{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 9.992769342010122,
  "global_step": 6910,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.894356005788712e-07,
      "loss": 5.3281,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.788712011577424e-07,
      "loss": 5.5625,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.683068017366137e-07,
      "loss": 5.5938,
      "step": 3
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1577424023154849e-06,
      "loss": 5.625,
      "step": 4
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.447178002894356e-06,
      "loss": 5.7031,
      "step": 5
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7366136034732275e-06,
      "loss": 5.7812,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.0260492040520984e-06,
      "loss": 6.6719,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3154848046309697e-06,
      "loss": 6.125,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.604920405209841e-06,
      "loss": 5.3438,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.894356005788712e-06,
      "loss": 5.0938,
      "step": 10
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.1837916063675832e-06,
      "loss": 5.0625,
      "step": 11
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.473227206946455e-06,
      "loss": 5.1875,
      "step": 12
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.762662807525326e-06,
      "loss": 5.7656,
      "step": 13
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.052098408104197e-06,
      "loss": 5.5156,
      "step": 14
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.341534008683068e-06,
      "loss": 5.1719,
      "step": 15
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.630969609261939e-06,
      "loss": 4.6719,
      "step": 16
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.920405209840811e-06,
      "loss": 4.2969,
      "step": 17
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.209840810419682e-06,
      "loss": 4.3594,
      "step": 18
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.499276410998553e-06,
      "loss": 4.4062,
      "step": 19
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.788712011577424e-06,
      "loss": 3.6875,
      "step": 20
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.078147612156296e-06,
      "loss": 3.6172,
      "step": 21
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.3675832127351665e-06,
      "loss": 3.4297,
      "step": 22
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.657018813314038e-06,
      "loss": 2.3203,
      "step": 23
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.94645441389291e-06,
      "loss": 2.5,
      "step": 24
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.2358900144717804e-06,
      "loss": 3.5859,
      "step": 25
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.525325615050652e-06,
      "loss": 2.9141,
      "step": 26
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.814761215629521e-06,
      "loss": 2.2188,
      "step": 27
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.104196816208394e-06,
      "loss": 3.4844,
      "step": 28
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.393632416787266e-06,
      "loss": 2.2109,
      "step": 29
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.683068017366136e-06,
      "loss": 1.9023,
      "step": 30
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.972503617945008e-06,
      "loss": 2.2031,
      "step": 31
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.261939218523879e-06,
      "loss": 2.3516,
      "step": 32
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.55137481910275e-06,
      "loss": 2.8203,
      "step": 33
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.840810419681621e-06,
      "loss": 1.7383,
      "step": 34
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0130246020260492e-05,
      "loss": 2.3672,
      "step": 35
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0419681620839364e-05,
      "loss": 3.1172,
      "step": 36
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0709117221418236e-05,
      "loss": 1.5898,
      "step": 37
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0998552821997107e-05,
      "loss": 2.5781,
      "step": 38
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.1287988422575977e-05,
      "loss": 1.1875,
      "step": 39
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.1577424023154848e-05,
      "loss": 0.9941,
      "step": 40
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.186685962373372e-05,
      "loss": 1.5625,
      "step": 41
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2156295224312592e-05,
      "loss": 2.1562,
      "step": 42
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2445730824891462e-05,
      "loss": 1.4414,
      "step": 43
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2735166425470333e-05,
      "loss": 0.6738,
      "step": 44
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.3024602026049203e-05,
      "loss": 0.3584,
      "step": 45
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.3314037626628076e-05,
      "loss": 0.3916,
      "step": 46
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.3603473227206948e-05,
      "loss": 1.0889,
      "step": 47
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.389290882778582e-05,
      "loss": 0.21,
      "step": 48
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4182344428364689e-05,
      "loss": 1.3892,
      "step": 49
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4471780028943561e-05,
      "loss": 0.5454,
      "step": 50
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4761215629522431e-05,
      "loss": 1.0078,
      "step": 51
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.5050651230101304e-05,
      "loss": 1.1934,
      "step": 52
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.5340086830680174e-05,
      "loss": 0.1655,
      "step": 53
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.5629522431259043e-05,
      "loss": 0.8237,
      "step": 54
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.5918958031837915e-05,
      "loss": 0.1689,
      "step": 55
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.6208393632416787e-05,
      "loss": 0.9995,
      "step": 56
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.649782923299566e-05,
      "loss": 0.8347,
      "step": 57
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.678726483357453e-05,
      "loss": 1.3672,
      "step": 58
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.70767004341534e-05,
      "loss": 1.1719,
      "step": 59
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.7366136034732272e-05,
      "loss": 1.0591,
      "step": 60
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.7655571635311145e-05,
      "loss": 0.5669,
      "step": 61
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.7945007235890017e-05,
      "loss": 0.8821,
      "step": 62
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.823444283646889e-05,
      "loss": 1.2715,
      "step": 63
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.8523878437047758e-05,
      "loss": 1.3438,
      "step": 64
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.881331403762663e-05,
      "loss": 0.7327,
      "step": 65
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.91027496382055e-05,
      "loss": 0.0759,
      "step": 66
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.939218523878437e-05,
      "loss": 0.468,
      "step": 67
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9681620839363243e-05,
      "loss": 0.5835,
      "step": 68
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9971056439942115e-05,
      "loss": 1.5748,
      "step": 69
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.0260492040520984e-05,
      "loss": 0.92,
      "step": 70
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.0549927641099856e-05,
      "loss": 1.1442,
      "step": 71
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.0839363241678728e-05,
      "loss": 0.0486,
      "step": 72
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.11287988422576e-05,
      "loss": 1.2773,
      "step": 73
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.1418234442836473e-05,
      "loss": 0.4969,
      "step": 74
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.170767004341534e-05,
      "loss": 0.0969,
      "step": 75
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.1997105643994213e-05,
      "loss": 0.1355,
      "step": 76
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.2286541244573082e-05,
      "loss": 0.1102,
      "step": 77
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.2575976845151954e-05,
      "loss": 0.4373,
      "step": 78
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.2865412445730827e-05,
      "loss": 0.6619,
      "step": 79
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.3154848046309695e-05,
      "loss": 0.5676,
      "step": 80
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.3444283646888568e-05,
      "loss": 0.6392,
      "step": 81
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.373371924746744e-05,
      "loss": 0.4504,
      "step": 82
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.4023154848046312e-05,
      "loss": 0.0518,
      "step": 83
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.4312590448625184e-05,
      "loss": 0.1421,
      "step": 84
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.4602026049204053e-05,
      "loss": 0.0485,
      "step": 85
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.4891461649782925e-05,
      "loss": 0.7021,
      "step": 86
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.5180897250361797e-05,
      "loss": 0.7388,
      "step": 87
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.5470332850940666e-05,
      "loss": 0.0339,
      "step": 88
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.5759768451519538e-05,
      "loss": 0.2732,
      "step": 89
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.6049204052098407e-05,
      "loss": 0.2972,
      "step": 90
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.6338639652677282e-05,
      "loss": 0.3364,
      "step": 91
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.662807525325615e-05,
      "loss": 1.4609,
      "step": 92
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.691751085383502e-05,
      "loss": 0.0658,
      "step": 93
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.7206946454413896e-05,
      "loss": 0.3535,
      "step": 94
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.7496382054992764e-05,
      "loss": 1.1055,
      "step": 95
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.778581765557164e-05,
      "loss": 1.4414,
      "step": 96
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.807525325615051e-05,
      "loss": 1.875,
      "step": 97
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.8364688856729377e-05,
      "loss": 0.8364,
      "step": 98
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.865412445730825e-05,
      "loss": 0.4307,
      "step": 99
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.8943560057887122e-05,
      "loss": 0.7852,
      "step": 100
    },
    {
      "epoch": 0.15,
      "learning_rate": 2.9232995658465994e-05,
      "loss": 0.2856,
      "step": 101
    },
    {
      "epoch": 0.15,
      "learning_rate": 2.9522431259044863e-05,
      "loss": 0.4302,
      "step": 102
    },
    {
      "epoch": 0.15,
      "learning_rate": 2.981186685962373e-05,
      "loss": 1.0903,
      "step": 103
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.0101302460202607e-05,
      "loss": 1.3477,
      "step": 104
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.0390738060781476e-05,
      "loss": 0.5117,
      "step": 105
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.068017366136035e-05,
      "loss": 0.5107,
      "step": 106
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.096960926193922e-05,
      "loss": 0.0619,
      "step": 107
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.1259044862518086e-05,
      "loss": 0.7822,
      "step": 108
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.1548480463096964e-05,
      "loss": 1.0176,
      "step": 109
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.183791606367583e-05,
      "loss": 0.0381,
      "step": 110
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.212735166425471e-05,
      "loss": 0.0771,
      "step": 111
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.2416787264833574e-05,
      "loss": 0.6042,
      "step": 112
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.2706222865412446e-05,
      "loss": 0.0258,
      "step": 113
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.299565846599132e-05,
      "loss": 0.6804,
      "step": 114
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.328509406657019e-05,
      "loss": 0.0977,
      "step": 115
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.357452966714906e-05,
      "loss": 0.0099,
      "step": 116
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.3863965267727935e-05,
      "loss": 0.4032,
      "step": 117
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.41534008683068e-05,
      "loss": 0.6332,
      "step": 118
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.444283646888567e-05,
      "loss": 0.8965,
      "step": 119
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.4732272069464545e-05,
      "loss": 0.4485,
      "step": 120
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.502170767004342e-05,
      "loss": 0.0394,
      "step": 121
    },
    {
      "epoch": 0.18,
      "learning_rate": 3.531114327062229e-05,
      "loss": 0.0372,
      "step": 122
    },
    {
      "epoch": 0.18,
      "learning_rate": 3.5600578871201154e-05,
      "loss": 0.3734,
      "step": 123
    },
    {
      "epoch": 0.18,
      "learning_rate": 3.5890014471780033e-05,
      "loss": 0.8356,
      "step": 124
    },
    {
      "epoch": 0.18,
      "learning_rate": 3.61794500723589e-05,
      "loss": 0.0643,
      "step": 125
    },
    {
      "epoch": 0.18,
      "learning_rate": 3.646888567293778e-05,
      "loss": 0.0152,
      "step": 126
    },
    {
      "epoch": 0.18,
      "learning_rate": 3.675832127351664e-05,
      "loss": 0.2809,
      "step": 127
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.7047756874095515e-05,
      "loss": 0.2233,
      "step": 128
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.733719247467439e-05,
      "loss": 0.5156,
      "step": 129
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.762662807525326e-05,
      "loss": 0.6548,
      "step": 130
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.791606367583213e-05,
      "loss": 1.1465,
      "step": 131
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.8205499276411e-05,
      "loss": 0.2869,
      "step": 132
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.8494934876989876e-05,
      "loss": 0.045,
      "step": 133
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.878437047756874e-05,
      "loss": 0.7078,
      "step": 134
    },
    {
      "epoch": 0.2,
      "learning_rate": 3.9073806078147614e-05,
      "loss": 0.0347,
      "step": 135
    },
    {
      "epoch": 0.2,
      "learning_rate": 3.9363241678726486e-05,
      "loss": 0.0981,
      "step": 136
    },
    {
      "epoch": 0.2,
      "learning_rate": 3.965267727930536e-05,
      "loss": 0.2703,
      "step": 137
    },
    {
      "epoch": 0.2,
      "learning_rate": 3.994211287988423e-05,
      "loss": 0.2527,
      "step": 138
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.02315484804631e-05,
      "loss": 0.0697,
      "step": 139
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.052098408104197e-05,
      "loss": 0.0642,
      "step": 140
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.081041968162084e-05,
      "loss": 0.7895,
      "step": 141
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.109985528219971e-05,
      "loss": 0.4514,
      "step": 142
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.1389290882778584e-05,
      "loss": 0.8613,
      "step": 143
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.1678726483357456e-05,
      "loss": 0.467,
      "step": 144
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.196816208393632e-05,
      "loss": 0.0636,
      "step": 145
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.22575976845152e-05,
      "loss": 0.5093,
      "step": 146
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.2547033285094066e-05,
      "loss": 0.8398,
      "step": 147
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.2836468885672945e-05,
      "loss": 0.262,
      "step": 148
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.312590448625181e-05,
      "loss": 0.7925,
      "step": 149
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.341534008683068e-05,
      "loss": 0.0974,
      "step": 150
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.3704775687409555e-05,
      "loss": 0.2759,
      "step": 151
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.399421128798843e-05,
      "loss": 0.3937,
      "step": 152
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.42836468885673e-05,
      "loss": 0.0422,
      "step": 153
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4573082489146165e-05,
      "loss": 0.8379,
      "step": 154
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.486251808972504e-05,
      "loss": 0.4929,
      "step": 155
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.515195369030391e-05,
      "loss": 0.0918,
      "step": 156
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.544138929088278e-05,
      "loss": 0.0916,
      "step": 157
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.573082489146165e-05,
      "loss": 0.0862,
      "step": 158
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.6020260492040525e-05,
      "loss": 0.1279,
      "step": 159
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.630969609261939e-05,
      "loss": 0.9927,
      "step": 160
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.659913169319827e-05,
      "loss": 0.3762,
      "step": 161
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.6888567293777135e-05,
      "loss": 0.3406,
      "step": 162
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.717800289435601e-05,
      "loss": 0.926,
      "step": 163
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.746743849493488e-05,
      "loss": 0.0336,
      "step": 164
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.775687409551375e-05,
      "loss": 0.9177,
      "step": 165
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8046309696092624e-05,
      "loss": 0.1737,
      "step": 166
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.833574529667149e-05,
      "loss": 1.1484,
      "step": 167
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.862518089725037e-05,
      "loss": 0.7409,
      "step": 168
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8914616497829233e-05,
      "loss": 0.9434,
      "step": 169
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.9204052098408106e-05,
      "loss": 0.9844,
      "step": 170
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.949348769898698e-05,
      "loss": 0.0173,
      "step": 171
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.978292329956585e-05,
      "loss": 0.5267,
      "step": 172
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.007235890014472e-05,
      "loss": 0.3511,
      "step": 173
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.0361794500723594e-05,
      "loss": 0.113,
      "step": 174
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.065123010130246e-05,
      "loss": 1.1113,
      "step": 175
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.094066570188133e-05,
      "loss": 0.9424,
      "step": 176
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.123010130246021e-05,
      "loss": 0.2178,
      "step": 177
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.1519536903039076e-05,
      "loss": 0.0814,
      "step": 178
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.180897250361795e-05,
      "loss": 0.104,
      "step": 179
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.2098408104196814e-05,
      "loss": 0.585,
      "step": 180
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.2387843704775686e-05,
      "loss": 0.6309,
      "step": 181
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.2677279305354565e-05,
      "loss": 0.1016,
      "step": 182
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.296671490593344e-05,
      "loss": 0.0828,
      "step": 183
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.32561505065123e-05,
      "loss": 0.0729,
      "step": 184
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.3545586107091175e-05,
      "loss": 0.0723,
      "step": 185
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.383502170767004e-05,
      "loss": 0.655,
      "step": 186
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.412445730824892e-05,
      "loss": 0.0266,
      "step": 187
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.441389290882779e-05,
      "loss": 0.3758,
      "step": 188
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.4703328509406656e-05,
      "loss": 0.8789,
      "step": 189
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.499276410998553e-05,
      "loss": 0.9509,
      "step": 190
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.5282199710564394e-05,
      "loss": 0.4947,
      "step": 191
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.557163531114328e-05,
      "loss": 0.4727,
      "step": 192
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.5861070911722145e-05,
      "loss": 0.4902,
      "step": 193
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.615050651230102e-05,
      "loss": 0.1271,
      "step": 194
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.643994211287988e-05,
      "loss": 0.5073,
      "step": 195
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.6729377713458755e-05,
      "loss": 0.6868,
      "step": 196
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.7018813314037634e-05,
      "loss": 0.0614,
      "step": 197
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.73082489146165e-05,
      "loss": 0.0624,
      "step": 198
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.759768451519537e-05,
      "loss": 0.0533,
      "step": 199
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.7887120115774244e-05,
      "loss": 0.6312,
      "step": 200
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.817655571635311e-05,
      "loss": 0.6699,
      "step": 201
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.846599131693199e-05,
      "loss": 1.2324,
      "step": 202
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.875542691751086e-05,
      "loss": 0.1017,
      "step": 203
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.9044862518089725e-05,
      "loss": 0.0933,
      "step": 204
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.93342981186686e-05,
      "loss": 0.0483,
      "step": 205
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.962373371924746e-05,
      "loss": 0.0848,
      "step": 206
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.991316931982635e-05,
      "loss": 0.5275,
      "step": 207
    },
    {
      "epoch": 0.3,
      "learning_rate": 6.0202604920405214e-05,
      "loss": 0.7949,
      "step": 208
    },
    {
      "epoch": 0.3,
      "learning_rate": 6.0492040520984086e-05,
      "loss": 1.2207,
      "step": 209
    },
    {
      "epoch": 0.3,
      "learning_rate": 6.078147612156295e-05,
      "loss": 0.582,
      "step": 210
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.107091172214182e-05,
      "loss": 0.6777,
      "step": 211
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.13603473227207e-05,
      "loss": 0.4133,
      "step": 212
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.164978292329957e-05,
      "loss": 0.1567,
      "step": 213
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.193921852387844e-05,
      "loss": 0.4307,
      "step": 214
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.22286541244573e-05,
      "loss": 0.4292,
      "step": 215
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.251808972503617e-05,
      "loss": 0.0686,
      "step": 216
    },
    {
      "epoch": 0.31,
      "learning_rate": 6.280752532561506e-05,
      "loss": 0.3237,
      "step": 217
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.309696092619393e-05,
      "loss": 0.4565,
      "step": 218
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.33863965267728e-05,
      "loss": 0.0735,
      "step": 219
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.367583212735166e-05,
      "loss": 0.9447,
      "step": 220
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.396526772793054e-05,
      "loss": 1.2578,
      "step": 221
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.425470332850942e-05,
      "loss": 0.0218,
      "step": 222
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.454413892908828e-05,
      "loss": 0.4785,
      "step": 223
    },
    {
      "epoch": 0.32,
      "learning_rate": 6.483357452966715e-05,
      "loss": 0.6276,
      "step": 224
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.512301013024601e-05,
      "loss": 0.0713,
      "step": 225
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.541244573082489e-05,
      "loss": 0.0536,
      "step": 226
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.570188133140377e-05,
      "loss": 0.1152,
      "step": 227
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.599131693198264e-05,
      "loss": 0.7061,
      "step": 228
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.62807525325615e-05,
      "loss": 0.0188,
      "step": 229
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.657018813314038e-05,
      "loss": 0.049,
      "step": 230
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.685962373371925e-05,
      "loss": 0.0456,
      "step": 231
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.714905933429813e-05,
      "loss": 0.0203,
      "step": 232
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.743849493487699e-05,
      "loss": 0.1289,
      "step": 233
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.772793053545587e-05,
      "loss": 1.2441,
      "step": 234
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.801736613603474e-05,
      "loss": 0.6691,
      "step": 235
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.83068017366136e-05,
      "loss": 0.6297,
      "step": 236
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.859623733719248e-05,
      "loss": 0.7148,
      "step": 237
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.888567293777135e-05,
      "loss": 0.0471,
      "step": 238
    },
    {
      "epoch": 0.35,
      "learning_rate": 6.917510853835022e-05,
      "loss": 0.026,
      "step": 239
    },
    {
      "epoch": 0.35,
      "learning_rate": 6.946454413892909e-05,
      "loss": 0.3683,
      "step": 240
    },
    {
      "epoch": 0.35,
      "learning_rate": 6.975397973950795e-05,
      "loss": 0.0953,
      "step": 241
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.004341534008683e-05,
