05,
      "loss": 0.3579,
      "step": 3664
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1490112994350283e-05,
      "loss": 0.2762,
      "step": 3665
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.148775894538607e-05,
      "loss": 0.3467,
      "step": 3666
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1485404896421844e-05,
      "loss": 0.3095,
      "step": 3667
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.148305084745763e-05,
      "loss": 0.4195,
      "step": 3668
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.148069679849341e-05,
      "loss": 0.503,
      "step": 3669
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.147834274952919e-05,
      "loss": 0.2321,
      "step": 3670
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.1475988700564973e-05,
      "loss": 0.2527,
      "step": 3671
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.147363465160076e-05,
      "loss": 0.2803,
      "step": 3672
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1471280602636534e-05,
      "loss": 0.2557,
      "step": 3673
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.146892655367232e-05,
      "loss": 0.3734,
      "step": 3674
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.14665725047081e-05,
      "loss": 0.2977,
      "step": 3675
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.146421845574388e-05,
      "loss": 0.4251,
      "step": 3676
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1461864406779663e-05,
      "loss": 0.3963,
      "step": 3677
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.145951035781545e-05,
      "loss": 0.4627,
      "step": 3678
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1457156308851224e-05,
      "loss": 0.5638,
      "step": 3679
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.145480225988701e-05,
      "loss": 0.3628,
      "step": 3680
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.145244821092279e-05,
      "loss": 0.3492,
      "step": 3681
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.145009416195857e-05,
      "loss": 0.4113,
      "step": 3682
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1447740112994353e-05,
      "loss": 0.3381,
      "step": 3683
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.144538606403013e-05,
      "loss": 0.5235,
      "step": 3684
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1443032015065914e-05,
      "loss": 0.3168,
      "step": 3685
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.14406779661017e-05,
      "loss": 0.4324,
      "step": 3686
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1438323917137476e-05,
      "loss": 0.5159,
      "step": 3687
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.143596986817326e-05,
      "loss": 0.3007,
      "step": 3688
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1433615819209043e-05,
      "loss": 0.3441,
      "step": 3689
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.143126177024482e-05,
      "loss": 0.2744,
      "step": 3690
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1428907721280604e-05,
      "loss": 0.2471,
      "step": 3691
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.142655367231639e-05,
      "loss": 0.2224,
      "step": 3692
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.1424199623352166e-05,
      "loss": 0.3197,
      "step": 3693
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.142184557438795e-05,
      "loss": 0.2769,
      "step": 3694
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1419491525423733e-05,
      "loss": 0.3861,
      "step": 3695
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.141713747645951e-05,
      "loss": 0.2618,
      "step": 3696
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1414783427495294e-05,
      "loss": 0.3247,
      "step": 3697
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.141242937853108e-05,
      "loss": 0.3266,
      "step": 3698
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1410075329566856e-05,
      "loss": 0.4857,
      "step": 3699
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.140772128060264e-05,
      "loss": 0.4455,
      "step": 3700
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1405367231638423e-05,
      "loss": 0.417,
      "step": 3701
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.14030131826742e-05,
      "loss": 0.2848,
      "step": 3702
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1400659133709984e-05,
      "loss": 0.3504,
      "step": 3703
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.139830508474576e-05,
      "loss": 0.2984,
      "step": 3704
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1395951035781546e-05,
      "loss": 0.3203,
      "step": 3705
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.139359698681733e-05,
      "loss": 0.2847,
      "step": 3706
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1391242937853107e-05,
      "loss": 0.3842,
      "step": 3707
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.138888888888889e-05,
      "loss": 0.2776,
      "step": 3708
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1386534839924674e-05,
      "loss": 0.3706,
      "step": 3709
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.138418079096045e-05,
      "loss": 0.3357,
      "step": 3710
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1381826741996236e-05,
      "loss": 0.3832,
      "step": 3711
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.137947269303202e-05,
      "loss": 0.3592,
      "step": 3712
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1377118644067797e-05,
      "loss": 0.2515,
      "step": 3713
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.137476459510358e-05,
      "loss": 0.3177,
      "step": 3714
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.1372410546139364e-05,
      "loss": 0.3099,
      "step": 3715
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.137005649717514e-05,
      "loss": 0.3384,
      "step": 3716
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1367702448210926e-05,
      "loss": 0.4076,
      "step": 3717
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.136534839924671e-05,
      "loss": 0.2519,
      "step": 3718
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1362994350282487e-05,
      "loss": 0.319,
      "step": 3719
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.136064030131827e-05,
      "loss": 0.2549,
      "step": 3720
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.135828625235405e-05,
      "loss": 0.3299,
      "step": 3721
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.135593220338983e-05,
      "loss": 0.3057,
      "step": 3722
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1353578154425616e-05,
      "loss": 0.3441,
      "step": 3723
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.135122410546139e-05,
      "loss": 0.2855,
      "step": 3724
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1348870056497177e-05,
      "loss": 0.3675,
      "step": 3725
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.134651600753296e-05,
      "loss": 0.3261,
      "step": 3726
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.134416195856874e-05,
      "loss": 0.4974,
      "step": 3727
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.134180790960452e-05,
      "loss": 0.3834,
      "step": 3728
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1339453860640306e-05,
      "loss": 0.3359,
      "step": 3729
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.133709981167608e-05,
      "loss": 0.379,
      "step": 3730
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1334745762711867e-05,
      "loss": 0.3222,
      "step": 3731
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.133239171374765e-05,
      "loss": 0.302,
      "step": 3732
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.133003766478343e-05,
      "loss": 0.2918,
      "step": 3733
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.132768361581921e-05,
      "loss": 0.3116,
      "step": 3734
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.1325329566854996e-05,
      "loss": 0.3656,
      "step": 3735
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.132297551789077e-05,
      "loss": 0.4063,
      "step": 3736
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.1320621468926557e-05,
      "loss": 0.182,
      "step": 3737
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.1318267419962334e-05,
      "loss": 0.2696,
      "step": 3738