      "loss": 0.4937,
      "step": 242
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.033285094066571e-05,
      "loss": 0.0798,
      "step": 243
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.062228654124458e-05,
      "loss": 0.042,
      "step": 244
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.091172214182344e-05,
      "loss": 0.0323,
      "step": 245
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.120115774240231e-05,
      "loss": 0.7539,
      "step": 246
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.149059334298119e-05,
      "loss": 0.0509,
      "step": 247
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.178002894356007e-05,
      "loss": 0.2661,
      "step": 248
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.206946454413893e-05,
      "loss": 0.0911,
      "step": 249
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.23589001447178e-05,
      "loss": 0.0252,
      "step": 250
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.264833574529666e-05,
      "loss": 0.8215,
      "step": 251
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.293777134587556e-05,
      "loss": 0.6487,
      "step": 252
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.322720694645442e-05,
      "loss": 0.2935,
      "step": 253
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.351664254703329e-05,
      "loss": 0.083,
      "step": 254
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.380607814761215e-05,
      "loss": 1.0977,
      "step": 255
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.409551374819103e-05,
      "loss": 0.0552,
      "step": 256
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.438494934876991e-05,
      "loss": 0.0669,
      "step": 257
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.467438494934877e-05,
      "loss": 1.0615,
      "step": 258
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.496382054992764e-05,
      "loss": 0.209,
      "step": 259
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.525325615050652e-05,
      "loss": 0.5072,
      "step": 260
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.55426917510854e-05,
      "loss": 0.5352,
      "step": 261
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.583212735166426e-05,
      "loss": 0.0402,
      "step": 262
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.612156295224313e-05,
      "loss": 0.2839,
      "step": 263
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.6410998552822e-05,
      "loss": 0.1694,
      "step": 264
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.670043415340087e-05,
      "loss": 0.7488,
      "step": 265
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.698986975397975e-05,
      "loss": 0.5178,
      "step": 266
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.727930535455862e-05,
      "loss": 1.1289,
      "step": 267
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.756874095513748e-05,
      "loss": 0.0543,
      "step": 268
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.785817655571636e-05,
      "loss": 0.541,
      "step": 269
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.814761215629523e-05,
      "loss": 0.8145,
      "step": 270
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.84370477568741e-05,
      "loss": 0.5225,
      "step": 271
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.872648335745297e-05,
      "loss": 0.0215,
      "step": 272
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.901591895803184e-05,
      "loss": 0.6094,
      "step": 273
    },
    {
      "epoch": 0.4,
      "learning_rate": 7.930535455861072e-05,
      "loss": 1.0547,
      "step": 274
    },
    {
      "epoch": 0.4,
      "learning_rate": 7.959479015918958e-05,
      "loss": 0.6432,
      "step": 275
    },
    {
      "epoch": 0.4,
      "learning_rate": 7.988422575976846e-05,
      "loss": 1.4297,
      "step": 276
    },
    {
      "epoch": 0.4,
      "learning_rate": 8.017366136034733e-05,
      "loss": 0.438,
      "step": 277
    },
    {
      "epoch": 0.4,
      "learning_rate": 8.04630969609262e-05,
      "loss": 0.0575,
      "step": 278
    },
    {
      "epoch": 0.4,
      "learning_rate": 8.075253256150507e-05,
      "loss": 0.058,
      "step": 279
    },
    {
      "epoch": 0.4,
      "learning_rate": 8.104196816208394e-05,
      "loss": 0.9927,
      "step": 280
    },
    {
      "epoch": 0.41,
      "learning_rate": 8.133140376266281e-05,
      "loss": 0.2661,
      "step": 281
    },
    {
      "epoch": 0.41,
      "learning_rate": 8.162083936324168e-05,
      "loss": 0.5847,
      "step": 282
    },
    {
      "epoch": 0.41,
      "learning_rate": 8.191027496382056e-05,
      "loss": 0.0798,
      "step": 283
    },
    {
      "epoch": 0.41,
      "learning_rate": 8.219971056439942e-05,
      "loss": 0.0494,
      "step": 284
    },
    {
      "epoch": 0.41,
      "learning_rate": 8.248914616497829e-05,
      "loss": 0.3375,
      "step": 285
    },
    {
      "epoch": 0.41,
      "learning_rate": 8.277858176555717e-05,
      "loss": 0.0316,
      "step": 286
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.306801736613605e-05,
      "loss": 0.3951,
      "step": 287
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.335745296671491e-05,
      "loss": 0.0169,
      "step": 288
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.364688856729378e-05,
      "loss": 1.4922,
      "step": 289
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.393632416787264e-05,
      "loss": 0.0526,
      "step": 290
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.422575976845152e-05,
      "loss": 0.7547,
      "step": 291
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.45151953690304e-05,
      "loss": 0.1838,
      "step": 292
    },
    {
      "epoch": 0.42,
      "learning_rate": 8.480463096960927e-05,
      "loss": 0.0503,
      "step": 293
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.509406657018813e-05,
      "loss": 0.189,
      "step": 294
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.538350217076701e-05,
      "loss": 0.502,
      "step": 295
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.567293777134589e-05,
      "loss": 0.2156,
      "step": 296
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.596237337192476e-05,
      "loss": 0.0355,
      "step": 297
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.625180897250362e-05,
      "loss": 0.2061,
      "step": 298
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.654124457308249e-05,
      "loss": 0.2424,
      "step": 299
    },
    {
      "epoch": 0.43,
      "learning_rate": 8.683068017366137e-05,
      "loss": 0.5928,
      "step": 300
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.712011577424024e-05,
      "loss": 0.6719,
      "step": 301
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.740955137481911e-05,
      "loss": 0.0627,
      "step": 302
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.769898697539797e-05,
      "loss": 0.5321,
      "step": 303
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.798842257597685e-05,
      "loss": 0.0397,
      "step": 304
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.827785817655572e-05,
      "loss": 1.0449,
      "step": 305
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.85672937771346e-05,
      "loss": 0.1439,
      "step": 306
    },
    {
      "epoch": 0.44,
      "learning_rate": 8.885672937771346e-05,
      "loss": 0.548,
      "step": 307
    },
    {
      "epoch": 0.45,
      "learning_rate": 8.914616497829233e-05,
      "loss": 0.4983,
      "step": 308
    },
    {
      "epoch": 0.45,
      "learning_rate": 8.943560057887121e-05,
      "loss": 0.5488,
      "step": 309
    },
    {
      "epoch": 0.45,
      "learning_rate": 8.972503617945007e-05,
      "loss": 0.0569,
      "step": 310
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.001447178002895e-05,
      "loss": 0.4834,
      "step": 311
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.030390738060782e-05,
      "loss": 0.6709,
      "step": 312
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.05933429811867e-05,
      "loss": 0.0659,
      "step": 313
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.088277858176556e-05,
      "loss": 0.521,
      "step": 314
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.117221418234443e-05,
      "loss": 0.0502,
      "step": 315
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.14616497829233e-05,
      "loss": 0.3198,
      "step": 316
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.175108538350217e-05,
      "loss": 0.0432,
      "step": 317
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.204052098408105e-05,
      "loss": 0.1003,
      "step": 318
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.232995658465992e-05,
      "loss": 0.5898,
      "step": 319
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.261939218523878e-05,
      "loss": 0.0759,
      "step": 320
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.290882778581766e-05,
      "loss": 0.1221,
      "step": 321
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.319826338639654e-05,
      "loss": 0.0482,
      "step": 322
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.34876989869754e-05,
      "loss": 0.8018,
      "step": 323
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.377713458755427e-05,
      "loss": 0.4397,
      "step": 324
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.406657018813314e-05,
      "loss": 0.1709,
      "step": 325
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.435600578871201e-05,
      "loss": 0.094,
      "step": 326
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.46454413892909e-05,
      "loss": 0.1226,
      "step": 327
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.493487698986976e-05,
      "loss": 1.1562,
      "step": 328
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.522431259044862e-05,
      "loss": 0.3064,
      "step": 329
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.55137481910275e-05,
      "loss": 0.5471,
      "step": 330
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.580318379160638e-05,
      "loss": 0.3022,
      "step": 331
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.609261939218525e-05,
      "loss": 0.2545,
      "step": 332
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.638205499276411e-05,
      "loss": 0.3213,
      "step": 333
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.667149059334298e-05,
      "loss": 0.1443,
      "step": 334
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.696092619392186e-05,
      "loss": 0.2532,
      "step": 335
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.725036179450074e-05,
      "loss": 0.2549,
      "step": 336
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.75397973950796e-05,
      "loss": 0.0463,
      "step": 337
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.782923299565847e-05,
      "loss": 0.0223,
      "step": 338
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.811866859623735e-05,
      "loss": 0.048,
      "step": 339
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.840810419681621e-05,
      "loss": 1.1521,
      "step": 340
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.869753979739509e-05,
      "loss": 0.1086,
      "step": 341
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.898697539797396e-05,
      "loss": 0.8707,
      "step": 342
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.927641099855282e-05,
      "loss": 0.1023,
      "step": 343
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.95658465991317e-05,
      "loss": 0.0659,
      "step": 344
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.985528219971057e-05,
      "loss": 0.5345,
      "step": 345
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010014471780028944,
      "loss": 0.0257,
      "step": 346
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001004341534008683,
      "loss": 0.0636,
      "step": 347
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010072358900144719,
      "loss": 0.0141,
      "step": 348
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010101302460202607,
      "loss": 0.0201,
      "step": 349
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010130246020260492,
      "loss": 0.0502,
      "step": 350
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001015918958031838,
      "loss": 0.2144,
      "step": 351
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010188133140376266,
      "loss": 0.3022,
      "step": 352
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010217076700434154,
      "loss": 0.4375,
      "step": 353
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010246020260492042,
      "loss": 0.0182,
      "step": 354
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010274963820549927,
      "loss": 0.0703,
      "step": 355
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010303907380607815,
      "loss": 0.3344,
      "step": 356
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00010332850940665702,
      "loss": 0.9077,
      "step": 357
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001036179450072359,
      "loss": 0.2336,
      "step": 358
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00010390738060781478,
      "loss": 0.709,
      "step": 359
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00010419681620839363,
      "loss": 0.3228,
      "step": 360
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001044862518089725,
      "loss": 0.2303,
      "step": 361
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00010477568740955137,
      "loss": 0.0859,
      "step": 362
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00010506512301013025,
      "loss": 0.1255,
      "step": 363
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00010535455861070913,
      "loss": 0.5679,
      "step": 364
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000105643994211288,
      "loss": 0.3262,
      "step": 365
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00010593342981186687,
      "loss": 0.064,
      "step": 366
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00010622286541244573,
      "loss": 0.0355,
      "step": 367
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001065123010130246,
      "loss": 0.582,
      "step": 368
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00010680173661360348,
      "loss": 0.2993,
      "step": 369
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00010709117221418235,
      "loss": 0.5687,
      "step": 370
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00010738060781476123,
      "loss": 0.2557,
      "step": 371
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00010767004341534008,
      "loss": 0.1971,
      "step": 372
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00010795947901591896,
      "loss": 0.1919,
      "step": 373
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00010824891461649784,
      "loss": 0.0218,
      "step": 374
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001085383502170767,
      "loss": 0.0707,
      "step": 375
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00010882778581765558,
      "loss": 0.4258,
      "step": 376
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00010911722141823443,
      "loss": 0.1049,
      "step": 377
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00010940665701881331,
      "loss": 0.9863,
      "step": 378
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001096960926193922,
      "loss": 0.283,
      "step": 379
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00010998552821997106,
      "loss": 0.8438,
      "step": 380
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00011027496382054994,
      "loss": 0.1466,
      "step": 381
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00011056439942112879,
      "loss": 0.7773,
      "step": 382
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00011085383502170768,
      "loss": 0.4629,
      "step": 383
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011114327062228656,
      "loss": 0.6543,
      "step": 384
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011143270622286541,
      "loss": 0.0559,
      "step": 385
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011172214182344429,
      "loss": 0.0292,
      "step": 386
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011201157742402316,
      "loss": 0.0813,
      "step": 387
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011230101302460203,
      "loss": 0.7441,
      "step": 388
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011259044862518091,
      "loss": 0.3094,
      "step": 389
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00011287988422575977,
      "loss": 0.178,
      "step": 390
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00011316931982633864,
      "loss": 0.1458,
      "step": 391
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00011345875542691751,
      "loss": 0.7825,
      "step": 392
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00011374819102749639,
      "loss": 0.2544,
      "step": 393
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00011403762662807527,
      "loss": 0.2808,
      "step": 394
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00011432706222865412,
      "loss": 0.0527,
      "step": 395
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000114616497829233,
      "loss": 0.1658,
      "step": 396
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00011490593342981186,
      "loss": 0.606,
      "step": 397
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00011519536903039074,
      "loss": 0.3398,
      "step": 398
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00011548480463096962,
      "loss": 0.0579,
      "step": 399
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00011577424023154849,
      "loss": 0.6958,
      "step": 400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00011606367583212737,
      "loss": 0.1011,
      "step": 401
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00011635311143270622,
      "loss": 0.457,
      "step": 402
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001166425470332851,
      "loss": 0.1709,
      "step": 403
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00011693198263386398,
      "loss": 0.2092,
      "step": 404
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00011722141823444284,
      "loss": 0.2481,
      "step": 405
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00011751085383502172,
      "loss": 0.4773,
      "step": 406
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00011780028943560057,
      "loss": 0.2705,
      "step": 407
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00011808972503617945,
      "loss": 0.502,
      "step": 408
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00011837916063675833,
      "loss": 0.1675,
      "step": 409
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001186685962373372,
      "loss": 0.259,
      "step": 410
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00011895803183791607,
      "loss": 0.9707,
      "step": 411
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00011924746743849493,
      "loss": 0.3896,
      "step": 412
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001195369030390738,
      "loss": 0.2913,
      "step": 413
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001198263386396527,
      "loss": 0.915,
      "step": 414
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00012011577424023155,
      "loss": 0.2295,
      "step": 415
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00012040520984081043,
      "loss": 0.7373,
      "step": 416
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001206946454413893,
      "loss": 0.0531,
      "step": 417
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00012098408104196817,
      "loss": 0.064,
      "step": 418
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00012127351664254705,
      "loss": 0.105,
      "step": 419
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001215629522431259,
      "loss": 0.3093,
      "step": 420
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00012185238784370478,
      "loss": 0.1482,
      "step": 421
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00012214182344428363,
      "loss": 0.0113,
      "step": 422
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001224312590448625,
      "loss": 0.2091,
      "step": 423
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001227206946454414,
      "loss": 0.0088,
      "step": 424
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00012301013024602027,
      "loss": 0.0457,
      "step": 425
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00012329956584659915,
      "loss": 0.3201,
      "step": 426
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000123589001447178,
      "loss": 0.5742,
      "step": 427
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00012387843704775688,
      "loss": 0.3484,
      "step": 428
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00012416787264833576,
      "loss": 0.0786,
      "step": 429
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001244573082489146,
      "loss": 0.489,
      "step": 430
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001247467438494935,
      "loss": 0.3994,
      "step": 431
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00012503617945007234,
      "loss": 0.5774,
      "step": 432
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00012532561505065125,
      "loss": 0.1578,
      "step": 433
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00012561505065123013,
      "loss": 0.0563,
      "step": 434
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00012590448625180898,
      "loss": 0.3621,
      "step": 435
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00012619392185238786,
      "loss": 0.0391,
      "step": 436
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001264833574529667,
      "loss": 0.9502,
      "step": 437
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001267727930535456,
      "loss": 0.2712,
      "step": 438
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00012706222865412447,
      "loss": 0.0184,
      "step": 439
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00012735166425470332,
      "loss": 0.049,
      "step": 440
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001276410998552822,
      "loss": 0.4179,
      "step": 441
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00012793053545586108,
      "loss": 0.0219,
      "step": 442
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00012821997105643996,
      "loss": 0.3604,
      "step": 443
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00012850940665701884,
      "loss": 0.4763,
      "step": 444
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001287988422575977,
      "loss": 0.4058,
      "step": 445
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00012908827785817657,
      "loss": 0.127,
      "step": 446
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00012937771345875542,
      "loss": 0.0172,
      "step": 447
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001296671490593343,
      "loss": 0.3008,
      "step": 448
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00012995658465991318,
      "loss": 0.8535,
      "step": 449
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00013024602026049203,
      "loss": 0.5227,
      "step": 450
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00013053545586107093,
      "loss": 0.6855,
      "step": 451
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00013082489146164979,
      "loss": 0.0535,
      "step": 452
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00013111432706222866,
      "loss": 0.396,
      "step": 453
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00013140376266280754,
      "loss": 0.4805,
      "step": 454
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0001316931982633864,
      "loss": 0.0789,
      "step": 455
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00013198263386396527,
      "loss": 0.0667,
      "step": 456
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00013227206946454413,
      "loss": 0.8062,
      "step": 457
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000132561505065123,
      "loss": 0.0636,
      "step": 458
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00013285094066570188,
      "loss": 0.3038,
      "step": 459
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00013314037626628076,
      "loss": 1.5527,
      "step": 460
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00013342981186685964,
      "loss": 0.6377,
      "step": 461
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001337192474674385,
      "loss": 0.2294,
      "step": 462
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00013400868306801737,
      "loss": 0.8262,
      "step": 463
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00013429811866859625,
      "loss": 0.0189,
      "step": 464
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001345875542691751,
      "loss": 0.4209,
      "step": 465
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00013487698986975398,
      "loss": 0.1531,
      "step": 466
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00013516642547033283,
      "loss": 0.0201,
      "step": 467
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00013545586107091174,
      "loss": 0.2681,
      "step": 468
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00013574529667149062,
      "loss": 0.3821,
      "step": 469
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00013603473227206947,
      "loss": 0.047,
      "step": 470
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00013632416787264835,
      "loss": 0.0713,
      "step": 471
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001366136034732272,
      "loss": 0.4363,
      "step": 472
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00013690303907380608,
      "loss": 0.0745,
      "step": 473
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00013719247467438496,
      "loss": 0.0414,
      "step": 474
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001374819102749638,
      "loss": 0.0825,
      "step": 475
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001377713458755427,
      "loss": 0.0565,
      "step": 476
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00013806078147612157,
      "loss": 0.0571,
      "step": 477
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00013835021707670045,
      "loss": 0.0342,
      "step": 478
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00013863965267727933,
      "loss": 0.6047,
      "step": 479
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00013892908827785818,
      "loss": 0.0559,
      "step": 480
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00013921852387843706,
      "loss": 0.1201,
      "step": 481
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001395079594790159,
      "loss": 0.4338,
      "step": 482
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001397973950795948,
      "loss": 0.2939,
      "step": 483
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014008683068017367,