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.131591337099812e-05,
      "loss": 0.4208,
      "step": 3739
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.13135593220339e-05,
      "loss": 0.3508,
      "step": 3740
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.131120527306968e-05,
      "loss": 0.3113,
      "step": 3741
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.130885122410546e-05,
      "loss": 0.4219,
      "step": 3742
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.1306497175141247e-05,
      "loss": 0.3859,
      "step": 3743
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.1304143126177024e-05,
      "loss": 0.2985,
      "step": 3744
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.130178907721281e-05,
      "loss": 0.4313,
      "step": 3745
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.129943502824859e-05,
      "loss": 0.4577,
      "step": 3746
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.129708097928437e-05,
      "loss": 0.4215,
      "step": 3747
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.129472693032015e-05,
      "loss": 0.2643,
      "step": 3748
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.1292372881355937e-05,
      "loss": 0.2658,
      "step": 3749
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.1290018832391714e-05,
      "loss": 0.406,
      "step": 3750
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.12876647834275e-05,
      "loss": 0.3976,
      "step": 3751
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.128531073446328e-05,
      "loss": 0.3477,
      "step": 3752
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.128295668549906e-05,
      "loss": 0.2937,
      "step": 3753
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.128060263653484e-05,
      "loss": 0.3756,
      "step": 3754
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.127824858757062e-05,
      "loss": 0.3364,
      "step": 3755
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.127589453860641e-05,
      "loss": 0.4324,
      "step": 3756
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.127354048964219e-05,
      "loss": 0.3508,
      "step": 3757
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1271186440677965e-05,
      "loss": 0.4147,
      "step": 3758
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.126883239171375e-05,
      "loss": 0.3941,
      "step": 3759
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.126647834274953e-05,
      "loss": 0.2753,
      "step": 3760
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.126412429378531e-05,
      "loss": 0.3193,
      "step": 3761
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1261770244821094e-05,
      "loss": 0.293,
      "step": 3762
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.125941619585688e-05,
      "loss": 0.3931,
      "step": 3763
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1257062146892655e-05,
      "loss": 0.3706,
      "step": 3764
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.125470809792844e-05,
      "loss": 0.2613,
      "step": 3765
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.125235404896422e-05,
      "loss": 0.3778,
      "step": 3766
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.125e-05,
      "loss": 0.2308,
      "step": 3767
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1247645951035784e-05,
      "loss": 0.4068,
      "step": 3768
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.124529190207157e-05,
      "loss": 0.3936,
      "step": 3769
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1242937853107345e-05,
      "loss": 0.3691,
      "step": 3770
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.124058380414313e-05,
      "loss": 0.3389,
      "step": 3771
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1238229755178906e-05,
      "loss": 0.2693,
      "step": 3772
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1235875706214697e-05,
      "loss": 0.2658,
      "step": 3773
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1233521657250474e-05,
      "loss": 0.3653,
      "step": 3774
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.123116760828625e-05,
      "loss": 0.3527,
      "step": 3775
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1228813559322035e-05,
      "loss": 0.3895,
      "step": 3776
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.122645951035782e-05,
      "loss": 0.275,
      "step": 3777
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.1224105461393596e-05,
      "loss": 0.415,
      "step": 3778
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.122175141242938e-05,
      "loss": 0.3604,
      "step": 3779
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1219397363465164e-05,
      "loss": 0.2926,
      "step": 3780
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.121704331450094e-05,
      "loss": 0.2788,
      "step": 3781
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1214689265536725e-05,
      "loss": 0.2117,
      "step": 3782
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.121233521657251e-05,
      "loss": 0.2379,
      "step": 3783
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1209981167608286e-05,
      "loss": 0.3198,
      "step": 3784
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.120762711864407e-05,
      "loss": 0.2637,
      "step": 3785
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1205273069679854e-05,
      "loss": 0.3162,
      "step": 3786
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.120291902071563e-05,
      "loss": 0.4621,
      "step": 3787
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1200564971751415e-05,
      "loss": 0.2862,
      "step": 3788
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.119821092278719e-05,
      "loss": 0.3977,
      "step": 3789
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.119585687382298e-05,
      "loss": 0.3807,
      "step": 3790
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.119350282485876e-05,
      "loss": 0.3578,
      "step": 3791
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.119114877589454e-05,
      "loss": 0.169,
      "step": 3792
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.118879472693033e-05,
      "loss": 0.3856,
      "step": 3793
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1186440677966105e-05,
      "loss": 0.3446,
      "step": 3794
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.118408662900188e-05,
      "loss": 0.3474,
      "step": 3795
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1181732580037666e-05,
      "loss": 0.2823,
      "step": 3796
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.117937853107345e-05,
      "loss": 0.2553,
      "step": 3797
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.117702448210923e-05,
      "loss": 0.2717,
      "step": 3798
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.117467043314501e-05,
      "loss": 0.286,
      "step": 3799
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.1172316384180795e-05,
      "loss": 0.2908,
      "step": 3800
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.116996233521657e-05,
      "loss": 0.4076,
      "step": 3801
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.1167608286252356e-05,
      "loss": 0.3347,
      "step": 3802
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.116525423728814e-05,
      "loss": 0.3313,
      "step": 3803
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.1162900188323924e-05,
      "loss": 0.2896,
      "step": 3804
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.11605461393597e-05,
      "loss": 0.3105,
      "step": 3805
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.115819209039548e-05,
      "loss": 0.381,
      "step": 3806
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.115583804143127e-05,
      "loss": 0.3326,
      "step": 3807
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.1153483992467046e-05,
      "loss": 0.2198,
      "step": 3808
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.115112994350282e-05,
      "loss": 0.2879,
      "step": 3809
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.1148775894538614e-05,
      "loss": 0.443,
      "step": 3810
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.114642184557439e-05,
      "loss": 0.3747,
      "step": 3811
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.114406779661017e-05,
      "loss": 0.4125,
      "step": 3812
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.114171374764595e-05,
      "loss": 0.438,
      "step": 3813