      "loss": 0.9219,
      "step": 484
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014037626628075252,
      "loss": 0.0867,
      "step": 485
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014066570188133143,
      "loss": 0.0616,
      "step": 486
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014095513748191028,
      "loss": 0.4156,
      "step": 487
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014124457308248916,
      "loss": 0.4788,
      "step": 488
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014153400868306804,
      "loss": 0.1907,
      "step": 489
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001418234442836469,
      "loss": 0.0765,
      "step": 490
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014211287988422577,
      "loss": 0.3845,
      "step": 491
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014240231548480462,
      "loss": 0.8008,
      "step": 492
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001426917510853835,
      "loss": 0.1536,
      "step": 493
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014298118668596238,
      "loss": 0.3577,
      "step": 494
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014327062228654125,
      "loss": 0.3896,
      "step": 495
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014356005788712013,
      "loss": 0.0389,
      "step": 496
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014384949348769899,
      "loss": 0.1799,
      "step": 497
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014413892908827786,
      "loss": 0.7617,
      "step": 498
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014442836468885674,
      "loss": 0.6846,
      "step": 499
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001447178002894356,
      "loss": 0.0329,
      "step": 500
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014500723589001447,
      "loss": 0.4768,
      "step": 501
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014529667149059333,
      "loss": 0.55,
      "step": 502
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014558610709117223,
      "loss": 0.2052,
      "step": 503
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001458755426917511,
      "loss": 0.0249,
      "step": 504
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014616497829232996,
      "loss": 0.051,
      "step": 505
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014645441389290884,
      "loss": 0.25,
      "step": 506
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001467438494934877,
      "loss": 0.2465,
      "step": 507
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014703328509406657,
      "loss": 0.1636,
      "step": 508
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014732272069464545,
      "loss": 0.3411,
      "step": 509
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001476121562952243,
      "loss": 0.5214,
      "step": 510
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014790159189580318,
      "loss": 0.1299,
      "step": 511
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014819102749638206,
      "loss": 0.752,
      "step": 512
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014848046309696094,
      "loss": 0.0596,
      "step": 513
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014876989869753982,
      "loss": 0.6367,
      "step": 514
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014905933429811867,
      "loss": 0.5713,
      "step": 515
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00014934876989869755,
      "loss": 0.9316,
      "step": 516
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001496382054992764,
      "loss": 0.6392,
      "step": 517
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00014992764109985528,
      "loss": 0.1814,
      "step": 518
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00015021707670043416,
      "loss": 0.3689,
      "step": 519
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00015050651230101304,
      "loss": 0.0558,
      "step": 520
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00015079594790159192,
      "loss": 0.0455,
      "step": 521
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001510853835021708,
      "loss": 0.4558,
      "step": 522
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00015137481910274965,
      "loss": 0.0275,
      "step": 523
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00015166425470332853,
      "loss": 0.3442,
      "step": 524
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00015195369030390738,
      "loss": 1.0703,
      "step": 525
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00015224312590448626,
      "loss": 0.1638,
      "step": 526
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00015253256150506514,
      "loss": 0.3613,
      "step": 527
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000152821997105644,
      "loss": 0.2158,
      "step": 528
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00015311143270622287,
      "loss": 0.0121,
      "step": 529
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00015340086830680175,
      "loss": 0.2,
      "step": 530
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00015369030390738063,
      "loss": 0.4082,
      "step": 531
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001539797395079595,
      "loss": 0.3195,
      "step": 532
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00015426917510853836,
      "loss": 0.4443,
      "step": 533
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00015455861070911724,
      "loss": 0.7148,
      "step": 534
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001548480463096961,
      "loss": 0.446,
      "step": 535
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00015513748191027497,
      "loss": 0.0763,
      "step": 536
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00015542691751085385,
      "loss": 0.6726,
      "step": 537
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00015571635311143272,
      "loss": 0.5931,
      "step": 538
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001560057887120116,
      "loss": 0.5104,
      "step": 539
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00015629522431259045,
      "loss": 0.1196,
      "step": 540
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00015658465991316933,
      "loss": 0.3542,
      "step": 541
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001568740955137482,
      "loss": 0.2173,
      "step": 542
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00015716353111432706,
      "loss": 0.04,
      "step": 543
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00015745296671490594,
      "loss": 0.342,
      "step": 544
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001577424023154848,
      "loss": 0.0574,
      "step": 545
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00015803183791606367,
      "loss": 0.5027,
      "step": 546
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00015832127351664255,
      "loss": 0.2007,
      "step": 547
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00015861070911722143,
      "loss": 0.4073,
      "step": 548
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001589001447178003,
      "loss": 0.0255,
      "step": 549
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00015918958031837916,
      "loss": 0.2272,
      "step": 550
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00015947901591895804,
      "loss": 0.2021,
      "step": 551
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00015976845151953692,
      "loss": 0.3438,
      "step": 552
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00016005788712011577,
      "loss": 0.1482,
      "step": 553
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00016034732272069465,
      "loss": 0.5955,
      "step": 554
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00016063675832127353,
      "loss": 0.0479,
      "step": 555
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001609261939218524,
      "loss": 0.4019,
      "step": 556
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001612156295224313,
      "loss": 0.4027,
      "step": 557
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00016150506512301014,
      "loss": 0.1664,
      "step": 558
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00016179450072358902,
      "loss": 0.0526,
      "step": 559
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00016208393632416787,
      "loss": 0.0497,
      "step": 560
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00016237337192474675,
      "loss": 0.5215,
      "step": 561
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00016266280752532563,
      "loss": 0.5849,
      "step": 562
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00016295224312590448,
      "loss": 0.472,
      "step": 563
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016324167872648336,
      "loss": 0.2003,
      "step": 564
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016353111432706224,
      "loss": 0.6387,
      "step": 565
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016382054992764112,
      "loss": 0.309,
      "step": 566
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016410998552822,
      "loss": 0.3745,
      "step": 567
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016439942112879885,
      "loss": 0.8184,
      "step": 568
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016468885672937773,
      "loss": 0.0325,
      "step": 569
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00016497829232995658,
      "loss": 0.2354,
      "step": 570
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016526772793053546,
      "loss": 0.0662,
      "step": 571
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016555716353111434,
      "loss": 0.0291,
      "step": 572
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016584659913169322,
      "loss": 0.0197,
      "step": 573
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001661360347322721,
      "loss": 0.4701,
      "step": 574
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016642547033285095,
      "loss": 0.4367,
      "step": 575
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016671490593342983,
      "loss": 0.9717,
      "step": 576
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001670043415340087,
      "loss": 0.349,
      "step": 577
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016729377713458756,
      "loss": 0.1098,
      "step": 578
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016758321273516644,
      "loss": 0.5322,
      "step": 579
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001678726483357453,
      "loss": 0.3179,
      "step": 580
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016816208393632417,
      "loss": 0.0468,
      "step": 581
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016845151953690305,
      "loss": 0.5146,
      "step": 582
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016874095513748192,
      "loss": 0.5583,
      "step": 583
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001690303907380608,
      "loss": 0.6935,
      "step": 584
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016931982633863965,
      "loss": 0.1995,
      "step": 585
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016960926193921853,
      "loss": 0.0502,
      "step": 586
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001698986975397974,
      "loss": 1.0508,
      "step": 587
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00017018813314037626,
      "loss": 0.5117,
      "step": 588
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00017047756874095514,
      "loss": 0.948,
      "step": 589
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00017076700434153402,
      "loss": 0.3545,
      "step": 590
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001710564399421129,
      "loss": 0.7126,
      "step": 591
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017134587554269178,
      "loss": 0.0356,
      "step": 592
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017163531114327063,
      "loss": 0.4066,
      "step": 593
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001719247467438495,
      "loss": 0.6196,
      "step": 594
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017221418234442836,
      "loss": 0.3135,
      "step": 595
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017250361794500724,
      "loss": 0.2045,
      "step": 596
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017279305354558612,
      "loss": 0.1234,
      "step": 597
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017308248914616497,
      "loss": 0.0611,
      "step": 598
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00017337192474674385,
      "loss": 0.3921,
      "step": 599
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00017366136034732273,
      "loss": 0.7344,
      "step": 600
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0001739507959479016,
      "loss": 0.6831,
      "step": 601
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0001742402315484805,
      "loss": 0.1262,
      "step": 602
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00017452966714905934,
      "loss": 0.2593,
      "step": 603
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00017481910274963822,
      "loss": 0.979,
      "step": 604
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00017510853835021707,
      "loss": 0.1428,
      "step": 605
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00017539797395079595,
      "loss": 0.0601,
      "step": 606
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00017568740955137483,
      "loss": 0.082,
      "step": 607
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001759768451519537,
      "loss": 1.3086,
      "step": 608
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001762662807525326,
      "loss": 0.0571,
      "step": 609
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00017655571635311144,
      "loss": 0.5908,
      "step": 610
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00017684515195369032,
      "loss": 0.1278,
      "step": 611
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001771345875542692,
      "loss": 0.5269,
      "step": 612
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017742402315484805,
      "loss": 0.3262,
      "step": 613
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017771345875542693,
      "loss": 0.0226,
      "step": 614
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017800289435600578,
      "loss": 0.4878,
      "step": 615
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017829232995658466,
      "loss": 0.2938,
      "step": 616
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017858176555716354,
      "loss": 0.0129,
      "step": 617
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017887120115774242,
      "loss": 0.4194,
      "step": 618
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001791606367583213,
      "loss": 0.0166,
      "step": 619
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00017945007235890015,
      "loss": 0.1968,
      "step": 620
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00017973950795947903,
      "loss": 0.0948,
      "step": 621
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001800289435600579,
      "loss": 0.2634,
      "step": 622
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00018031837916063676,
      "loss": 0.2871,
      "step": 623
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00018060781476121564,
      "loss": 0.4049,
      "step": 624
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00018089725036179451,
      "loss": 0.2789,
      "step": 625
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001811866859623734,
      "loss": 0.022,
      "step": 626
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00018147612156295227,
      "loss": 0.0571,
      "step": 627
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00018176555716353112,
      "loss": 0.1159,
      "step": 628
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00018205499276411,
      "loss": 0.0524,
      "step": 629
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00018234442836468885,
      "loss": 0.2393,
      "step": 630
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00018263386396526773,
      "loss": 0.5359,
      "step": 631
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001829232995658466,
      "loss": 0.2319,
      "step": 632
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00018321273516642546,
      "loss": 0.1389,
      "step": 633
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00018350217076700434,
      "loss": 0.2798,
      "step": 634
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00018379160636758322,
      "loss": 0.0769,
      "step": 635
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0001840810419681621,
      "loss": 0.1543,
      "step": 636
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00018437047756874098,
      "loss": 0.0673,
      "step": 637
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00018465991316931983,
      "loss": 0.2407,
      "step": 638
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0001849493487698987,
      "loss": 0.0511,
      "step": 639
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00018523878437047756,
      "loss": 0.1172,
      "step": 640
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00018552821997105644,
      "loss": 0.4808,
      "step": 641
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00018581765557163532,
      "loss": 0.1897,
      "step": 642
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001861070911722142,
      "loss": 0.0951,
      "step": 643
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00018639652677279308,
      "loss": 0.0674,
      "step": 644
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00018668596237337193,
      "loss": 0.0357,
      "step": 645
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001869753979739508,
      "loss": 0.3633,
      "step": 646
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001872648335745297,
      "loss": 0.0094,
      "step": 647
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00018755426917510854,
      "loss": 0.3036,
      "step": 648
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00018784370477568742,
      "loss": 0.464,
      "step": 649
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00018813314037626627,
      "loss": 0.1885,
      "step": 650
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00018842257597684515,
      "loss": 0.4131,
      "step": 651
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00018871201157742403,
      "loss": 0.6436,
      "step": 652
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001890014471780029,
      "loss": 0.061,
      "step": 653
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001892908827785818,
      "loss": 0.0765,
      "step": 654
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00018958031837916064,
      "loss": 0.0526,
      "step": 655
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00018986975397973952,
      "loss": 0.6934,
      "step": 656
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001901591895803184,
      "loss": 0.0625,
      "step": 657
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00019044862518089725,
      "loss": 0.228,
      "step": 658
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00019073806078147613,
      "loss": 1.2227,
      "step": 659
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.000191027496382055,
      "loss": 0.6377,
      "step": 660
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00019131693198263389,
      "loss": 0.0963,
      "step": 661
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00019160636758321276,
      "loss": 0.0621,
      "step": 662
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00019189580318379162,
      "loss": 0.3513,
      "step": 663
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001921852387843705,
      "loss": 0.0582,
      "step": 664
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00019247467438494935,
      "loss": 0.0352,
      "step": 665
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00019276410998552823,
      "loss": 0.7373,
      "step": 666
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001930535455861071,
      "loss": 0.022,
      "step": 667
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00019334298118668596,
      "loss": 0.7031,
      "step": 668
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00019363241678726484,
      "loss": 0.1364,
      "step": 669
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00019392185238784371,
      "loss": 0.052,
      "step": 670
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001942112879884226,
      "loss": 0.2917,
      "step": 671
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00019450072358900147,
      "loss": 0.4948,
      "step": 672
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00019479015918958032,
      "loss": 0.0405,
      "step": 673
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001950795947901592,
      "loss": 0.2135,
      "step": 674
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00019536903039073805,
      "loss": 0.2072,
      "step": 675
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00019565846599131693,
      "loss": 0.5903,
      "step": 676
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001959479015918958,
      "loss": 0.0421,
      "step": 677
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001962373371924747,
      "loss": 0.2101,
      "step": 678
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00019652677279305357,
      "loss": 0.229,
      "step": 679
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00019681620839363242,
      "loss": 0.1355,
      "step": 680
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001971056439942113,
      "loss": 0.7979,
      "step": 681
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00019739507959479018,
      "loss": 0.6934,
      "step": 682
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00019768451519536903,
      "loss": 0.6152,
      "step": 683
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001979739507959479,
      "loss": 0.0513,
      "step": 684
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00019826338639652676,
      "loss": 0.1348,
      "step": 685
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00019855282199710564,
      "loss": 0.0345,
      "step": 686
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00019884225759768452,
      "loss": 0.4137,
      "step": 687
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001991316931982634,
      "loss": 0.4524,
      "step": 688
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00019942112879884228,
      "loss": 0.4225,
      "step": 689
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00019971056439942113,
      "loss": 0.0942,
      "step": 690
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 0.0252,
      "step": 691
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00019996784048882457,
      "loss": 0.3408,
      "step": 692
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00019993568097764915,
      "loss": 0.3481,
      "step": 693
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001999035214664737,
      "loss": 0.386,
      "step": 694
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0001998713619552983,
      "loss": 0.2822,
      "step": 695
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00019983920244412288,
      "loss": 0.0194,
      "step": 696
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00019980704293294743,
      "loss": 0.3878,
      "step": 697
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.000199774883421772,
      "loss": 0.417,
      "step": 698
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00019974272391059657,
      "loss": 0.3496,
      "step": 699
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00019971056439942113,
      "loss": 0.2631,
      "step": 700
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00019967840488824571,
      "loss": 0.321,
      "step": 701
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00019964624537707027,
      "loss": 0.3857,
      "step": 702
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00019961408586589486,
      "loss": 0.0278,
      "step": 703
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0001995819263547194,
      "loss": 0.0323,
      "step": 704
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.000199549766843544,
      "loss": 0.0267,
      "step": 705
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00019951760733236855,
      "loss": 0.0706,
      "step": 706
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0001994854478211931,
      "loss": 0.0236,
      "step": 707
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0001994532883100177,
      "loss": 0.1183,
      "step": 708
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00019942112879884228,
      "loss": 0.0308,
      "step": 709
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00019938896928766684,
      "loss": 0.1538,
      "step": 710
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0001993568097764914,
      "loss": 0.3545,
      "step": 711
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00019932465026531598,
      "loss": 0.3336,
      "step": 712
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00019929249075414053,
      "loss": 0.1243,
      "step": 713
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00019926033124296512,
      "loss": 0.6566,
      "step": 714
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0001992281717317897,
      "loss": 0.3848,
      "step": 715
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019919601222061426,
      "loss": 0.0545,
      "step": 716
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019916385270943882,
      "loss": 0.0125,
      "step": 717
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0001991316931982634,
      "loss": 0.1392,
      "step": 718
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019909953368708796,
      "loss": 0.0367,
      "step": 719
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019906737417591254,
      "loss": 0.4795,
      "step": 720
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019903521466473712,
      "loss": 0.7667,
      "step": 721
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00019900305515356168,
      "loss": 0.0344,
      "step": 722
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019897089564238624,
      "loss": 0.0776,
      "step": 723
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019893873613121082,
      "loss": 0.0356,
      "step": 724
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019890657662003538,
      "loss": 0.0447,
      "step": 725
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019887441710885994,
      "loss": 0.0123,
      "step": 726
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019884225759768452,
      "loss": 0.6309,
      "step": 727
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0001988100980865091,
      "loss": 0.0251,