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.1139359698681736e-05,
      "loss": 0.2555,
      "step": 3814
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.113700564971751e-05,
      "loss": 0.4151,
      "step": 3815
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.11346516007533e-05,
      "loss": 0.3101,
      "step": 3816
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.113229755178908e-05,
      "loss": 0.3484,
      "step": 3817
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.112994350282486e-05,
      "loss": 0.1995,
      "step": 3818
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.112758945386064e-05,
      "loss": 0.2558,
      "step": 3819
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.1125235404896426e-05,
      "loss": 0.225,
      "step": 3820
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.112288135593221e-05,
      "loss": 0.2593,
      "step": 3821
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.112052730696799e-05,
      "loss": 0.2417,
      "step": 3822
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.1118173258003764e-05,
      "loss": 0.4145,
      "step": 3823
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.1115819209039555e-05,
      "loss": 0.3914,
      "step": 3824
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.111346516007533e-05,
      "loss": 0.2144,
      "step": 3825
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.111111111111111e-05,
      "loss": 0.2748,
      "step": 3826
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.11087570621469e-05,
      "loss": 0.3054,
      "step": 3827
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.110640301318268e-05,
      "loss": 0.3674,
      "step": 3828
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.1104048964218454e-05,
      "loss": 0.3268,
      "step": 3829
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.110169491525424e-05,
      "loss": 0.3073,
      "step": 3830
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.109934086629002e-05,
      "loss": 0.3917,
      "step": 3831
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.10969868173258e-05,
      "loss": 0.3872,
      "step": 3832
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.109463276836158e-05,
      "loss": 0.2811,
      "step": 3833
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.109227871939737e-05,
      "loss": 0.2092,
      "step": 3834
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.108992467043315e-05,
      "loss": 0.3113,
      "step": 3835
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.108757062146893e-05,
      "loss": 0.2555,
      "step": 3836
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.108521657250471e-05,
      "loss": 0.4124,
      "step": 3837
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.1082862523540496e-05,
      "loss": 0.452,
      "step": 3838
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.108050847457627e-05,
      "loss": 0.3926,
      "step": 3839
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.107815442561205e-05,
      "loss": 0.324,
      "step": 3840
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.107580037664784e-05,
      "loss": 0.2636,
      "step": 3841
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.107344632768362e-05,
      "loss": 0.375,
      "step": 3842
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1071092278719395e-05,
      "loss": 0.4922,
      "step": 3843
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1068738229755186e-05,
      "loss": 0.2856,
      "step": 3844
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.106638418079096e-05,
      "loss": 0.4623,
      "step": 3845
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.106403013182674e-05,
      "loss": 0.4265,
      "step": 3846
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1061676082862524e-05,
      "loss": 0.41,
      "step": 3847
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.105932203389831e-05,
      "loss": 0.3249,
      "step": 3848
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1056967984934085e-05,
      "loss": 0.4372,
      "step": 3849
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.105461393596987e-05,
      "loss": 0.193,
      "step": 3850
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.105225988700565e-05,
      "loss": 0.2949,
      "step": 3851
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.104990583804144e-05,
      "loss": 0.3255,
      "step": 3852
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1047551789077214e-05,
      "loss": 0.4105,
      "step": 3853
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1045197740113e-05,
      "loss": 0.3291,
      "step": 3854
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.104284369114878e-05,
      "loss": 0.2861,
      "step": 3855
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.104048964218456e-05,
      "loss": 0.3837,
      "step": 3856
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1038135593220336e-05,
      "loss": 0.4354,
      "step": 3857
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.103578154425613e-05,
      "loss": 0.2645,
      "step": 3858
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1033427495291904e-05,
      "loss": 0.241,
      "step": 3859
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.103107344632768e-05,
      "loss": 0.4367,
      "step": 3860
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.102871939736347e-05,
      "loss": 0.2625,
      "step": 3861
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.102636534839925e-05,
      "loss": 0.3016,
      "step": 3862
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.1024011299435026e-05,
      "loss": 0.3225,
      "step": 3863
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.102165725047081e-05,
      "loss": 0.3062,
      "step": 3864
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.1019303201506594e-05,
      "loss": 0.3372,
      "step": 3865
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.101694915254237e-05,
      "loss": 0.3093,
      "step": 3866
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.1014595103578155e-05,
      "loss": 0.2754,
      "step": 3867
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.101224105461394e-05,
      "loss": 0.3875,
      "step": 3868
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.100988700564972e-05,
      "loss": 0.3933,
      "step": 3869
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.10075329566855e-05,
      "loss": 0.2748,
      "step": 3870
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.1005178907721284e-05,
      "loss": 0.3746,
      "step": 3871
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.100282485875707e-05,
      "loss": 0.3024,
      "step": 3872
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.1000470809792845e-05,
      "loss": 0.3646,
      "step": 3873
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.099811676082862e-05,
      "loss": 0.2623,
      "step": 3874
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.099576271186441e-05,
      "loss": 0.3259,
      "step": 3875
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.099340866290019e-05,
      "loss": 0.2963,
      "step": 3876
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.099105461393597e-05,
      "loss": 0.3251,
      "step": 3877
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.098870056497176e-05,
      "loss": 0.326,
      "step": 3878
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.0986346516007535e-05,
      "loss": 0.3634,
      "step": 3879
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.098399246704331e-05,
      "loss": 0.2534,
      "step": 3880
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.0981638418079096e-05,
      "loss": 0.4424,
      "step": 3881
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.097928436911488e-05,
      "loss": 0.3983,
      "step": 3882
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.0976930320150664e-05,
      "loss": 0.3034,
      "step": 3883
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.097457627118644e-05,
      "loss": 0.2319,
      "step": 3884
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.0972222222222225e-05,
      "loss": 0.4273,
      "step": 3885
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.096986817325801e-05,
      "loss": 0.4201,
      "step": 3886
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0967514124293786e-05,
      "loss": 0.3083,
      "step": 3887
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.096516007532957e-05,
      "loss": 0.2564,
      "step": 3888
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0962806026365354e-05,
      "loss": 0.2283,
      "step": 3889
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.096045197740113e-05,
      "loss": 0.3893,
      "step": 3890
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.095809792843691e-05,
      "loss": 0.2499,
      "step": 3891
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.09557438794727e-05,
      "loss": 0.4107,
      "step": 3892
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0953389830508476e-05,
      "loss": 0.4139,
      "step": 3893
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.095103578154425e-05,
      "loss": 0.2816,
      "step": 3894
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0948681732580044e-05,
      "loss": 0.3475,
      "step": 3895
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.094632768361582e-05,
      "loss": 0.2481,
      "step": 3896
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.09439736346516e-05,
      "loss": 0.4145,
      "step": 3897
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.094161958568738e-05,
      "loss": 0.2912,
      "step": 3898
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0939265536723166e-05,
      "loss": 0.3049,
      "step": 3899
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.093691148775895e-05,
      "loss": 0.2344,
      "step": 3900
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.093455743879473e-05,
      "loss": 0.3539,
      "step": 3901
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.093220338983051e-05,
      "loss": 0.3548,
      "step": 3902
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0929849340866295e-05,
      "loss": 0.3616,
      "step": 3903
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.092749529190207e-05,
      "loss": 0.3833,
      "step": 3904
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.0925141242937856e-05,
      "loss": 0.5541,
      "step": 3905
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.092278719397364e-05,
      "loss": 0.37,
      "step": 3906
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.092043314500942e-05,
      "loss": 0.4103,
      "step": 3907
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0918079096045194e-05,
      "loss": 0.2338,
      "step": 3908
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0915725047080985e-05,
      "loss": 0.3992,
      "step": 3909
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.091337099811676e-05,
      "loss": 0.3082,
      "step": 3910
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.091101694915254e-05,
      "loss": 0.3155,
      "step": 3911
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.090866290018833e-05,
      "loss": 0.2585,
      "step": 3912
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.090630885122411e-05,
      "loss": 0.2913,
      "step": 3913
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.090395480225989e-05,
      "loss": 0.4282,
      "step": 3914
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.090160075329567e-05,
      "loss": 0.3979,
      "step": 3915
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.089924670433145e-05,
      "loss": 0.4054,
      "step": 3916
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0896892655367236e-05,
      "loss": 0.3468,
      "step": 3917
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.089453860640301e-05,
      "loss": 0.2972,
      "step": 3918
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.08921845574388e-05,
      "loss": 0.2338,
      "step": 3919
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.088983050847458e-05,
      "loss": 0.2222,
      "step": 3920
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.088747645951036e-05,
      "loss": 0.3032,
      "step": 3921
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.088512241054614e-05,
      "loss": 0.3055,
      "step": 3922
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0882768361581926e-05,
      "loss": 0.3839,
      "step": 3923
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.08804143126177e-05,
      "loss": 0.3487,
      "step": 3924
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.087806026365348e-05,
      "loss": 0.2537,
      "step": 3925
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.087570621468927e-05,
      "loss": 0.4674,
      "step": 3926
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.087335216572505e-05,
      "loss": 0.3391,
      "step": 3927
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.0870998116760825e-05,
      "loss": 0.2722,
      "step": 3928
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.0868644067796616e-05,
      "loss": 0.3276,
      "step": 3929
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.086629001883239e-05,
      "loss": 0.4798,
      "step": 3930
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.086393596986818e-05,
      "loss": 0.2786,
      "step": 3931
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.0861581920903954e-05,
      "loss": 0.2328,
      "step": 3932
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.085922787193974e-05,
      "loss": 0.2901,
      "step": 3933
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.085687382297552e-05,
      "loss": 0.4259,
      "step": 3934
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.08545197740113e-05,
      "loss": 0.3105,
      "step": 3935
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.085216572504708e-05,
      "loss": 0.4319,
      "step": 3936
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.084981167608287e-05,
      "loss": 0.3159,
      "step": 3937
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.0847457627118644e-05,
      "loss": 0.4941,
      "step": 3938
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.084510357815443e-05,
      "loss": 0.3292,
      "step": 3939
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.084274952919021e-05,
      "loss": 0.3927,
      "step": 3940
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.084039548022599e-05,
      "loss": 0.3256,
      "step": 3941
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.083804143126177e-05,
      "loss": 0.3369,
      "step": 3942
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.083568738229756e-05,
      "loss": 0.3514,
      "step": 3943
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.0833333333333334e-05,
      "loss": 0.219,
      "step": 3944
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.083097928436911e-05,
      "loss": 0.417,
      "step": 3945
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.08286252354049e-05,
      "loss": 0.4043,
      "step": 3946
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.082627118644068e-05,
      "loss": 0.3731,
      "step": 3947
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.082391713747646e-05,
      "loss": 0.2747,
      "step": 3948
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.082156308851224e-05,
      "loss": 0.3463,
      "step": 3949
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0819209039548024e-05,
      "loss": 0.3274,
      "step": 3950
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.081685499058381e-05,
      "loss": 0.327,
      "step": 3951
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0814500941619585e-05,
      "loss": 0.3824,
      "step": 3952
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.081214689265537e-05,
      "loss": 0.3412,
      "step": 3953
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.080979284369115e-05,
      "loss": 0.3981,
      "step": 3954
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.080743879472693e-05,
      "loss": 0.2758,
      "step": 3955
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0805084745762714e-05,
      "loss": 0.2184,
      "step": 3956
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.08027306967985e-05,
      "loss": 0.4266,
      "step": 3957
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0800376647834275e-05,
      "loss": 0.4013,
      "step": 3958
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.079802259887006e-05,
      "loss": 0.3451,
      "step": 3959
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.079566854990584e-05,
      "loss": 0.2536,
      "step": 3960
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.079331450094162e-05,
      "loss": 0.3675,
      "step": 3961
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0790960451977404e-05,
      "loss": 0.4952,
      "step": 3962
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.078860640301319e-05,