      "step": 728
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00019877793857533366,
      "loss": 0.0093,
      "step": 729
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019874577906415825,
      "loss": 0.4271,
      "step": 730
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0001987136195529828,
      "loss": 0.2076,
      "step": 731
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019868146004180736,
      "loss": 0.5195,
      "step": 732
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019864930053063194,
      "loss": 0.227,
      "step": 733
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019861714101945653,
      "loss": 0.0182,
      "step": 734
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019858498150828109,
      "loss": 0.0089,
      "step": 735
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00019855282199710564,
      "loss": 0.7885,
      "step": 736
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019852066248593023,
      "loss": 0.2239,
      "step": 737
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019848850297475478,
      "loss": 0.3374,
      "step": 738
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019845634346357934,
      "loss": 0.2722,
      "step": 739
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019842418395240395,
      "loss": 0.5312,
      "step": 740
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001983920244412285,
      "loss": 0.2378,
      "step": 741
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019835986493005307,
      "loss": 1.2676,
      "step": 742
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00019832770541887765,
      "loss": 0.0294,
      "step": 743
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0001982955459077022,
      "loss": 0.2363,
      "step": 744
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00019826338639652676,
      "loss": 0.0473,
      "step": 745
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00019823122688535137,
      "loss": 0.3779,
      "step": 746
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00019819906737417593,
      "loss": 0.1676,
      "step": 747
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0001981669078630005,
      "loss": 0.2056,
      "step": 748
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00019813474835182507,
      "loss": 0.0484,
      "step": 749
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00019810258884064963,
      "loss": 0.6602,
      "step": 750
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00019807042932947419,
      "loss": 0.4297,
      "step": 751
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00019803826981829877,
      "loss": 0.2424,
      "step": 752
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00019800611030712335,
      "loss": 0.149,
      "step": 753
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001979739507959479,
      "loss": 0.2679,
      "step": 754
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001979417912847725,
      "loss": 0.1884,
      "step": 755
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00019790963177359705,
      "loss": 0.543,
      "step": 756
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001978774722624216,
      "loss": 0.1717,
      "step": 757
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001978453127512462,
      "loss": 0.0874,
      "step": 758
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00019781315324007078,
      "loss": 0.1072,
      "step": 759
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00019778099372889533,
      "loss": 0.0767,
      "step": 760
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001977488342177199,
      "loss": 0.3069,
      "step": 761
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00019771667470654448,
      "loss": 0.4237,
      "step": 762
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00019768451519536903,
      "loss": 0.0507,
      "step": 763
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001976523556841936,
      "loss": 0.395,
      "step": 764
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0001976201961730182,
      "loss": 0.0614,
      "step": 765
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00019758803666184276,
      "loss": 0.3135,
      "step": 766
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00019755587715066731,
      "loss": 0.0808,
      "step": 767
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0001975237176394919,
      "loss": 0.4873,
      "step": 768
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00019749155812831646,
      "loss": 0.0158,
      "step": 769
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.000197459398617141,
      "loss": 0.5422,
      "step": 770
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0001974272391059656,
      "loss": 0.1547,
      "step": 771
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019739507959479018,
      "loss": 0.46,
      "step": 772
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019736292008361474,
      "loss": 0.208,
      "step": 773
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019733076057243932,
      "loss": 0.0283,
      "step": 774
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019729860106126388,
      "loss": 0.038,
      "step": 775
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019726644155008844,
      "loss": 0.9277,
      "step": 776
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00019723428203891302,
      "loss": 0.0224,
      "step": 777
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001972021225277376,
      "loss": 0.1313,
      "step": 778
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00019716996301656216,
      "loss": 0.213,
      "step": 779
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00019713780350538674,
      "loss": 0.0822,
      "step": 780
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001971056439942113,
      "loss": 0.071,
      "step": 781
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00019707348448303586,
      "loss": 0.352,
      "step": 782
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00019704132497186044,
      "loss": 0.189,
      "step": 783
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.000197009165460685,
      "loss": 0.7429,
      "step": 784
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019697700594950958,
      "loss": 0.3916,
      "step": 785
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019694484643833414,
      "loss": 0.5176,
      "step": 786
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019691268692715872,
      "loss": 0.3162,
      "step": 787
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019688052741598328,
      "loss": 0.0395,
      "step": 788
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019684836790480784,
      "loss": 0.1326,
      "step": 789
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00019681620839363242,
      "loss": 0.1671,
      "step": 790
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.000196784048882457,
      "loss": 0.5352,
      "step": 791
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00019675188937128156,
      "loss": 0.0375,
      "step": 792
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00019671972986010615,
      "loss": 0.1194,
      "step": 793
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0001966875703489307,
      "loss": 0.25,
      "step": 794
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00019665541083775526,
      "loss": 0.0228,
      "step": 795
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00019662325132657985,
      "loss": 0.0889,
      "step": 796
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0001965910918154044,
      "loss": 0.1284,
      "step": 797
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.000196558932304229,
      "loss": 0.0215,
      "step": 798
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00019652677279305357,
      "loss": 0.5572,
      "step": 799
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00019649461328187813,
      "loss": 0.1245,
      "step": 800
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00019646245377070268,
      "loss": 0.0206,
      "step": 801
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00019643029425952727,
      "loss": 0.0119,
      "step": 802
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00019639813474835183,
      "loss": 0.0217,
      "step": 803
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001963659752371764,
      "loss": 0.3288,
      "step": 804
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.000196333815726001,
      "loss": 0.0081,
      "step": 805
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00019630165621482555,
      "loss": 0.0589,
      "step": 806
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0001962694967036501,
      "loss": 0.126,
      "step": 807
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0001962373371924747,
      "loss": 0.5117,
      "step": 808
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00019620517768129925,
      "loss": 0.1243,
      "step": 809
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00019617301817012383,
      "loss": 0.021,
      "step": 810
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0001961408586589484,
      "loss": 0.7065,
      "step": 811
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00019610869914777297,
      "loss": 0.0422,
      "step": 812
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00019607653963659753,
      "loss": 0.3604,
      "step": 813
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001960443801254221,
      "loss": 0.0692,
      "step": 814
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00019601222061424667,
      "loss": 0.5098,
      "step": 815
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00019598006110307123,
      "loss": 0.3054,
      "step": 816
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001959479015918958,
      "loss": 0.3528,
      "step": 817
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001959157420807204,
      "loss": 0.2984,
      "step": 818
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00019588358256954495,
      "loss": 0.0104,
      "step": 819
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0001958514230583695,
      "loss": 0.2346,
      "step": 820
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0001958192635471941,
      "loss": 0.1403,
      "step": 821
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00019578710403601865,
      "loss": 0.2032,
      "step": 822
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00019575494452484324,
      "loss": 0.3615,
      "step": 823
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00019572278501366782,
      "loss": 0.0606,
      "step": 824
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00019569062550249238,
      "loss": 0.0425,
      "step": 825
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00019565846599131693,
      "loss": 0.0184,
      "step": 826
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019562630648014152,
      "loss": 0.2004,
      "step": 827
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019559414696896607,
      "loss": 0.3335,
      "step": 828
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019556198745779063,
      "loss": 0.2844,
      "step": 829
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019552982794661522,
      "loss": 0.4512,
      "step": 830
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001954976684354398,
      "loss": 0.5098,
      "step": 831
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019546550892426436,
      "loss": 0.1321,
      "step": 832
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00019543334941308894,
      "loss": 0.1355,
      "step": 833
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0001954011899019135,
      "loss": 0.5576,
      "step": 834
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019536903039073805,
      "loss": 0.4808,
      "step": 835
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019533687087956264,
      "loss": 0.0499,
      "step": 836
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019530471136838722,
      "loss": 0.022,
      "step": 837
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019527255185721178,
      "loss": 0.0367,
      "step": 838
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019524039234603634,
      "loss": 0.0673,
      "step": 839
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00019520823283486092,
      "loss": 0.0881,
      "step": 840
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019517607332368548,
      "loss": 0.624,
      "step": 841
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019514391381251006,
      "loss": 0.0612,
      "step": 842
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019511175430133465,
      "loss": 0.1339,
      "step": 843
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0001950795947901592,
      "loss": 0.4902,
      "step": 844
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019504743527898376,
      "loss": 0.0575,
      "step": 845
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00019501527576780834,
      "loss": 0.0995,
      "step": 846
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0001949831162566329,
      "loss": 0.015,
      "step": 847
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019495095674545746,
      "loss": 0.5713,
      "step": 848
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019491879723428207,
      "loss": 0.1938,
      "step": 849
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019488663772310663,
      "loss": 0.1047,
      "step": 850
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019485447821193118,
      "loss": 0.0495,
      "step": 851
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019482231870075577,
      "loss": 0.5898,
      "step": 852
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019479015918958032,
      "loss": 0.786,
      "step": 853
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00019475799967840488,
      "loss": 0.5889,
      "step": 854
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00019472584016722947,
      "loss": 0.6982,
      "step": 855
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00019469368065605405,
      "loss": 0.7012,
      "step": 856
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001946615211448786,
      "loss": 0.0637,
      "step": 857
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001946293616337032,
      "loss": 0.0812,
      "step": 858
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00019459720212252775,
      "loss": 0.3584,
      "step": 859
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001945650426113523,
      "loss": 0.0645,
      "step": 860
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0001945328831001769,
      "loss": 0.0497,
      "step": 861
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019450072358900147,
      "loss": 0.2715,
      "step": 862
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019446856407782603,
      "loss": 0.0211,
      "step": 863
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019443640456665059,
      "loss": 0.1919,
      "step": 864
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019440424505547517,
      "loss": 0.2264,
      "step": 865
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019437208554429973,
      "loss": 0.3857,
      "step": 866
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00019433992603312428,
      "loss": 0.3017,
      "step": 867
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001943077665219489,
      "loss": 0.0172,
      "step": 868
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00019427560701077345,
      "loss": 0.9082,
      "step": 869
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.000194243447499598,
      "loss": 0.492,
      "step": 870
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001942112879884226,
      "loss": 0.0222,
      "step": 871
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00019417912847724715,
      "loss": 0.0169,
      "step": 872
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001941469689660717,
      "loss": 0.009,
      "step": 873
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001941148094548963,
      "loss": 0.073,
      "step": 874
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019408264994372088,
      "loss": 0.0075,
      "step": 875
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019405049043254543,
      "loss": 0.0551,
      "step": 876
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019401833092137002,
      "loss": 0.3218,
      "step": 877
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019398617141019457,
      "loss": 0.3012,
      "step": 878
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019395401189901913,
      "loss": 0.0089,
      "step": 879
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00019392185238784371,
      "loss": 0.2744,
      "step": 880
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0001938896928766683,
      "loss": 0.1006,
      "step": 881
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019385753336549286,
      "loss": 0.0089,
      "step": 882
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019382537385431744,
      "loss": 0.1519,
      "step": 883
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.000193793214343142,
      "loss": 0.3051,
      "step": 884
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019376105483196655,
      "loss": 0.1578,
      "step": 885
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019372889532079114,
      "loss": 0.7026,
      "step": 886
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0001936967358096157,
      "loss": 0.6621,
      "step": 887
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00019366457629844028,
      "loss": 0.259,
      "step": 888
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019363241678726484,
      "loss": 0.3672,
      "step": 889
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019360025727608942,
      "loss": 0.0356,
      "step": 890
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019356809776491398,
      "loss": 0.0346,
      "step": 891
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019353593825373853,
      "loss": 0.5264,
      "step": 892
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019350377874256312,
      "loss": 0.1868,
      "step": 893
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001934716192313877,
      "loss": 0.2217,
      "step": 894
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00019343945972021226,
      "loss": 0.0532,
      "step": 895
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00019340730020903684,
      "loss": 0.3773,
      "step": 896
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0001933751406978614,
      "loss": 0.6562,
      "step": 897
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00019334298118668596,
      "loss": 0.1519,
      "step": 898
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00019331082167551054,
      "loss": 0.3264,
      "step": 899
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0001932786621643351,
      "loss": 0.2769,
      "step": 900
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00019324650265315968,
      "loss": 0.7891,
      "step": 901
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00019321434314198427,
      "loss": 0.076,
      "step": 902
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00019318218363080882,
      "loss": 0.5869,
      "step": 903
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00019315002411963338,
      "loss": 0.0352,
      "step": 904
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00019311786460845796,
      "loss": 0.5879,
      "step": 905
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00019308570509728252,
      "loss": 0.0486,
      "step": 906
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0001930535455861071,
      "loss": 0.0986,
      "step": 907
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0001930213860749317,
      "loss": 0.0199,
      "step": 908
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00019298922656375625,
      "loss": 0.1453,
      "step": 909
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0001929570670525808,
      "loss": 0.9766,
      "step": 910
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0001929249075414054,
      "loss": 0.2535,
      "step": 911
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00019289274803022994,
      "loss": 0.4243,
      "step": 912
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00019286058851905453,
      "loss": 0.0386,
      "step": 913
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00019282842900787908,
      "loss": 0.4512,
      "step": 914
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00019279626949670367,
      "loss": 0.0309,
      "step": 915
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00019276410998552823,
      "loss": 0.3744,
      "step": 916
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00019273195047435278,
      "loss": 0.09,
      "step": 917
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00019269979096317737,
      "loss": 0.681,
      "step": 918
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00019266763145200192,
      "loss": 0.3566,
      "step": 919
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001926354719408265,
      "loss": 0.0121,
      "step": 920
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001926033124296511,
      "loss": 0.47,
      "step": 921
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00019257115291847565,
      "loss": 0.562,
      "step": 922
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001925389934073002,
      "loss": 0.0125,
      "step": 923
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0001925068338961248,
      "loss": 0.2288,
      "step": 924
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00019247467438494935,
      "loss": 0.0442,
      "step": 925
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00019244251487377393,
      "loss": 0.126,
      "step": 926
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00019241035536259851,
      "loss": 0.0255,
      "step": 927
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00019237819585142307,
      "loss": 0.4468,
      "step": 928
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00019234603634024763,
      "loss": 0.5752,
      "step": 929
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0001923138768290722,
      "loss": 0.8468,
      "step": 930
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00019228171731789677,
      "loss": 0.2637,
      "step": 931
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00019224955780672133,
      "loss": 0.4243,
      "step": 932
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00019221739829554594,
      "loss": 0.2211,
      "step": 933
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0001921852387843705,
      "loss": 0.9297,
      "step": 934
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00019215307927319505,
      "loss": 0.3627,
      "step": 935
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00019212091976201964,
      "loss": 0.7842,
      "step": 936
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.0001920887602508442,
      "loss": 0.0265,
      "step": 937
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00019205660073966875,
      "loss": 0.069,
      "step": 938
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00019202444122849333,
      "loss": 0.3796,
      "step": 939
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00019199228171731792,
      "loss": 0.0178,
      "step": 940
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00019196012220614248,
      "loss": 0.1887,
      "step": 941
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00019192796269496703,
      "loss": 0.006,
      "step": 942
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00019189580318379162,
      "loss": 0.1221,
      "step": 943
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019186364367261617,
      "loss": 0.4207,
      "step": 944
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019183148416144076,
      "loss": 0.2934,
      "step": 945
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019179932465026534,
      "loss": 0.295,
      "step": 946
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0001917671651390899,
      "loss": 0.8662,
      "step": 947
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019173500562791446,
      "loss": 0.2792,
      "step": 948
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00019170284611673904,
      "loss": 0.2147,
      "step": 949
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0001916706866055636,
      "loss": 0.3335,
      "step": 950
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019163852709438815,
      "loss": 0.0969,
      "step": 951
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019160636758321276,
      "loss": 0.2739,
      "step": 952
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019157420807203732,
      "loss": 0.4346,
      "step": 953
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019154204856086188,
      "loss": 0.0239,
      "step": 954
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019150988904968646,
      "loss": 0.2833,
      "step": 955
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019147772953851102,
      "loss": 0.024,
      "step": 956
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00019144557002733558,
      "loss": 0.0626,
      "step": 957
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0001914134105161602,
      "loss": 0.0427,
      "step": 958
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00019138125100498474,
      "loss": 0.353,
      "step": 959
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0001913490914938093,
      "loss": 0.3364,
      "step": 960
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00019131693198263389,
      "loss": 0.3345,
      "step": 961
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00019128477247145844,
      "loss": 0.0337,
      "step": 962
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.000191252612960283,
      "loss": 0.0411,
      "step": 963
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00019122045344910758,
      "loss": 0.3207,
      "step": 964
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019118829393793217,
      "loss": 0.2686,
      "step": 965
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019115613442675672,
      "loss": 0.2738,
      "step": 966
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019112397491558128,
      "loss": 0.072,
      "step": 967
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019109181540440587,
      "loss": 0.2651,
      "step": 968
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00019105965589323042,
      "loss": 0.0262,
      "step": 969
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.000191027496382055,
      "loss": 0.2111,
      "step": 970
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0001909953368708796,
      "loss": 0.0267,
      "step": 971
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00019096317735970415,
      "loss": 0.198,
      "step": 972
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0001909310178485287,
      "loss": 0.203,
      "step": 973
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0001908988583373533,
      "loss": 0.0707,
      "step": 974
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00019086669882617785,
      "loss": 0.0169,
      "step": 975
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0001908345393150024,
      "loss": 0.3423,
      "step": 976
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00019080237980382699,
      "loss": 1.1914,
      "step": 977
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00019077022029265157,
      "loss": 0.6133,
      "step": 978
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00019073806078147613,
      "loss": 0.1249,
      "step": 979
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0001907059012703007,
      "loss": 0.9961,
      "step": 980
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00019067374175912527,
      "loss": 0.6514,
      "step": 981
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00019064158224794983,
      "loss": 0.2559,
      "step": 982
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0001906094227367744,
      "loss": 1.457,
      "step": 983
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.000190577263225599,
      "loss": 0.0573,
      "step": 984
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00019054510371442355,
      "loss": 0.2492,
      "step": 985
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00019051294420324813,
      "loss": 0.1501,
      "step": 986
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0001904807846920727,
      "loss": 0.308,
      "step": 987
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00019044862518089725,
      "loss": 0.2319,
      "step": 988
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00019041646566972183,
      "loss": 0.0435,
      "step": 989
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0001903843061585464,
      "loss": 0.2747,
      "step": 990
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00019035214664737097,
      "loss": 0.069,
      "step": 991
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00019031998713619553,
      "loss": 0.0928,
      "step": 992
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019028782762502011,
      "loss": 0.0871,
      "step": 993
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019025566811384467,
      "loss": 0.2117,
      "step": 994
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019022350860266926,
      "loss": 0.0742,
      "step": 995
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0001901913490914938,
      "loss": 0.0276,
      "step": 996
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0001901591895803184,
      "loss": 0.0288,
      "step": 997
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019012703006914295,
      "loss": 0.0482,
      "step": 998
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00019009487055796754,
      "loss": 0.0405,
      "step": 999
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.0001900627110467921,
      "loss": 0.4185,
      "step": 1000
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00019003055153561665,
      "loss": 0.052,
      "step": 1001
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00018999839202444124,
      "loss": 0.2536,
      "step": 1002
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00018996623251326582,
      "loss": 0.0666,
      "step": 1003
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00018993407300209038,
      "loss": 0.1761,
      "step": 1004
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00018990191349091496,
      "loss": 0.1955,
      "step": 1005
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00018986975397973952,
      "loss": 0.6914,
      "step": 1006
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00018983759446856407,
      "loss": 0.3228,
      "step": 1007
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00018980543495738866,
      "loss": 0.3789,
      "step": 1008
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00018977327544621322,
      "loss": 0.196,
      "step": 1009
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0001897411159350378,
      "loss": 0.5684,
      "step": 1010
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00018970895642386238,
      "loss": 0.1248,
      "step": 1011
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00018967679691268694,
      "loss": 0.4219,
      "step": 1012
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0001896446374015115,
      "loss": 0.4936,
      "step": 1013
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00018961247789033608,
      "loss": 0.085,
      "step": 1014
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00018958031837916064,
      "loss": 0.7246,
      "step": 1015
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00018954815886798522,
      "loss": 0.1851,
      "step": 1016
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00018951599935680978,
      "loss": 0.6416,
      "step": 1017
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00018948383984563436,
      "loss": 0.2969,
      "step": 1018
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00018945168033445892,
      "loss": 0.0389,
      "step": 1019
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0001894195208232835,
      "loss": 0.7307,
      "step": 1020
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00018938736131210806,
      "loss": 0.5123,
      "step": 1021
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00018935520180093262,
      "loss": 0.6025,
      "step": 1022
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0001893230422897572,
      "loss": 0.136,
      "step": 1023
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0001892908827785818,
      "loss": 0.1534,
      "step": 1024
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00018925872326740634,
      "loss": 0.2471,
      "step": 1025
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0001892265637562309,
      "loss": 0.2319,
      "step": 1026
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00018919440424505548,
      "loss": 0.2034,
      "step": 1027
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00018916224473388004,
      "loss": 0.5246,
      "step": 1028
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00018913008522270463,
      "loss": 0.0864,
      "step": 1029
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0001890979257115292,
      "loss": 0.328,
      "step": 1030
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00018906576620035377,
      "loss": 0.0555,
      "step": 1031
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00018903360668917832,
      "loss": 0.2856,
      "step": 1032
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0001890014471780029,
      "loss": 0.0161,
      "step": 1033
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00018896928766682746,
      "loss": 0.5574,
      "step": 1034
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00018893712815565202,
      "loss": 0.0394,
      "step": 1035
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00018890496864447663,
      "loss": 0.5786,
      "step": 1036
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0001888728091333012,
      "loss": 0.6406,
      "step": 1037
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00018884064962212575,
      "loss": 0.0483,
      "step": 1038
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00018880849011095033,
      "loss": 0.3685,
      "step": 1039
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0001887763305997749,
      "loss": 0.5945,
      "step": 1040
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00018874417108859944,
      "loss": 0.22,
      "step": 1041
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00018871201157742403,
      "loss": 0.24,
      "step": 1042
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0001886798520662486,
      "loss": 0.319,
      "step": 1043
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00018864769255507317,
      "loss": 0.1765,
      "step": 1044
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00018861553304389775,
      "loss": 0.3809,
      "step": 1045
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0001885833735327223,
      "loss": 0.1314,
      "step": 1046
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00018855121402154687,
      "loss": 0.3179,
      "step": 1047
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00018851905451037145,
      "loss": 0.2455,
      "step": 1048
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00018848689499919604,
      "loss": 0.3326,
      "step": 1049
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001884547354880206,
      "loss": 0.4805,
      "step": 1050
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00018842257597684515,
      "loss": 0.4333,
      "step": 1051
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00018839041646566973,
      "loss": 0.3871,
      "step": 1052
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001883582569544943,
      "loss": 0.0124,
      "step": 1053
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00018832609744331885,
      "loss": 0.3501,
      "step": 1054
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018829393793214346,
      "loss": 0.0182,
      "step": 1055
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018826177842096802,
      "loss": 0.5273,
      "step": 1056
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018822961890979257,
      "loss": 1.0527,
      "step": 1057
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018819745939861716,
      "loss": 0.0105,
      "step": 1058
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018816529988744171,
      "loss": 0.0106,
      "step": 1059
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018813314037626627,
      "loss": 0.1983,
      "step": 1060
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00018810098086509088,
      "loss": 0.4026,
      "step": 1061
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00018806882135391544,
      "loss": 0.035,
      "step": 1062
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00018803666184274,
      "loss": 0.4431,
      "step": 1063
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00018800450233156458,
      "loss": 0.2834,
      "step": 1064
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00018797234282038914,
      "loss": 0.6074,
      "step": 1065
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0001879401833092137,
      "loss": 0.0211,
      "step": 1066
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00018790802379803828,
      "loss": 0.0605,
      "step": 1067
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00018787586428686286,
      "loss": 0.1231,
      "step": 1068
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.00018784370477568742,
      "loss": 0.2887,
      "step": 1069
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.000187811545264512,
      "loss": 0.6699,
      "step": 1070
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.00018777938575333656,
      "loss": 0.5364,
      "step": 1071
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.00018774722624216112,
      "loss": 0.2706,
      "step": 1072
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0001877150667309857,
      "loss": 0.312,
      "step": 1073
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.00018768290721981029,
      "loss": 0.5498,
      "step": 1074
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.00018765074770863484,
      "loss": 0.7256,
      "step": 1075
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0001876185881974594,
      "loss": 0.1799,
      "step": 1076
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.00018758642868628398,
      "loss": 0.1133,
      "step": 1077
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.00018755426917510854,
      "loss": 0.0159,
      "step": 1078
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0001875221096639331,
      "loss": 0.3236,
      "step": 1079
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.00018748995015275768,
      "loss": 0.1868,
      "step": 1080
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.00018745779064158227,
      "loss": 0.6394,
      "step": 1081
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.00018742563113040682,
      "loss": 0.498,
      "step": 1082
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0001873934716192314,
      "loss": 0.1713,
      "step": 1083
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.00018736131210805596,
      "loss": 0.0169,
      "step": 1084
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.00018732915259688052,
      "loss": 0.6252,
      "step": 1085
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0001872969930857051,
      "loss": 0.0166,
      "step": 1086
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0001872648335745297,
      "loss": 0.2162,
      "step": 1087
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.00018723267406335425,
      "loss": 0.2617,
      "step": 1088
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.00018720051455217883,
      "loss": 0.5088,
      "step": 1089
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0001871683550410034,
      "loss": 0.0519,
      "step": 1090
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.00018713619552982794,
      "loss": 0.1472,
      "step": 1091
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.00018710403601865253,
      "loss": 0.2544,
      "step": 1092
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0001870718765074771,
      "loss": 0.0144,
      "step": 1093
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.00018703971699630167,
      "loss": 0.4844,
      "step": 1094
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.00018700755748512625,
      "loss": 0.9346,
      "step": 1095
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0001869753979739508,
      "loss": 0.5234,
      "step": 1096
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.00018694323846277537,
      "loss": 0.0341,
      "step": 1097
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.00018691107895159995,
      "loss": 0.1488,
      "step": 1098
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0001868789194404245,
      "loss": 0.0436,
      "step": 1099
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0001868467599292491,
      "loss": 0.0294,
      "step": 1100
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.00018681460041807365,
      "loss": 0.7031,
      "step": 1101
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.00018678244090689823,
      "loss": 0.1275,
      "step": 1102
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.0001867502813957228,
      "loss": 0.1259,
      "step": 1103
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00018671812188454735,
      "loss": 0.1619,
      "step": 1104
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00018668596237337193,
      "loss": 0.9824,
      "step": 1105
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00018665380286219651,
      "loss": 0.2413,
      "step": 1106
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00018662164335102107,
      "loss": 0.0363,
      "step": 1107
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00018658948383984566,
      "loss": 0.5703,
      "step": 1108
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.0001865573243286702,
      "loss": 0.0309,
      "step": 1109
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.00018652516481749477,
      "loss": 0.4009,
      "step": 1110
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.00018649300530631935,
      "loss": 0.4575,
      "step": 1111
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0001864608457951439,
      "loss": 0.0128,
      "step": 1112
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0001864286862839685,
      "loss": 0.3362,
      "step": 1113
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.00018639652677279308,
      "loss": 0.0386,
      "step": 1114
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.00018636436726161764,
      "loss": 0.1039,
      "step": 1115
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0001863322077504422,
      "loss": 1.1211,
      "step": 1116
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00018630004823926678,
      "loss": 0.3276,
      "step": 1117
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00018626788872809133,
      "loss": 0.0742,
      "step": 1118
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00018623572921691592,
      "loss": 0.2123,
      "step": 1119
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00018620356970574047,
      "loss": 0.3115,
      "step": 1120
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00018617141019456506,
      "loss": 0.0569,
      "step": 1121
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.00018613925068338962,
      "loss": 0.3574,
      "step": 1122
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.0001861070911722142,
      "loss": 0.0503,
      "step": 1123
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00018607493166103876,
      "loss": 0.2169,
      "step": 1124
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00018604277214986331,
      "loss": 0.4248,
      "step": 1125
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0001860106126386879,
      "loss": 0.5635,
      "step": 1126
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00018597845312751248,
      "loss": 0.4785,
      "step": 1127
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00018594629361633704,
      "loss": 0.0535,
      "step": 1128
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0001859141341051616,
      "loss": 0.2972,
      "step": 1129
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00018588197459398618,
      "loss": 0.2052,
      "step": 1130
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00018584981508281074,
      "loss": 0.0143,
      "step": 1131
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00018581765557163532,
      "loss": 0.3059,
      "step": 1132
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0001857854960604599,
      "loss": 0.0747,
      "step": 1133
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00018575333654928446,
      "loss": 0.3878,
      "step": 1134
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00018572117703810902,
      "loss": 0.4272,
      "step": 1135
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0001856890175269336,
      "loss": 0.041,
      "step": 1136
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.00018565685801575816,
      "loss": 0.4455,
      "step": 1137
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018562469850458274,
      "loss": 0.6536,
      "step": 1138
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018559253899340733,
      "loss": 0.4277,
      "step": 1139
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018556037948223188,
      "loss": 0.093,
      "step": 1140
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018552821997105644,
      "loss": 0.2984,
      "step": 1141
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018549606045988103,
      "loss": 0.0369,
      "step": 1142
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018546390094870558,
      "loss": 0.0335,
      "step": 1143
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00018543174143753014,
      "loss": 0.3027,
      "step": 1144
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00018539958192635472,
      "loss": 0.4346,
      "step": 1145
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0001853674224151793,
      "loss": 0.3199,
      "step": 1146
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00018533526290400387,
      "loss": 0.2645,
      "step": 1147
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00018530310339282845,
      "loss": 0.5078,
      "step": 1148
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.000185270943881653,
      "loss": 0.0616,
      "step": 1149
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00018523878437047756,
      "loss": 0.3135,
      "step": 1150
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00018520662485930215,
      "loss": 0.0333,
      "step": 1151
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00018517446534812673,
      "loss": 0.0086,
      "step": 1152
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.0001851423058369513,
      "loss": 0.0484,
      "step": 1153
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00018511014632577585,
      "loss": 0.1115,
      "step": 1154
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00018507798681460043,
      "loss": 0.7109,
      "step": 1155
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00018504582730342499,
      "loss": 0.0742,
      "step": 1156
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00018501366779224954,
      "loss": 0.4846,
      "step": 1157
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.00018498150828107415,
      "loss": 0.5688,
      "step": 1158
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.0001849493487698987,
      "loss": 0.0156,
      "step": 1159
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00018491718925872327,
      "loss": 0.8037,
      "step": 1160
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00018488502974754785,
      "loss": 0.1136,
      "step": 1161
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.0001848528702363724,
      "loss": 0.0808,
      "step": 1162
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00018482071072519697,
      "loss": 0.0624,
      "step": 1163
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00018478855121402158,
      "loss": 0.0318,
      "step": 1164
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00018475639170284613,
      "loss": 0.0629,
      "step": 1165
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0001847242321916707,
      "loss": 0.0481,
      "step": 1166
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.00018469207268049528,
      "loss": 0.2267,
      "step": 1167
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.00018465991316931983,
      "loss": 1.2656,
      "step": 1168
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0001846277536581444,
      "loss": 0.0562,
      "step": 1169
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.00018459559414696897,
      "loss": 0.2952,
      "step": 1170
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.00018456343463579356,
      "loss": 0.2499,
      "step": 1171
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.00018453127512461811,
      "loss": 0.4171,
      "step": 1172
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0001844991156134427,
      "loss": 0.1648,
      "step": 1173
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.00018446695610226726,
      "loss": 0.1066,
      "step": 1174
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0001844347965910918,
      "loss": 0.0715,
      "step": 1175
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0001844026370799164,
      "loss": 0.2234,
      "step": 1176
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.00018437047756874098,
      "loss": 0.6753,
      "step": 1177
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.00018433831805756554,
      "loss": 0.012,
      "step": 1178
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0001843061585463901,
      "loss": 0.2018,
      "step": 1179
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.00018427399903521468,
      "loss": 0.0097,
      "step": 1180
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.00018424183952403924,
      "loss": 0.8384,
      "step": 1181
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0001842096800128638,
      "loss": 0.2432,
      "step": 1182
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0001841775205016884,
      "loss": 0.583,
      "step": 1183
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.00018414536099051296,
      "loss": 0.0136,
      "step": 1184
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.00018411320147933752,
      "loss": 0.2087,
      "step": 1185
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0001840810419681621,
      "loss": 0.6973,
      "step": 1186
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.00018404888245698666,
      "loss": 0.3066,
      "step": 1187
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.00018401672294581122,
      "loss": 0.3223,
      "step": 1188
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0001839845634346358,
      "loss": 0.1769,
      "step": 1189
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.00018395240392346038,
      "loss": 0.2669,
      "step": 1190
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.00018392024441228494,
      "loss": 0.0819,
      "step": 1191
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.00018388808490110952,
      "loss": 0.2836,
      "step": 1192
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00018385592538993408,
      "loss": 0.219,
      "step": 1193
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00018382376587875864,
      "loss": 0.5264,
      "step": 1194
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00018379160636758322,
      "loss": 0.0989,
      "step": 1195
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0001837594468564078,
      "loss": 0.3771,
      "step": 1196
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00018372728734523236,
      "loss": 0.1754,
      "step": 1197
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.00018369512783405695,
      "loss": 0.0994,
      "step": 1198
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0001836629683228815,
      "loss": 0.0092,
      "step": 1199
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.00018363080881170606,
      "loss": 0.75,
      "step": 1200
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.00018359864930053065,
      "loss": 0.2956,
      "step": 1201
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0001835664897893552,
      "loss": 0.0122,
      "step": 1202
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0001835343302781798,
      "loss": 0.4326,
      "step": 1203
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.00018350217076700434,
      "loss": 0.228,
      "step": 1204
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.00018347001125582893,
      "loss": 0.0306,
      "step": 1205
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.00018343785174465348,
      "loss": 0.0407,
      "step": 1206
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.00018340569223347804,
      "loss": 0.1704,
      "step": 1207
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.00018337353272230263,
      "loss": 0.5367,
      "step": 1208
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0001833413732111272,
      "loss": 0.1538,
      "step": 1209
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.00018330921369995177,
      "loss": 0.2784,
      "step": 1210
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.00018327705418877635,
      "loss": 0.0811,
      "step": 1211
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0001832448946776009,
      "loss": 0.0466,
      "step": 1212