      "loss": 0.3285,
      "step": 3963
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0786252354048965e-05,
      "loss": 0.4946,
      "step": 3964
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.078389830508475e-05,
      "loss": 0.3073,
      "step": 3965
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0781544256120526e-05,
      "loss": 0.2401,
      "step": 3966
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.077919020715631e-05,
      "loss": 0.1595,
      "step": 3967
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0776836158192094e-05,
      "loss": 0.262,
      "step": 3968
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.077448210922787e-05,
      "loss": 0.2517,
      "step": 3969
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.0772128060263655e-05,
      "loss": 0.3074,
      "step": 3970
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.076977401129944e-05,
      "loss": 0.3043,
      "step": 3971
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0767419962335216e-05,
      "loss": 0.4418,
      "step": 3972
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0765065913371e-05,
      "loss": 0.3923,
      "step": 3973
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0762711864406784e-05,
      "loss": 0.4242,
      "step": 3974
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.076035781544256e-05,
      "loss": 0.3371,
      "step": 3975
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0758003766478345e-05,
      "loss": 0.2932,
      "step": 3976
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.075564971751413e-05,
      "loss": 0.3478,
      "step": 3977
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0753295668549906e-05,
      "loss": 0.4157,
      "step": 3978
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.075094161958569e-05,
      "loss": 0.4641,
      "step": 3979
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0748587570621474e-05,
      "loss": 0.2319,
      "step": 3980
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.074623352165725e-05,
      "loss": 0.259,
      "step": 3981
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0743879472693035e-05,
      "loss": 0.239,
      "step": 3982
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.074152542372881e-05,
      "loss": 0.3564,
      "step": 3983
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0739171374764596e-05,
      "loss": 0.2858,
      "step": 3984
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.073681732580038e-05,
      "loss": 0.2963,
      "step": 3985
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.073446327683616e-05,
      "loss": 0.3234,
      "step": 3986
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.073210922787194e-05,
      "loss": 0.3731,
      "step": 3987
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0729755178907725e-05,
      "loss": 0.2214,
      "step": 3988
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.07274011299435e-05,
      "loss": 0.2631,
      "step": 3989
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0725047080979286e-05,
      "loss": 0.3288,
      "step": 3990
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.072269303201507e-05,
      "loss": 0.33,
      "step": 3991
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.072033898305085e-05,
      "loss": 0.302,
      "step": 3992
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.071798493408663e-05,
      "loss": 0.2778,
      "step": 3993
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.0715630885122415e-05,
      "loss": 0.2978,
      "step": 3994
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.071327683615819e-05,
      "loss": 0.2444,
      "step": 3995
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.0710922787193976e-05,
      "loss": 0.4302,
      "step": 3996
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.070856873822976e-05,
      "loss": 0.3779,
      "step": 3997
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.070621468926554e-05,
      "loss": 0.4369,
      "step": 3998
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.070386064030132e-05,
      "loss": 0.5371,
      "step": 3999
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.07015065913371e-05,
      "loss": 0.3552,
      "step": 4000
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.069915254237288e-05,
      "loss": 0.3527,
      "step": 4001
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.0696798493408666e-05,
      "loss": 0.2474,
      "step": 4002
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.0694444444444444e-05,
      "loss": 0.4436,
      "step": 4003
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.069209039548023e-05,
      "loss": 0.4707,
      "step": 4004
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.068973634651601e-05,
      "loss": 0.3694,
      "step": 4005
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.068738229755179e-05,
      "loss": 0.4218,
      "step": 4006
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.068502824858757e-05,
      "loss": 0.2799,
      "step": 4007
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.0682674199623356e-05,
      "loss": 0.2431,
      "step": 4008
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.0680320150659134e-05,
      "loss": 0.4536,
      "step": 4009
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.067796610169492e-05,
      "loss": 0.3038,
      "step": 4010
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.06756120527307e-05,
      "loss": 0.4319,
      "step": 4011
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.067325800376648e-05,
      "loss": 0.3389,
      "step": 4012
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.067090395480226e-05,
      "loss": 0.5023,
      "step": 4013
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0668549905838046e-05,
      "loss": 0.333,
      "step": 4014
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0666195856873824e-05,
      "loss": 0.2637,
      "step": 4015
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.066384180790961e-05,
      "loss": 0.4157,
      "step": 4016
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.066148775894539e-05,
      "loss": 0.3155,
      "step": 4017
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.065913370998117e-05,
      "loss": 0.2457,
      "step": 4018
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.065677966101695e-05,
      "loss": 0.3484,
      "step": 4019
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.065442561205273e-05,
      "loss": 0.3363,
      "step": 4020
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0652071563088513e-05,
      "loss": 0.3313,
      "step": 4021
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.06497175141243e-05,
      "loss": 0.3192,
      "step": 4022
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0647363465160075e-05,
      "loss": 0.2936,
      "step": 4023
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.064500941619586e-05,
      "loss": 0.4256,
      "step": 4024
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.064265536723164e-05,
      "loss": 0.4565,
      "step": 4025
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.064030131826742e-05,
      "loss": 0.3231,
      "step": 4026
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0637947269303203e-05,
      "loss": 0.3051,
      "step": 4027
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.063559322033899e-05,
      "loss": 0.4167,
      "step": 4028
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0633239171374765e-05,
      "loss": 0.3367,
      "step": 4029
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.063088512241055e-05,
      "loss": 0.3292,
      "step": 4030
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.062853107344633e-05,
      "loss": 0.2795,
      "step": 4031
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.062617702448211e-05,
      "loss": 0.3199,
      "step": 4032
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.0623822975517893e-05,
      "loss": 0.3875,
      "step": 4033
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.062146892655368e-05,
      "loss": 0.1801,
      "step": 4034
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0619114877589455e-05,
      "loss": 0.3432,
      "step": 4035
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.061676082862524e-05,
      "loss": 0.4207,
      "step": 4036
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0614406779661016e-05,
      "loss": 0.3677,
      "step": 4037
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.06120527306968e-05,
      "loss": 0.2848,
      "step": 4038
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0609698681732583e-05,
      "loss": 0.5048,
      "step": 4039
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.060734463276836e-05,
      "loss": 0.3228,
      "step": 4040
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0604990583804145e-05,
      "loss": 0.4409,
      "step": 4041
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.060263653483993e-05,
      "loss": 0.3238,
      "step": 4042
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0600282485875706e-05,
      "loss": 0.2269,
      "step": 4043
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.059792843691149e-05,
      "loss": 0.3655,
      "step": 4044
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0595574387947273e-05,
      "loss": 0.3629,
      "step": 4045
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.059322033898305e-05,
      "loss": 0.373,
      "step": 4046
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0590866290018835e-05,
      "loss": 0.2742,
      "step": 4047
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.058851224105462e-05,
      "loss": 0.246,
      "step": 4048
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0586158192090396e-05,
      "loss": 0.3108,
      "step": 4049
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.058380414312618e-05,
      "loss": 0.2551,
      "step": 4050
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0581450094161963e-05,
      "loss": 0.3146,
      "step": 4051
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.057909604519774e-05,
      "loss": 0.2494,
      "step": 4052
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0576741996233525e-05,
      "loss": 0.3607,
      "step": 4053
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.05743879472693e-05,
      "loss": 0.3487,
      "step": 4054
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0572033898305086e-05,
      "loss": 0.2877,
      "step": 4055
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.056967984934087e-05,
      "loss": 0.3115,
      "step": 4056
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.056732580037665e-05,
      "loss": 0.4524,
      "step": 4057
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.056497175141243e-05,
      "loss": 0.2934,
      "step": 4058
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.0562617702448215e-05,
      "loss": 0.243,
      "step": 4059
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.056026365348399e-05,
      "loss": 0.3829,
      "step": 4060
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.0557909604519776e-05,
      "loss": 0.2745,
      "step": 4061
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.055555555555556e-05,
      "loss": 0.449,
      "step": 4062
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.055320150659134e-05,
      "loss": 0.3634,
      "step": 4063
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.055084745762712e-05,
      "loss": 0.3846,
      "step": 4064
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.0548493408662905e-05,
      "loss": 0.3212,
      "step": 4065
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.054613935969868e-05,
      "loss": 0.3974,
      "step": 4066
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.0543785310734466e-05,
      "loss": 0.2839,
      "step": 4067
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.054143126177025e-05,
      "loss": 0.2755,
      "step": 4068
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.053907721280603e-05,
      "loss": 0.4286,
      "step": 4069
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.053672316384181e-05,
      "loss": 0.3433,
      "step": 4070
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.053436911487759e-05,
      "loss": 0.2627,
      "step": 4071
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.053201506591337e-05,
      "loss": 0.3768,
      "step": 4072
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.0529661016949156e-05,
      "loss": 0.4877,
      "step": 4073
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.052730696798493e-05,
      "loss": 0.2945,
      "step": 4074
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.052495291902072e-05,
      "loss": 0.3174,
      "step": 4075
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.05225988700565e-05,
      "loss": 0.3846,
      "step": 4076
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.052024482109228e-05,
      "loss": 0.3584,
      "step": 4077
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.051789077212806e-05,
      "loss": 0.2786,
      "step": 4078
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.0515536723163846e-05,
      "loss": 0.3202,
      "step": 4079
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.051318267419962e-05,
      "loss": 0.3457,
      "step": 4080
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.051082862523541e-05,
      "loss": 0.3405,
      "step": 4081
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.050847457627119e-05,
      "loss": 0.3477,
      "step": 4082
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.050612052730697e-05,
      "loss": 0.2889,
      "step": 4083
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.050376647834275e-05,
      "loss": 0.3,
      "step": 4084
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.0501412429378536e-05,
      "loss": 0.182,
      "step": 4085
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.049905838041431e-05,
      "loss": 0.3993,
      "step": 4086
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.04967043314501e-05,
      "loss": 0.4038,
      "step": 4087
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.0494350282485874e-05,
      "loss": 0.485,
      "step": 4088
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.0491996233521665e-05,
      "loss": 0.2953,
      "step": 4089
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.048964218455744e-05,
      "loss": 0.3622,
      "step": 4090
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.048728813559322e-05,
      "loss": 0.461,
      "step": 4091
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.0484934086629e-05,
      "loss": 0.3463,
      "step": 4092
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.048258003766479e-05,
      "loss": 0.2809,
      "step": 4093
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.0480225988700564e-05,
      "loss": 0.2423,
      "step": 4094
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.047787193973635e-05,
      "loss": 0.3256,
      "step": 4095
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.047551789077213e-05,
      "loss": 0.423,
      "step": 4096
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.047316384180791e-05,
      "loss": 0.2598,
      "step": 4097
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.047080979284369e-05,
      "loss": 0.4071,
      "step": 4098
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.046845574387948e-05,
      "loss": 0.3208,
      "step": 4099
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0466101694915254e-05,
      "loss": 0.3755,
      "step": 4100
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.046374764595104e-05,
      "loss": 0.3786,
      "step": 4101
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.046139359698682e-05,
      "loss": 0.3978,
      "step": 4102
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.04590395480226e-05,
      "loss": 0.4359,
      "step": 4103
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.045668549905838e-05,
      "loss": 0.3159,
      "step": 4104
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.045433145009416e-05,
      "loss": 0.2923,
      "step": 4105
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.045197740112995e-05,
      "loss": 0.3218,
      "step": 4106
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.044962335216573e-05,
      "loss": 0.3317,
      "step": 4107
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0447269303201505e-05,
      "loss": 0.396,
      "step": 4108
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0444915254237296e-05,
      "loss": 0.3119,
      "step": 4109
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.044256120527307e-05,
      "loss": 0.2331,
      "step": 4110
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.044020715630885e-05,
      "loss": 0.3991,
      "step": 4111
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0437853107344634e-05,
      "loss": 0.3683,
      "step": 4112