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.00018321273516642546,
      "loss": 0.0638,
      "step": 1213
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.00018318057565525005,
      "loss": 0.0652,
      "step": 1214
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0001831484161440746,
      "loss": 0.5955,
      "step": 1215
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0001831162566328992,
      "loss": 0.4305,
      "step": 1216
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.00018308409712172377,
      "loss": 0.4907,
      "step": 1217
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.00018305193761054833,
      "loss": 0.0278,
      "step": 1218
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0001830197780993729,
      "loss": 0.0686,
      "step": 1219
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.00018298761858819747,
      "loss": 0.3708,
      "step": 1220
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.00018295545907702203,
      "loss": 0.4863,
      "step": 1221
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0001829232995658466,
      "loss": 0.2507,
      "step": 1222
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0001828911400546712,
      "loss": 0.1841,
      "step": 1223
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.00018285898054349575,
      "loss": 0.8984,
      "step": 1224
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0001828268210323203,
      "loss": 0.6284,
      "step": 1225
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0001827946615211449,
      "loss": 0.3675,
      "step": 1226
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.00018276250200996945,
      "loss": 0.5925,
      "step": 1227
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.000182730342498794,
      "loss": 0.4229,
      "step": 1228
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0001826981829876186,
      "loss": 0.0353,
      "step": 1229
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.00018266602347644318,
      "loss": 0.0444,
      "step": 1230
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.00018263386396526773,
      "loss": 0.2043,
      "step": 1231
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0001826017044540923,
      "loss": 0.5171,
      "step": 1232
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.00018256954494291687,
      "loss": 0.3355,
      "step": 1233
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.00018253738543174143,
      "loss": 0.7305,
      "step": 1234
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00018250522592056602,
      "loss": 0.0635,
      "step": 1235
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0001824730664093906,
      "loss": 0.0251,
      "step": 1236
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00018244090689821516,
      "loss": 0.1512,
      "step": 1237
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00018240874738703971,
      "loss": 0.0754,
      "step": 1238
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0001823765878758643,
      "loss": 0.1655,
      "step": 1239
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00018234442836468885,
      "loss": 0.3105,
      "step": 1240
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00018231226885351344,
      "loss": 0.1657,
      "step": 1241
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018228010934233802,
      "loss": 0.0334,
      "step": 1242
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018224794983116258,
      "loss": 0.2773,
      "step": 1243
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018221579031998714,
      "loss": 0.3327,
      "step": 1244
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018218363080881172,
      "loss": 0.4805,
      "step": 1245
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018215147129763628,
      "loss": 0.323,
      "step": 1246
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018211931178646083,
      "loss": 0.0529,
      "step": 1247
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.00018208715227528545,
      "loss": 0.0443,
      "step": 1248
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00018205499276411,
      "loss": 0.3828,
      "step": 1249
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00018202283325293456,
      "loss": 0.6518,
      "step": 1250
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00018199067374175914,
      "loss": 0.7816,
      "step": 1251
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0001819585142305837,
      "loss": 0.2339,
      "step": 1252
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00018192635471940826,
      "loss": 0.2413,
      "step": 1253
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00018189419520823284,
      "loss": 0.2589,
      "step": 1254
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00018186203569705743,
      "loss": 0.2373,
      "step": 1255
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00018182987618588198,
      "loss": 0.1313,
      "step": 1256
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00018179771667470654,
      "loss": 1.1133,
      "step": 1257
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00018176555716353112,
      "loss": 0.0735,
      "step": 1258
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00018173339765235568,
      "loss": 0.0523,
      "step": 1259
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00018170123814118027,
      "loss": 0.1436,
      "step": 1260
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00018166907863000485,
      "loss": 0.1799,
      "step": 1261
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.0001816369191188294,
      "loss": 0.7617,
      "step": 1262
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00018160475960765396,
      "loss": 0.0324,
      "step": 1263
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00018157260009647855,
      "loss": 0.1908,
      "step": 1264
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.0001815404405853031,
      "loss": 0.3379,
      "step": 1265
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00018150828107412766,
      "loss": 0.0172,
      "step": 1266
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00018147612156295227,
      "loss": 0.0453,
      "step": 1267
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.00018144396205177683,
      "loss": 0.1499,
      "step": 1268
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00018141180254060139,
      "loss": 0.0984,
      "step": 1269
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00018137964302942597,
      "loss": 0.4248,
      "step": 1270
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00018134748351825053,
      "loss": 0.5939,
      "step": 1271
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00018131532400707508,
      "loss": 0.1431,
      "step": 1272
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00018128316449589967,
      "loss": 0.458,
      "step": 1273
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00018125100498472425,
      "loss": 0.5557,
      "step": 1274
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0001812188454735488,
      "loss": 0.2148,
      "step": 1275
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0001811866859623734,
      "loss": 0.046,
      "step": 1276
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.00018115452645119795,
      "loss": 0.7124,
      "step": 1277
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0001811223669400225,
      "loss": 0.0284,
      "step": 1278
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0001810902074288471,
      "loss": 0.3447,
      "step": 1279
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.00018105804791767168,
      "loss": 0.0402,
      "step": 1280
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.00018102588840649623,
      "loss": 0.0349,
      "step": 1281
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0001809937288953208,
      "loss": 0.0183,
      "step": 1282
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.00018096156938414537,
      "loss": 0.0629,
      "step": 1283
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.00018092940987296993,
      "loss": 0.0102,
      "step": 1284
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.00018089725036179451,
      "loss": 0.4517,
      "step": 1285
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0001808650908506191,
      "loss": 0.998,
      "step": 1286
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.00018083293133944366,
      "loss": 0.2278,
      "step": 1287
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0001808007718282682,
      "loss": 0.1328,
      "step": 1288
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0001807686123170928,
      "loss": 0.7578,
      "step": 1289
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.00018073645280591735,
      "loss": 0.0531,
      "step": 1290
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0001807042932947419,
      "loss": 0.4336,
      "step": 1291
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0001806721337835665,
      "loss": 0.2167,
      "step": 1292
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.00018063997427239108,
      "loss": 0.2501,
      "step": 1293
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.00018060781476121564,
      "loss": 0.4304,
      "step": 1294
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.00018057565525004022,
      "loss": 0.8184,
      "step": 1295
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.00018054349573886478,
      "loss": 0.0942,
      "step": 1296
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.00018051133622768933,
      "loss": 1.062,
      "step": 1297
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.00018047917671651392,
      "loss": 0.0394,
      "step": 1298
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0001804470172053385,
      "loss": 0.0308,
      "step": 1299
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.00018041485769416306,
      "loss": 0.1382,
      "step": 1300
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.00018038269818298764,
      "loss": 0.2758,
      "step": 1301
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0001803505386718122,
      "loss": 0.2295,
      "step": 1302
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.00018031837916063676,
      "loss": 0.0132,
      "step": 1303
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.00018028621964946134,
      "loss": 0.2671,
      "step": 1304
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.0001802540601382859,
      "loss": 0.2633,
      "step": 1305
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.00018022190062711048,
      "loss": 0.0056,
      "step": 1306
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.00018018974111593504,
      "loss": 0.1809,
      "step": 1307
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.00018015758160475962,
      "loss": 0.6835,
      "step": 1308
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.00018012542209358418,
      "loss": 0.4326,
      "step": 1309
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.00018009326258240876,
      "loss": 0.2063,
      "step": 1310
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00018006110307123332,
      "loss": 0.7178,
      "step": 1311
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0001800289435600579,
      "loss": 0.0146,
      "step": 1312
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00017999678404888246,
      "loss": 0.041,
      "step": 1313
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00017996462453770705,
      "loss": 0.5034,
      "step": 1314
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0001799324650265316,
      "loss": 0.7383,
      "step": 1315
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00017990030551535616,
      "loss": 0.0416,
      "step": 1316
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.00017986814600418074,
      "loss": 0.5586,
      "step": 1317
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.0001798359864930053,
      "loss": 0.3043,
      "step": 1318
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00017980382698182988,
      "loss": 0.2427,
      "step": 1319
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00017977166747065447,
      "loss": 0.4307,
      "step": 1320
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00017973950795947903,
      "loss": 0.0485,
      "step": 1321
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00017970734844830358,
      "loss": 0.7734,
      "step": 1322
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00017967518893712817,
      "loss": 0.0746,
      "step": 1323
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.00017964302942595272,
      "loss": 0.0929,
      "step": 1324
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0001796108699147773,
      "loss": 0.6484,
      "step": 1325
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0001795787104036019,
      "loss": 0.365,
      "step": 1326
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.00017954655089242645,
      "loss": 0.0298,
      "step": 1327
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.000179514391381251,
      "loss": 0.2339,
      "step": 1328
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0001794822318700756,
      "loss": 0.0439,
      "step": 1329
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.00017945007235890015,
      "loss": 0.477,
      "step": 1330
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.00017941791284772473,
      "loss": 0.0862,
      "step": 1331
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0001793857533365493,
      "loss": 0.1543,
      "step": 1332
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.00017935359382537387,
      "loss": 0.6621,
      "step": 1333
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.00017932143431419843,
      "loss": 0.0511,
      "step": 1334
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.000179289274803023,
      "loss": 0.0583,
      "step": 1335
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.00017925711529184757,
      "loss": 0.0653,
      "step": 1336
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.00017922495578067213,
      "loss": 0.3071,
      "step": 1337
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0001791927962694967,
      "loss": 0.0223,
      "step": 1338
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0001791606367583213,
      "loss": 0.022,
      "step": 1339
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.00017912847724714585,
      "loss": 0.4209,
      "step": 1340
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0001790963177359704,
      "loss": 0.351,
      "step": 1341
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.000179064158224795,
      "loss": 0.019,
      "step": 1342
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.00017903199871361955,
      "loss": 0.4434,
      "step": 1343
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.00017899983920244413,
      "loss": 0.1196,
      "step": 1344
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017896767969126872,
      "loss": 0.0281,
      "step": 1345
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017893552018009327,
      "loss": 0.025,
      "step": 1346
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017890336066891783,
      "loss": 0.2275,
      "step": 1347
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017887120115774242,
      "loss": 0.091,
      "step": 1348
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017883904164656697,
      "loss": 0.2425,
      "step": 1349
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017880688213539153,
      "loss": 0.0718,
      "step": 1350
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00017877472262421614,
      "loss": 0.0259,
      "step": 1351
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0001787425631130407,
      "loss": 0.3746,
      "step": 1352
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00017871040360186525,
      "loss": 0.1061,
      "step": 1353
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00017867824409068984,
      "loss": 0.6211,
      "step": 1354
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0001786460845795144,
      "loss": 0.1288,
      "step": 1355
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00017861392506833895,
      "loss": 0.25,
      "step": 1356
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00017858176555716354,
      "loss": 0.1924,
      "step": 1357
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00017854960604598812,
      "loss": 0.0574,
      "step": 1358
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00017851744653481268,
      "loss": 0.0112,
      "step": 1359
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00017848528702363726,
      "loss": 0.0092,
      "step": 1360
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00017845312751246182,
      "loss": 0.4004,
      "step": 1361
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00017842096800128638,
      "loss": 0.3518,
      "step": 1362
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00017838880849011096,
      "loss": 0.0601,
      "step": 1363
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.00017835664897893554,
      "loss": 0.1061,
      "step": 1364
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0001783244894677601,
      "loss": 0.4639,
      "step": 1365
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00017829232995658466,
      "loss": 0.2524,
      "step": 1366
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00017826017044540924,
      "loss": 0.0637,
      "step": 1367
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.0001782280109342338,
      "loss": 0.0394,
      "step": 1368
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00017819585142305836,
      "loss": 0.0407,
      "step": 1369
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00017816369191188297,
      "loss": 0.7366,
      "step": 1370
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00017813153240070752,
      "loss": 0.2137,
      "step": 1371
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00017809937288953208,
      "loss": 0.3064,
      "step": 1372
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.00017806721337835667,
      "loss": 0.0546,
      "step": 1373
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.00017803505386718122,
      "loss": 0.457,
      "step": 1374
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.00017800289435600578,
      "loss": 0.2714,
      "step": 1375
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0001779707348448304,
      "loss": 0.0291,
      "step": 1376
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.00017793857533365495,
      "loss": 0.0305,
      "step": 1377
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0001779064158224795,
      "loss": 1.0703,
      "step": 1378
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0001778742563113041,
      "loss": 0.3367,
      "step": 1379
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.00017784209680012865,
      "loss": 0.0811,
      "step": 1380
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0001778099372889532,
      "loss": 0.0503,
      "step": 1381
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.00017777777777777779,
      "loss": 0.3013,
      "step": 1382
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.00017774561826660237,
      "loss": 0.4683,
      "step": 1383
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.00017771345875542693,
      "loss": 0.1169,
      "step": 1384
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0001776812992442515,
      "loss": 0.333,
      "step": 1385
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.00017764913973307607,
      "loss": 0.0984,
      "step": 1386
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.00017761698022190063,
      "loss": 0.209,
      "step": 1387
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0001775848207107252,
      "loss": 0.0814,
      "step": 1388
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0001775526611995498,
      "loss": 0.4619,
      "step": 1389
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.00017752050168837435,
      "loss": 0.2397,
      "step": 1390
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0001774883421771989,
      "loss": 0.1909,
      "step": 1391
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0001774561826660235,
      "loss": 0.0117,
      "step": 1392
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.00017742402315484805,
      "loss": 0.0206,
      "step": 1393
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0001773918636436726,
      "loss": 0.178,
      "step": 1394
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0001773597041324972,
      "loss": 0.1544,
      "step": 1395
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.00017732754462132177,
      "loss": 0.5376,
      "step": 1396
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.00017729538511014633,
      "loss": 0.232,
      "step": 1397
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.00017726322559897091,
      "loss": 0.3546,
      "step": 1398
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.00017723106608779547,
      "loss": 0.4072,
      "step": 1399
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.00017719890657662003,
      "loss": 0.5112,
      "step": 1400
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0001771667470654446,
      "loss": 0.5982,
      "step": 1401
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0001771345875542692,
      "loss": 0.0368,
      "step": 1402
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00017710242804309375,
      "loss": 0.3184,
      "step": 1403
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00017707026853191834,
      "loss": 0.1766,
      "step": 1404
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0001770381090207429,
      "loss": 0.0089,
      "step": 1405
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00017700594950956745,
      "loss": 0.0193,
      "step": 1406
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00017697378999839204,
      "loss": 0.0452,
      "step": 1407
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.0001769416304872166,
      "loss": 0.4775,
      "step": 1408
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00017690947097604118,
      "loss": 0.4712,
      "step": 1409
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00017687731146486573,
      "loss": 0.2158,
      "step": 1410
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00017684515195369032,
      "loss": 0.0389,
      "step": 1411
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00017681299244251487,
      "loss": 0.4897,
      "step": 1412
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00017678083293133946,
      "loss": 0.0214,
      "step": 1413
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00017674867342016402,
      "loss": 0.4355,
      "step": 1414
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0001767165139089886,
      "loss": 0.1139,
      "step": 1415
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00017668435439781316,
      "loss": 0.3818,
      "step": 1416
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00017665219488663774,
      "loss": 0.1379,
      "step": 1417
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0001766200353754623,
      "loss": 0.3631,
      "step": 1418
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00017658787586428685,
      "loss": 0.1336,
      "step": 1419
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00017655571635311144,
      "loss": 0.0152,
      "step": 1420
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.00017652355684193602,
      "loss": 0.0558,
      "step": 1421
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00017649139733076058,
      "loss": 0.0133,
      "step": 1422
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00017645923781958516,
      "loss": 0.0412,
      "step": 1423
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00017642707830840972,
      "loss": 0.6738,
      "step": 1424
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00017639491879723428,
      "loss": 0.0502,
      "step": 1425
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00017636275928605886,
      "loss": 0.4653,
      "step": 1426
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00017633059977488342,
      "loss": 0.1294,
      "step": 1427
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.000176298440263708,
      "loss": 0.1364,
      "step": 1428
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0001762662807525326,
      "loss": 0.1282,
      "step": 1429
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.00017623412124135714,
      "loss": 0.0245,
      "step": 1430
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0001762019617301817,
      "loss": 0.1396,
      "step": 1431
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.00017616980221900628,
      "loss": 0.4551,
      "step": 1432
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.00017613764270783084,
      "loss": 0.4282,
      "step": 1433
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.00017610548319665543,
      "loss": 0.2305,
      "step": 1434
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00017607332368547998,
      "loss": 0.4277,
      "step": 1435
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00017604116417430457,
      "loss": 0.2585,
      "step": 1436
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00017600900466312912,
      "loss": 0.0382,
      "step": 1437
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0001759768451519537,
      "loss": 0.0378,
      "step": 1438
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00017594468564077826,
      "loss": 0.0575,
      "step": 1439
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00017591252612960282,
      "loss": 0.394,
      "step": 1440
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0001758803666184274,
      "loss": 0.0131,
      "step": 1441
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.000175848207107252,
      "loss": 0.415,
      "step": 1442
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.00017581604759607655,
      "loss": 0.2232,
      "step": 1443
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0001757838880849011,
      "loss": 0.2194,
      "step": 1444
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0001757517285737257,
      "loss": 0.0507,
      "step": 1445
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.00017571956906255024,
      "loss": 0.0937,
      "step": 1446
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.00017568740955137483,
      "loss": 0.0722,
      "step": 1447
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0001756552500401994,
      "loss": 0.2513,
      "step": 1448
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00017562309052902397,
      "loss": 0.5186,
      "step": 1449
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00017559093101784853,
      "loss": 0.4209,
      "step": 1450
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0001755587715066731,
      "loss": 0.2663,
      "step": 1451
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00017552661199549767,
      "loss": 0.2131,
      "step": 1452