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.043549905838042e-05,
      "loss": 0.2445,
      "step": 4113
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0433145009416195e-05,
      "loss": 0.3701,
      "step": 4114
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.043079096045198e-05,
      "loss": 0.2558,
      "step": 4115
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.042843691148776e-05,
      "loss": 0.2462,
      "step": 4116
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.042608286252354e-05,
      "loss": 0.2542,
      "step": 4117
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0423728813559324e-05,
      "loss": 0.2379,
      "step": 4118
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.042137476459511e-05,
      "loss": 0.343,
      "step": 4119
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.041902071563089e-05,
      "loss": 0.2379,
      "step": 4120
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.041666666666667e-05,
      "loss": 0.3339,
      "step": 4121
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0414312617702446e-05,
      "loss": 0.2884,
      "step": 4122
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.041195856873824e-05,
      "loss": 0.2946,
      "step": 4123
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0409604519774014e-05,
      "loss": 0.3045,
      "step": 4124
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.040725047080979e-05,
      "loss": 0.3311,
      "step": 4125
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.040489642184558e-05,
      "loss": 0.2595,
      "step": 4126
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.040254237288136e-05,
      "loss": 0.3373,
      "step": 4127
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0400188323917136e-05,
      "loss": 0.2607,
      "step": 4128
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.039783427495292e-05,
      "loss": 0.2906,
      "step": 4129
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0395480225988704e-05,
      "loss": 0.3553,
      "step": 4130
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.039312617702448e-05,
      "loss": 0.3675,
      "step": 4131
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0390772128060265e-05,
      "loss": 0.3345,
      "step": 4132
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.038841807909605e-05,
      "loss": 0.3459,
      "step": 4133
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0386064030131826e-05,
      "loss": 0.3159,
      "step": 4134
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.038370998116761e-05,
      "loss": 0.3895,
      "step": 4135
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0381355932203394e-05,
      "loss": 0.2382,
      "step": 4136
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.037900188323918e-05,
      "loss": 0.2767,
      "step": 4137
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0376647834274955e-05,
      "loss": 0.2279,
      "step": 4138
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.037429378531073e-05,
      "loss": 0.2259,
      "step": 4139
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.037193973634652e-05,
      "loss": 0.4026,
      "step": 4140
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.03695856873823e-05,
      "loss": 0.367,
      "step": 4141
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.036723163841808e-05,
      "loss": 0.3445,
      "step": 4142
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.036487758945387e-05,
      "loss": 0.2946,
      "step": 4143
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0362523540489645e-05,
      "loss": 0.2931,
      "step": 4144
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.036016949152542e-05,
      "loss": 0.41,
      "step": 4145
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0357815442561206e-05,
      "loss": 0.4136,
      "step": 4146
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.035546139359699e-05,
      "loss": 0.3578,
      "step": 4147
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.035310734463277e-05,
      "loss": 0.4197,
      "step": 4148
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.035075329566855e-05,
      "loss": 0.3107,
      "step": 4149
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0348399246704335e-05,
      "loss": 0.3232,
      "step": 4150
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.034604519774011e-05,
      "loss": 0.4379,
      "step": 4151
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0343691148775896e-05,
      "loss": 0.3631,
      "step": 4152
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.034133709981168e-05,
      "loss": 0.2728,
      "step": 4153
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0338983050847464e-05,
      "loss": 0.3852,
      "step": 4154
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.033662900188324e-05,
      "loss": 0.3041,
      "step": 4155
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.033427495291902e-05,
      "loss": 0.4163,
      "step": 4156
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.033192090395481e-05,
      "loss": 0.2762,
      "step": 4157
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0329566854990586e-05,
      "loss": 0.1958,
      "step": 4158
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.032721280602636e-05,
      "loss": 0.316,
      "step": 4159
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.0324858757062154e-05,
      "loss": 0.2706,
      "step": 4160
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.032250470809793e-05,
      "loss": 0.3394,
      "step": 4161
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.032015065913371e-05,
      "loss": 0.2022,
      "step": 4162
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.031779661016949e-05,
      "loss": 0.2266,
      "step": 4163
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.0315442561205276e-05,
      "loss": 0.2958,
      "step": 4164
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.031308851224105e-05,
      "loss": 0.5119,
      "step": 4165
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.031073446327684e-05,
      "loss": 0.2262,
      "step": 4166
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.030838041431262e-05,
      "loss": 0.1969,
      "step": 4167
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.0306026365348405e-05,
      "loss": 0.2742,
      "step": 4168
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.030367231638418e-05,
      "loss": 0.3414,
      "step": 4169
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.0301318267419966e-05,
      "loss": 0.3796,
      "step": 4170
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.029896421845575e-05,
      "loss": 0.2796,
      "step": 4171
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.029661016949153e-05,
      "loss": 0.4311,
      "step": 4172
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.0294256120527304e-05,
      "loss": 0.3744,
      "step": 4173
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.0291902071563095e-05,
      "loss": 0.4297,
      "step": 4174
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.028954802259887e-05,
      "loss": 0.3757,
      "step": 4175
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.028719397363465e-05,
      "loss": 0.2206,
      "step": 4176
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.028483992467044e-05,
      "loss": 0.349,
      "step": 4177
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.028248587570622e-05,
      "loss": 0.1441,
      "step": 4178
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.0280131826741994e-05,
      "loss": 0.3211,
      "step": 4179
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.027777777777778e-05,
      "loss": 0.2869,
      "step": 4180
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.027542372881356e-05,
      "loss": 0.2487,
      "step": 4181
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.027306967984934e-05,
      "loss": 0.273,
      "step": 4182
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.027071563088512e-05,
      "loss": 0.241,
      "step": 4183
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.026836158192091e-05,
      "loss": 0.2336,
      "step": 4184
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.026600753295669e-05,
      "loss": 0.2987,
      "step": 4185
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.026365348399247e-05,
      "loss": 0.3111,
      "step": 4186
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.026129943502825e-05,