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00017549445248432222,
      "loss": 0.0486,
      "step": 1453
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.00017546229297314684,
      "loss": 0.067,
      "step": 1454
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0001754301334619714,
      "loss": 0.0178,
      "step": 1455
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00017539797395079595,
      "loss": 0.0964,
      "step": 1456
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00017536581443962053,
      "loss": 0.7441,
      "step": 1457
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.0001753336549284451,
      "loss": 0.2605,
      "step": 1458
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00017530149541726965,
      "loss": 0.8633,
      "step": 1459
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00017526933590609423,
      "loss": 0.1082,
      "step": 1460
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00017523717639491882,
      "loss": 0.0782,
      "step": 1461
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00017520501688374337,
      "loss": 0.3411,
      "step": 1462
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00017517285737256796,
      "loss": 0.0447,
      "step": 1463
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00017514069786139251,
      "loss": 0.1625,
      "step": 1464
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00017510853835021707,
      "loss": 0.3384,
      "step": 1465
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00017507637883904166,
      "loss": 0.1653,
      "step": 1466
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00017504421932786624,
      "loss": 0.2932,
      "step": 1467
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.0001750120598166908,
      "loss": 0.0236,
      "step": 1468
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00017497990030551535,
      "loss": 0.1301,
      "step": 1469
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00017494774079433994,
      "loss": 0.308,
      "step": 1470
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.0001749155812831645,
      "loss": 0.2839,
      "step": 1471
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00017488342177198905,
      "loss": 0.7676,
      "step": 1472
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00017485126226081366,
      "loss": 0.0378,
      "step": 1473
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00017481910274963822,
      "loss": 0.0452,
      "step": 1474
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00017478694323846278,
      "loss": 0.1438,
      "step": 1475
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.00017475478372728736,
      "loss": 0.0236,
      "step": 1476
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00017472262421611192,
      "loss": 0.0728,
      "step": 1477
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00017469046470493647,
      "loss": 0.1371,
      "step": 1478
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00017465830519376109,
      "loss": 0.1774,
      "step": 1479
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00017462614568258564,
      "loss": 0.0717,
      "step": 1480
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0001745939861714102,
      "loss": 0.3423,
      "step": 1481
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00017456182666023478,
      "loss": 0.0059,
      "step": 1482
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.00017452966714905934,
      "loss": 0.0123,
      "step": 1483
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0001744975076378839,
      "loss": 0.0057,
      "step": 1484
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.00017446534812670848,
      "loss": 0.1255,
      "step": 1485
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.00017443318861553307,
      "loss": 0.011,
      "step": 1486
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.00017440102910435762,
      "loss": 0.3121,
      "step": 1487
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0001743688695931822,
      "loss": 0.1583,
      "step": 1488
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.00017433671008200676,
      "loss": 0.2458,
      "step": 1489
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.00017430455057083132,
      "loss": 0.0442,
      "step": 1490
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0001742723910596559,
      "loss": 0.114,
      "step": 1491
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0001742402315484805,
      "loss": 0.3412,
      "step": 1492
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.00017420807203730505,
      "loss": 0.127,
      "step": 1493
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0001741759125261296,
      "loss": 0.5459,
      "step": 1494
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.00017414375301495419,
      "loss": 0.3782,
      "step": 1495
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.00017411159350377874,
      "loss": 0.012,
      "step": 1496
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0001740794339926033,
      "loss": 0.2576,
      "step": 1497
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.00017404727448142788,
      "loss": 0.0093,
      "step": 1498
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.00017401511497025247,
      "loss": 0.0206,
      "step": 1499
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.00017398295545907703,
      "loss": 0.3408,
      "step": 1500
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0001739507959479016,
      "loss": 0.2435,
      "step": 1501
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.00017391863643672617,
      "loss": 0.2646,
      "step": 1502
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.00017388647692555072,
      "loss": 0.427,
      "step": 1503
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0001738543174143753,
      "loss": 0.1873,
      "step": 1504
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.0001738221579031999,
      "loss": 0.2759,
      "step": 1505
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.00017378999839202445,
      "loss": 0.175,
      "step": 1506
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.00017375783888084903,
      "loss": 0.0141,
      "step": 1507
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.0001737256793696736,
      "loss": 0.4731,
      "step": 1508
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.00017369351985849815,
      "loss": 0.0511,
      "step": 1509
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.00017366136034732273,
      "loss": 0.0182,
      "step": 1510
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00017362920083614731,
      "loss": 0.1714,
      "step": 1511
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00017359704132497187,
      "loss": 0.0319,
      "step": 1512
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00017356488181379646,
      "loss": 0.4464,
      "step": 1513
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.000173532722302621,
      "loss": 0.1361,
      "step": 1514
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00017350056279144557,
      "loss": 0.0107,
      "step": 1515
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.00017346840328027015,
      "loss": 0.0139,
      "step": 1516
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.0001734362437690947,
      "loss": 0.1436,
      "step": 1517
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0001734040842579193,
      "loss": 0.1785,
      "step": 1518
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00017337192474674385,
      "loss": 0.243,
      "step": 1519
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00017333976523556844,
      "loss": 0.2253,
      "step": 1520
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.000173307605724393,
      "loss": 0.2687,
      "step": 1521
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00017327544621321755,
      "loss": 0.0188,
      "step": 1522
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00017324328670204213,
      "loss": 0.2349,
      "step": 1523
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00017321112719086672,
      "loss": 0.0208,
      "step": 1524
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00017317896767969127,
      "loss": 0.2988,
      "step": 1525
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00017314680816851586,
      "loss": 0.1503,
      "step": 1526
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00017311464865734042,
      "loss": 0.2207,
      "step": 1527
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00017308248914616497,
      "loss": 0.0321,
      "step": 1528
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00017305032963498956,
      "loss": 0.2839,
      "step": 1529
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.00017301817012381411,
      "loss": 0.0342,
      "step": 1530
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.0001729860106126387,
      "loss": 0.3098,
      "step": 1531
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00017295385110146328,
      "loss": 0.1734,
      "step": 1532
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00017292169159028784,
      "loss": 0.318,
      "step": 1533
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0001728895320791124,
      "loss": 0.0157,
      "step": 1534
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00017285737256793698,
      "loss": 0.5957,
      "step": 1535
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00017282521305676154,
      "loss": 0.3463,
      "step": 1536
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00017279305354558612,
      "loss": 0.1742,
      "step": 1537
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0001727608940344107,
      "loss": 0.0176,
      "step": 1538
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00017272873452323526,
      "loss": 0.2388,
      "step": 1539
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00017269657501205982,
      "loss": 0.028,
      "step": 1540
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0001726644155008844,
      "loss": 0.2358,
      "step": 1541
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00017263225598970896,
      "loss": 0.2746,
      "step": 1542
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00017260009647853352,
      "loss": 0.1541,
      "step": 1543
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0001725679369673581,
      "loss": 0.2676,
      "step": 1544
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.00017253577745618268,
      "loss": 0.1106,
      "step": 1545
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00017250361794500724,
      "loss": 0.208,
      "step": 1546
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.0001724714584338318,
      "loss": 0.3488,
      "step": 1547
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00017243929892265638,
      "loss": 0.0475,
      "step": 1548
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00017240713941148094,
      "loss": 0.2738,
      "step": 1549
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00017237497990030552,
      "loss": 0.0111,
      "step": 1550
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.0001723428203891301,
      "loss": 0.3933,
      "step": 1551
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00017231066087795466,
      "loss": 0.2181,
      "step": 1552
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00017227850136677922,
      "loss": 0.0124,
      "step": 1553
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0001722463418556038,
      "loss": 0.1319,
      "step": 1554
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00017221418234442836,
      "loss": 0.2083,
      "step": 1555
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00017218202283325292,
      "loss": 0.0042,
      "step": 1556
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00017214986332207753,
      "loss": 0.0425,
      "step": 1557
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0001721177038109021,
      "loss": 0.1321,
      "step": 1558
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00017208554429972664,
      "loss": 0.064,
      "step": 1559
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00017205338478855123,
      "loss": 0.4538,
      "step": 1560
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00017202122527737579,
      "loss": 0.2072,
      "step": 1561
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00017198906576620034,
      "loss": 0.0178,
      "step": 1562
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00017195690625502495,
      "loss": 0.2131,
      "step": 1563
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.0001719247467438495,
      "loss": 0.0525,
      "step": 1564
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00017189258723267407,
      "loss": 0.022,
      "step": 1565
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00017186042772149865,
      "loss": 0.3335,
      "step": 1566
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.0001718282682103232,
      "loss": 0.206,
      "step": 1567
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00017179610869914777,
      "loss": 0.2913,
      "step": 1568
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00017176394918797235,
      "loss": 0.2917,
      "step": 1569
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00017173178967679693,
      "loss": 0.0248,
      "step": 1570
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.0001716996301656215,
      "loss": 0.5918,
      "step": 1571
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00017166747065444605,
      "loss": 0.1367,
      "step": 1572
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00017163531114327063,
      "loss": 0.0558,
      "step": 1573
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0001716031516320952,
      "loss": 0.1682,
      "step": 1574
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00017157099212091977,
      "loss": 0.4746,
      "step": 1575
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00017153883260974436,
      "loss": 0.0621,
      "step": 1576
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00017150667309856891,
      "loss": 0.1,
      "step": 1577
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00017147451358739347,
      "loss": 0.4941,
      "step": 1578
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00017144235407621806,
      "loss": 0.3308,
      "step": 1579
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0001714101945650426,
      "loss": 0.3972,
      "step": 1580
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.00017137803505386717,
      "loss": 0.1355,
      "step": 1581
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.00017134587554269178,
      "loss": 0.2352,
      "step": 1582
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.00017131371603151634,
      "loss": 0.4062,
      "step": 1583
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0001712815565203409,
      "loss": 0.2715,
      "step": 1584
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.00017124939700916548,
      "loss": 0.2617,
      "step": 1585
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.00017121723749799004,
      "loss": 0.3064,
      "step": 1586
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0001711850779868146,
      "loss": 0.0356,
      "step": 1587
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00017115291847563918,
      "loss": 0.2171,
      "step": 1588
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00017112075896446376,
      "loss": 0.0757,
      "step": 1589
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00017108859945328832,
      "loss": 0.2402,
      "step": 1590
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0001710564399421129,
      "loss": 0.0119,
      "step": 1591
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00017102428043093746,
      "loss": 0.2251,
      "step": 1592
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00017099212091976202,
      "loss": 0.1385,
      "step": 1593
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0001709599614085866,
      "loss": 0.2054,
      "step": 1594
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.00017092780189741118,
      "loss": 0.1274,
      "step": 1595
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.00017089564238623574,
      "loss": 0.1765,
      "step": 1596
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0001708634828750603,
      "loss": 0.6494,
      "step": 1597
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.00017083132336388488,
      "loss": 0.0504,
      "step": 1598
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.00017079916385270944,
      "loss": 0.2953,
      "step": 1599
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.00017076700434153402,
      "loss": 0.3096,
      "step": 1600
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00017073484483035858,
      "loss": 0.0234,
      "step": 1601
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00017070268531918316,
      "loss": 0.1962,
      "step": 1602
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00017067052580800772,
      "loss": 0.1768,
      "step": 1603
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0001706383662968323,
      "loss": 0.012,
      "step": 1604
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00017060620678565686,
      "loss": 0.2014,
      "step": 1605
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.00017057404727448142,
      "loss": 0.2881,
      "step": 1606
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.000170541887763306,
      "loss": 0.158,
      "step": 1607
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.0001705097282521306,
      "loss": 0.4688,
      "step": 1608
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.00017047756874095514,
      "loss": 0.0106,
      "step": 1609
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.00017044540922977973,
      "loss": 0.008,
      "step": 1610
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.00017041324971860428,
      "loss": 0.2302,
      "step": 1611
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.00017038109020742884,
      "loss": 0.2566,
      "step": 1612
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.00017034893069625343,
      "loss": 0.0629,
      "step": 1613
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.000170316771185078,
      "loss": 0.0368,
      "step": 1614
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00017028461167390257,
      "loss": 0.2354,
      "step": 1615
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00017025245216272715,
      "loss": 0.0455,
      "step": 1616
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0001702202926515517,
      "loss": 0.0187,
      "step": 1617
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00017018813314037626,
      "loss": 0.2166,
      "step": 1618
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00017015597362920085,
      "loss": 0.1858,
      "step": 1619
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0001701238141180254,
      "loss": 0.1926,
      "step": 1620
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.00017009165460685,
      "loss": 0.0081,
      "step": 1621
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.00017005949509567455,
      "loss": 0.0862,
      "step": 1622
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.00017002733558449913,
      "loss": 0.0774,
      "step": 1623
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0001699951760733237,
      "loss": 0.204,
      "step": 1624
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.00016996301656214827,
      "loss": 0.1578,
      "step": 1625
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.00016993085705097283,
      "loss": 0.3604,
      "step": 1626
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0001698986975397974,
      "loss": 0.0543,
      "step": 1627
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.00016986653802862197,
      "loss": 0.0081,
      "step": 1628
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.00016983437851744655,
      "loss": 0.0377,
      "step": 1629
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0001698022190062711,
      "loss": 0.0327,
      "step": 1630
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.00016977005949509567,
      "loss": 0.5742,
      "step": 1631
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.00016973789998392025,
      "loss": 0.1239,
      "step": 1632
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0001697057404727448,
      "loss": 0.2592,
      "step": 1633
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0001696735809615694,
      "loss": 0.5004,
      "step": 1634
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.00016964142145039398,
      "loss": 0.051,
      "step": 1635
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.00016960926193921853,
      "loss": 0.0735,
      "step": 1636
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0001695771024280431,
      "loss": 0.3551,
      "step": 1637
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.00016954494291686767,
      "loss": 0.4069,
      "step": 1638
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.00016951278340569223,
      "loss": 0.0287,
      "step": 1639
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.00016948062389451682,
      "loss": 0.0249,
      "step": 1640
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0001694484643833414,
      "loss": 0.1749,
      "step": 1641
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.00016941630487216596,
      "loss": 0.0287,
      "step": 1642
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00016938414536099051,
      "loss": 0.5527,
      "step": 1643
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0001693519858498151,
      "loss": 0.0955,
      "step": 1644
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00016931982633863965,
      "loss": 0.0499,
      "step": 1645
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0001692876668274642,
      "loss": 0.1488,
      "step": 1646
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0001692555073162888,
      "loss": 0.0203,
      "step": 1647
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00016922334780511338,
      "loss": 0.1826,
      "step": 1648
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00016919118829393794,
      "loss": 0.0136,
      "step": 1649
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00016915902878276252,
      "loss": 0.5833,
      "step": 1650
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00016912686927158708,
      "loss": 0.1033,
      "step": 1651
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00016909470976041163,
      "loss": 0.0281,
      "step": 1652
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00016906255024923622,
      "loss": 0.0058,
      "step": 1653
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0001690303907380608,
      "loss": 0.1567,
      "step": 1654
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00016899823122688536,
      "loss": 0.2866,
      "step": 1655
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.00016896607171570992,
      "loss": 0.2544,
      "step": 1656
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.0001689339122045345,
      "loss": 0.0928,
      "step": 1657
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00016890175269335906,
      "loss": 0.008,
      "step": 1658
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00016886959318218364,
      "loss": 0.0258,
      "step": 1659
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00016883743367100823,
      "loss": 0.1749,
      "step": 1660
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00016880527415983278,
      "loss": 0.3221,
      "step": 1661
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00016877311464865734,
      "loss": 0.3672,
      "step": 1662
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00016874095513748192,
      "loss": 0.2632,
      "step": 1663
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00016870879562630648,
      "loss": 0.2772,
      "step": 1664
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00016867663611513104,
      "loss": 0.2218,
      "step": 1665
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00016864447660395565,
      "loss": 0.0139,
      "step": 1666
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.0001686123170927802,
      "loss": 0.0659,
      "step": 1667
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00016858015758160476,
      "loss": 0.0154,
      "step": 1668
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00016854799807042935,
      "loss": 0.007,
      "step": 1669
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.0001685158385592539,
      "loss": 0.0093,
      "step": 1670
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.00016848367904807846,
      "loss": 0.2729,
      "step": 1671
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.00016845151953690305,
      "loss": 0.2322,
      "step": 1672
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.00016841936002572763,
      "loss": 0.2107,
      "step": 1673
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.00016838720051455219,
      "loss": 0.0215,
      "step": 1674
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.00016835504100337677,
      "loss": 0.0486,
      "step": 1675
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.00016832288149220133,
      "loss": 0.1523,
      "step": 1676
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00016829072198102588,
      "loss": 0.0129,
      "step": 1677
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00016825856246985047,
      "loss": 0.0858,
      "step": 1678
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00016822640295867505,
      "loss": 0.4778,
      "step": 1679
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0001681942434474996,
      "loss": 0.3769,
      "step": 1680
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00016816208393632417,
      "loss": 0.3506,
      "step": 1681
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00016812992442514875,
      "loss": 0.2158,
      "step": 1682
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0001680977649139733,
      "loss": 0.0997,
      "step": 1683
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00016806560540279786,
      "loss": 0.0117,
      "step": 1684
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00016803344589162248,
      "loss": 0.2346,
      "step": 1685
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00016800128638044703,
      "loss": 0.7344,
      "step": 1686
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.0001679691268692716,
      "loss": 0.1329,
      "step": 1687
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00016793696735809617,
      "loss": 0.4749,
      "step": 1688
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00016790480784692073,
      "loss": 0.1582,
      "step": 1689
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.0001678726483357453,
      "loss": 0.2216,
      "step": 1690
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.00016784048882456987,
      "loss": 0.0297,
      "step": 1691
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.00016780832931339446,