      "loss": 0.2296,
      "step": 4187
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.0258945386064036e-05,
      "loss": 0.292,
      "step": 4188
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.025659133709981e-05,
      "loss": 0.3515,
      "step": 4189
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.025423728813559e-05,
      "loss": 0.4625,
      "step": 4190
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.025188323917138e-05,
      "loss": 0.2747,
      "step": 4191
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.024952919020716e-05,
      "loss": 0.4395,
      "step": 4192
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.0247175141242935e-05,
      "loss": 0.3793,
      "step": 4193
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.0244821092278726e-05,
      "loss": 0.4036,
      "step": 4194
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.02424670433145e-05,
      "loss": 0.2878,
      "step": 4195
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.024011299435028e-05,
      "loss": 0.4269,
      "step": 4196
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.0237758945386064e-05,
      "loss": 0.2707,
      "step": 4197
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.023540489642185e-05,
      "loss": 0.2906,
      "step": 4198
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.023305084745763e-05,
      "loss": 0.2977,
      "step": 4199
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.023069679849341e-05,
      "loss": 0.4789,
      "step": 4200
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.022834274952919e-05,
      "loss": 0.3775,
      "step": 4201
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.022598870056498e-05,
      "loss": 0.2239,
      "step": 4202
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.0223634651600754e-05,
      "loss": 0.2504,
      "step": 4203
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.022128060263654e-05,
      "loss": 0.2584,
      "step": 4204
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.021892655367232e-05,
      "loss": 0.2977,
      "step": 4205
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.02165725047081e-05,
      "loss": 0.2211,
      "step": 4206
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0214218455743876e-05,
      "loss": 0.3079,
      "step": 4207
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.021186440677967e-05,
      "loss": 0.3754,
      "step": 4208
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0209510357815444e-05,
      "loss": 0.4358,
      "step": 4209
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.020715630885122e-05,
      "loss": 0.4616,
      "step": 4210
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.020480225988701e-05,
      "loss": 0.3434,
      "step": 4211
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.020244821092279e-05,
      "loss": 0.2921,
      "step": 4212
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0200094161958566e-05,
      "loss": 0.4058,
      "step": 4213
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.019774011299435e-05,
      "loss": 0.4353,
      "step": 4214
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0195386064030134e-05,
      "loss": 0.2333,
      "step": 4215
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.019303201506592e-05,
      "loss": 0.3496,
      "step": 4216
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0190677966101695e-05,
      "loss": 0.2516,
      "step": 4217
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.018832391713748e-05,
      "loss": 0.285,
      "step": 4218
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.018596986817326e-05,
      "loss": 0.4167,
      "step": 4219
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.018361581920904e-05,
      "loss": 0.3132,
      "step": 4220
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0181261770244824e-05,
      "loss": 0.371,
      "step": 4221
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.017890772128061e-05,
      "loss": 0.2976,
      "step": 4222
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.0176553672316385e-05,
      "loss": 0.3813,
      "step": 4223
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.017419962335216e-05,
      "loss": 0.4538,
      "step": 4224
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.017184557438795e-05,
      "loss": 0.2605,
      "step": 4225
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.016949152542373e-05,
      "loss": 0.2254,
      "step": 4226
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.016713747645951e-05,
      "loss": 0.5101,
      "step": 4227
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.01647834274953e-05,
      "loss": 0.3143,
      "step": 4228
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0162429378531075e-05,
      "loss": 0.3559,
      "step": 4229
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.016007532956685e-05,
      "loss": 0.3853,
      "step": 4230
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0157721280602636e-05,
      "loss": 0.2722,
      "step": 4231
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.015536723163842e-05,
      "loss": 0.2807,
      "step": 4232
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0153013182674204e-05,
      "loss": 0.2709,
      "step": 4233
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.015065913370998e-05,
      "loss": 0.2954,
      "step": 4234
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0148305084745765e-05,
      "loss": 0.3649,
      "step": 4235
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.014595103578155e-05,
      "loss": 0.3102,
      "step": 4236
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0143596986817326e-05,
      "loss": 0.4962,
      "step": 4237
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.014124293785311e-05,
      "loss": 0.3302,
      "step": 4238
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0138888888888894e-05,
      "loss": 0.2986,
      "step": 4239
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.013653483992467e-05,
      "loss": 0.2826,
      "step": 4240
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.013418079096045e-05,
      "loss": 0.3632,
      "step": 4241
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.013182674199624e-05,
      "loss": 0.3283,
      "step": 4242
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0129472693032016e-05,
      "loss": 0.4383,
      "step": 4243
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.012711864406779e-05,
      "loss": 0.2267,
      "step": 4244
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0124764595103584e-05,
      "loss": 0.2532,
      "step": 4245
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.012241054613936e-05,
      "loss": 0.2331,
      "step": 4246
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.0120056497175145e-05,
      "loss": 0.2804,
      "step": 4247
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.011770244821092e-05,
      "loss": 0.2507,
      "step": 4248
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.0115348399246706e-05,
      "loss": 0.3725,
      "step": 4249
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.011299435028249e-05,
      "loss": 0.2826,
      "step": 4250
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.011064030131827e-05,
      "loss": 0.1912,
      "step": 4251
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.010828625235405e-05,
      "loss": 0.2806,
      "step": 4252
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.0105932203389835e-05,
      "loss": 0.2734,
      "step": 4253
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.010357815442561e-05,
      "loss": 0.3386,
      "step": 4254
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.0101224105461396e-05,
      "loss": 0.4069,
      "step": 4255
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.009887005649718e-05,
      "loss": 0.4076,
      "step": 4256
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.009651600753296e-05,
      "loss": 0.3156,
      "step": 4257
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.009416195856874e-05,
      "loss": 0.1351,
      "step": 4258
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8527589346019441,
      "eval_loss": 0.35805782675743103,
      "eval_runtime": 22.4157,
      "eval_samples_per_second": 1519.155,
      "eval_steps_per_second": 23.778,
      "step": 4258
    }
  ],
  "max_steps": 21290,
  "num_train_epochs": 10,
  "total_flos": 2752615365760932.0,
  "trial_name": null,
  "trial_params": null
}