      "loss": 0.2529,
      "step": 1692
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.000167776169802219,
      "loss": 0.0373,
      "step": 1693
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0001677440102910436,
      "loss": 0.347,
      "step": 1694
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.00016771185077986815,
      "loss": 0.1392,
      "step": 1695
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0001676796912686927,
      "loss": 0.0433,
      "step": 1696
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0001676475317575173,
      "loss": 0.0459,
      "step": 1697
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00016761537224634188,
      "loss": 0.2012,
      "step": 1698
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00016758321273516644,
      "loss": 0.0261,
      "step": 1699
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.000167551053223991,
      "loss": 0.1572,
      "step": 1700
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00016751889371281558,
      "loss": 0.3038,
      "step": 1701
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00016748673420164013,
      "loss": 0.0451,
      "step": 1702
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00016745457469046472,
      "loss": 0.0455,
      "step": 1703
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.0001674224151792893,
      "loss": 0.3909,
      "step": 1704
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.00016739025566811386,
      "loss": 0.017,
      "step": 1705
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.00016735809615693842,
      "loss": 0.4316,
      "step": 1706
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.000167325936645763,
      "loss": 0.0135,
      "step": 1707
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.00016729377713458756,
      "loss": 0.3873,
      "step": 1708
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0001672616176234121,
      "loss": 0.0713,
      "step": 1709
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0001672294581122367,
      "loss": 0.3954,
      "step": 1710
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.00016719729860106128,
      "loss": 0.171,
      "step": 1711
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00016716513908988584,
      "loss": 0.272,
      "step": 1712
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00016713297957871042,
      "loss": 0.2456,
      "step": 1713
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00016710082006753498,
      "loss": 0.3362,
      "step": 1714
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00016706866055635954,
      "loss": 0.1274,
      "step": 1715
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00016703650104518412,
      "loss": 0.0745,
      "step": 1716
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0001670043415340087,
      "loss": 0.1982,
      "step": 1717
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.00016697218202283326,
      "loss": 0.4167,
      "step": 1718
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00016694002251165785,
      "loss": 0.1379,
      "step": 1719
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0001669078630004824,
      "loss": 0.0204,
      "step": 1720
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00016687570348930696,
      "loss": 0.2314,
      "step": 1721
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00016684354397813154,
      "loss": 0.094,
      "step": 1722
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0001668113844669561,
      "loss": 0.0729,
      "step": 1723
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00016677922495578068,
      "loss": 0.0383,
      "step": 1724
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00016674706544460524,
      "loss": 0.4639,
      "step": 1725
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00016671490593342983,
      "loss": 0.2704,
      "step": 1726
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00016668274642225438,
      "loss": 0.2522,
      "step": 1727
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00016665058691107897,
      "loss": 0.1787,
      "step": 1728
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00016661842739990352,
      "loss": 0.2781,
      "step": 1729
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.0001665862678887281,
      "loss": 0.0497,
      "step": 1730
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00016655410837755266,
      "loss": 0.2639,
      "step": 1731
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.00016652194886637725,
      "loss": 0.0533,
      "step": 1732
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0001664897893552018,
      "loss": 0.0938,
      "step": 1733
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.00016645762984402636,
      "loss": 0.1753,
      "step": 1734
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.00016642547033285095,
      "loss": 0.286,
      "step": 1735
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0001663933108216755,
      "loss": 0.458,
      "step": 1736
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0001663611513105001,
      "loss": 0.2529,
      "step": 1737
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.00016632899179932467,
      "loss": 0.0336,
      "step": 1738
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.00016629683228814923,
      "loss": 0.0185,
      "step": 1739
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.00016626467277697379,
      "loss": 0.5234,
      "step": 1740
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.00016623251326579837,
      "loss": 0.3333,
      "step": 1741
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.00016620035375462293,
      "loss": 0.5117,
      "step": 1742
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0001661681942434475,
      "loss": 0.0285,
      "step": 1743
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0001661360347322721,
      "loss": 0.3501,
      "step": 1744
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.00016610387522109665,
      "loss": 0.2922,
      "step": 1745
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0001660717157099212,
      "loss": 0.0815,
      "step": 1746
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.0001660395561987458,
      "loss": 0.1758,
      "step": 1747
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.00016600739668757035,
      "loss": 0.0198,
      "step": 1748
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.00016597523717639493,
      "loss": 0.3853,
      "step": 1749
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.0001659430776652195,
      "loss": 0.2095,
      "step": 1750
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.00016591091815404407,
      "loss": 0.5811,
      "step": 1751
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.00016587875864286863,
      "loss": 0.1929,
      "step": 1752
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00016584659913169322,
      "loss": 0.1608,
      "step": 1753
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00016581443962051777,
      "loss": 0.2416,
      "step": 1754
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00016578228010934233,
      "loss": 0.4052,
      "step": 1755
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00016575012059816691,
      "loss": 0.2181,
      "step": 1756
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0001657179610869915,
      "loss": 0.3525,
      "step": 1757
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00016568580157581605,
      "loss": 0.158,
      "step": 1758
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0001656536420646406,
      "loss": 0.4443,
      "step": 1759
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.0001656214825534652,
      "loss": 0.0185,
      "step": 1760
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00016558932304228975,
      "loss": 0.192,
      "step": 1761
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00016555716353111434,
      "loss": 0.0534,
      "step": 1762
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00016552500401993892,
      "loss": 0.0596,
      "step": 1763
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00016549284450876348,
      "loss": 0.1501,
      "step": 1764
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00016546068499758803,
      "loss": 0.3413,
      "step": 1765
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.00016542852548641262,
      "loss": 0.0476,
      "step": 1766
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00016539636597523718,
      "loss": 0.0573,
      "step": 1767
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00016536420646406173,
      "loss": 0.0079,
      "step": 1768
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00016533204695288634,
      "loss": 0.4541,
      "step": 1769
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.0001652998874417109,
      "loss": 0.082,
      "step": 1770
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00016526772793053546,
      "loss": 0.0681,
      "step": 1771
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00016523556841936004,
      "loss": 0.189,
      "step": 1772
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.0001652034089081846,
      "loss": 0.1296,
      "step": 1773
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016517124939700916,
      "loss": 0.5586,
      "step": 1774
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016513908988583374,
      "loss": 0.024,
      "step": 1775
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016510693037465832,
      "loss": 0.1633,
      "step": 1776
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016507477086348288,
      "loss": 0.3184,
      "step": 1777
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016504261135230747,
      "loss": 0.4739,
      "step": 1778
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016501045184113202,
      "loss": 0.1322,
      "step": 1779
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00016497829232995658,
      "loss": 0.2874,
      "step": 1780
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.00016494613281878116,
      "loss": 0.1083,
      "step": 1781
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.00016491397330760575,
      "loss": 0.2228,
      "step": 1782
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0001648818137964303,
      "loss": 0.1611,
      "step": 1783
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.00016484965428525486,
      "loss": 0.3369,
      "step": 1784
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.00016481749477407945,
      "loss": 0.0122,
      "step": 1785
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.000164785335262904,
      "loss": 0.0317,
      "step": 1786
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.00016475317575172856,
      "loss": 0.2966,
      "step": 1787
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00016472101624055317,
      "loss": 0.146,
      "step": 1788
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00016468885672937773,
      "loss": 0.0584,
      "step": 1789
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00016465669721820228,
      "loss": 0.1643,
      "step": 1790
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00016462453770702687,
      "loss": 0.2365,
      "step": 1791
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00016459237819585143,
      "loss": 0.0609,
      "step": 1792
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00016456021868467598,
      "loss": 0.0488,
      "step": 1793
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.0001645280591735006,
      "loss": 0.5989,
      "step": 1794
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.00016449589966232515,
      "loss": 0.1776,
      "step": 1795
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0001644637401511497,
      "loss": 0.1645,
      "step": 1796
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0001644315806399743,
      "loss": 0.0588,
      "step": 1797
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.00016439942112879885,
      "loss": 0.0313,
      "step": 1798
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0001643672616176234,
      "loss": 0.2765,
      "step": 1799
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.000164335102106448,
      "loss": 0.0788,
      "step": 1800
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.00016430294259527257,
      "loss": 0.137,
      "step": 1801
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00016427078308409713,
      "loss": 0.0137,
      "step": 1802
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00016423862357292171,
      "loss": 0.4941,
      "step": 1803
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00016420646406174627,
      "loss": 0.0987,
      "step": 1804
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00016417430455057083,
      "loss": 0.0638,
      "step": 1805
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.0001641421450393954,
      "loss": 0.0332,
      "step": 1806
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00016410998552822,
      "loss": 0.0114,
      "step": 1807
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.00016407782601704455,
      "loss": 0.0179,
      "step": 1808
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0001640456665058691,
      "loss": 0.5625,
      "step": 1809
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0001640135069946937,
      "loss": 0.0345,
      "step": 1810
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.00016398134748351825,
      "loss": 0.1685,
      "step": 1811
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0001639491879723428,
      "loss": 0.1735,
      "step": 1812
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0001639170284611674,
      "loss": 0.1466,
      "step": 1813
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.00016388486894999198,
      "loss": 0.2954,
      "step": 1814
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.00016385270943881653,
      "loss": 0.1772,
      "step": 1815
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00016382054992764112,
      "loss": 0.1562,
      "step": 1816
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00016378839041646567,
      "loss": 0.1105,
      "step": 1817
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00016375623090529023,
      "loss": 0.1384,
      "step": 1818
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00016372407139411482,
      "loss": 0.2349,
      "step": 1819
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0001636919118829394,
      "loss": 0.1816,
      "step": 1820
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00016365975237176396,
      "loss": 0.2266,
      "step": 1821
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.00016362759286058854,
      "loss": 0.0424,
      "step": 1822
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0001635954333494131,
      "loss": 0.087,
      "step": 1823
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.00016356327383823765,
      "loss": 0.2419,
      "step": 1824
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.00016353111432706224,
      "loss": 0.6549,
      "step": 1825
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0001634989548158868,
      "loss": 0.1173,
      "step": 1826
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.00016346679530471138,
      "loss": 0.0299,
      "step": 1827
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.00016343463579353596,
      "loss": 0.4004,
      "step": 1828
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.00016340247628236052,
      "loss": 0.0555,
      "step": 1829
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.00016337031677118508,
      "loss": 0.0529,
      "step": 1830
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.00016333815726000966,
      "loss": 0.0267,
      "step": 1831
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.00016330599774883422,
      "loss": 0.0415,
      "step": 1832
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.0001632738382376588,
      "loss": 0.3406,
      "step": 1833
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.00016324167872648336,
      "loss": 0.5527,
      "step": 1834
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.00016320951921530794,
      "loss": 0.1747,
      "step": 1835
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.0001631773597041325,
      "loss": 0.1993,
      "step": 1836
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00016314520019295706,
      "loss": 0.0702,
      "step": 1837
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00016311304068178164,
      "loss": 0.0735,
      "step": 1838
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00016308088117060623,
      "loss": 0.2521,
      "step": 1839
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00016304872165943078,
      "loss": 0.2911,
      "step": 1840
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00016301656214825537,
      "loss": 0.0173,
      "step": 1841
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.00016298440263707992,
      "loss": 0.0466,
      "step": 1842
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.00016295224312590448,
      "loss": 0.25,
      "step": 1843
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.00016292008361472906,
      "loss": 0.0391,
      "step": 1844
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.00016288792410355362,
      "loss": 0.038,
      "step": 1845
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0001628557645923782,
      "loss": 0.1174,
      "step": 1846
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0001628236050812028,
      "loss": 0.4521,
      "step": 1847
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.00016279144557002735,
      "loss": 0.3975,
      "step": 1848
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0001627592860588519,
      "loss": 0.124,
      "step": 1849
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0001627271265476765,
      "loss": 0.0085,
      "step": 1850
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.00016269496703650104,
      "loss": 0.1329,
      "step": 1851
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.00016266280752532563,
      "loss": 0.1874,
      "step": 1852
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0001626306480141502,
      "loss": 0.3021,
      "step": 1853
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.00016259848850297477,
      "loss": 0.0092,
      "step": 1854
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.00016256632899179933,
      "loss": 0.2493,
      "step": 1855
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0001625341694806239,
      "loss": 0.2693,
      "step": 1856
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.00016250200996944847,
      "loss": 0.6055,
      "step": 1857
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.00016246985045827302,
      "loss": 0.2944,
      "step": 1858
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0001624376909470976,
      "loss": 0.0911,
      "step": 1859
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0001624055314359222,
      "loss": 0.048,
      "step": 1860
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.00016237337192474675,
      "loss": 0.2705,
      "step": 1861
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0001623412124135713,
      "loss": 0.2197,
      "step": 1862
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0001623090529023959,
      "loss": 0.5763,
      "step": 1863
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016227689339122045,
      "loss": 0.207,
      "step": 1864
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016224473388004503,
      "loss": 0.3257,
      "step": 1865
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016221257436886962,
      "loss": 0.0339,
      "step": 1866
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016218041485769417,
      "loss": 0.293,
      "step": 1867
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016214825534651873,
      "loss": 0.3452,
      "step": 1868
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016211609583534331,
      "loss": 0.4609,
      "step": 1869
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00016208393632416787,
      "loss": 0.2551,
      "step": 1870
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00016205177681299243,
      "loss": 0.1295,
      "step": 1871
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00016201961730181704,
      "loss": 0.0521,
      "step": 1872
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.0001619874577906416,
      "loss": 0.1946,
      "step": 1873
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00016195529827946615,
      "loss": 0.0767,
      "step": 1874
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00016192313876829074,
      "loss": 0.2735,
      "step": 1875
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.0001618909792571153,
      "loss": 0.2793,
      "step": 1876
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.00016185881974593985,
      "loss": 0.1774,
      "step": 1877
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016182666023476446,
      "loss": 0.2725,
      "step": 1878
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016179450072358902,
      "loss": 0.0097,
      "step": 1879
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016176234121241358,
      "loss": 0.1431,
      "step": 1880
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016173018170123816,
      "loss": 0.3501,
      "step": 1881
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016169802219006272,
      "loss": 0.3003,
      "step": 1882
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016166586267888727,
      "loss": 0.2848,
      "step": 1883
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00016163370316771186,
      "loss": 0.0099,
      "step": 1884
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00016160154365653644,
      "loss": 0.122,
      "step": 1885
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.000161569384145361,
      "loss": 0.4888,
      "step": 1886
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00016153722463418556,
      "loss": 0.0296,
      "step": 1887
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00016150506512301014,
      "loss": 0.3306,
      "step": 1888
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.0001614729056118347,
      "loss": 0.3263,
      "step": 1889
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00016144074610065928,
      "loss": 0.1064,
      "step": 1890
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00016140858658948387,
      "loss": 0.4111,
      "step": 1891
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00016137642707830842,
      "loss": 0.0181,
      "step": 1892
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00016134426756713298,
      "loss": 0.2513,
      "step": 1893
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00016131210805595756,
      "loss": 0.3185,
      "step": 1894
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00016127994854478212,
      "loss": 0.1161,
      "step": 1895
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00016124778903360668,
      "loss": 0.1246,
      "step": 1896
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.0001612156295224313,
      "loss": 0.0745,
      "step": 1897
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.00016118347001125585,
      "loss": 0.5488,
      "step": 1898
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0001611513105000804,
      "loss": 0.0399,
      "step": 1899
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.00016111915098890499,
      "loss": 0.3871,
      "step": 1900
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.00016108699147772954,
      "loss": 0.6504,
      "step": 1901
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0001610548319665541,
      "loss": 0.0786,
      "step": 1902
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.00016102267245537868,
      "loss": 0.0549,
      "step": 1903
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.00016099051294420327,
      "loss": 0.3979,
      "step": 1904
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.00016095835343302783,
      "loss": 0.7207,
      "step": 1905
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0001609261939218524,
      "loss": 0.061,
      "step": 1906
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.00016089403441067697,
      "loss": 0.2457,
      "step": 1907
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.00016086187489950152,
      "loss": 0.21,
      "step": 1908
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0001608297153883261,
      "loss": 0.1514,
      "step": 1909
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0001607975558771507,
      "loss": 0.3848,
      "step": 1910
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.00016076539636597525,
      "loss": 0.2281,
      "step": 1911
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0001607332368547998,
      "loss": 0.3379,
      "step": 1912
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0001607010773436244,
      "loss": 0.2444,
      "step": 1913
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.00016066891783244895,
      "loss": 0.3932,
      "step": 1914
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.00016063675832127353,
      "loss": 0.0188,
      "step": 1915
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0001606045988100981,
      "loss": 0.1398,
      "step": 1916
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.00016057243929892267,
      "loss": 0.377,
      "step": 1917
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.00016054027978774723,
      "loss": 0.074,
      "step": 1918
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0001605081202765718,
      "loss": 0.1083,
      "step": 1919
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.00016047596076539637,
      "loss": 0.2512,
      "step": 1920
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.00016044380125422093,
      "loss": 0.1392,
      "step": 1921
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0001604116417430455,
      "loss": 0.0454,
      "step": 1922
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0001603794822318701,
      "loss": 0.0328,
      "step": 1923
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.00016034732272069465,
      "loss": 0.1243,
      "step": 1924
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.00016031516320951924,
      "loss": 0.5527,
      "step": 1925
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.0001602830036983438,
      "loss": 0.4385,
      "step": 1926
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.00016025084418716835,
      "loss": 0.0394,
      "step": 1927
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.00016021868467599293,
      "loss": 0.2795,
      "step": 1928
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.0001601865251648175,
      "loss": 0.3065,
      "step": 1929
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.00016015436565364207,
      "loss": 0.0449,
      "step": 1930
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.00016012220614246666,
      "loss": 0.0168,
      "step": 1931
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.00016009004663129122,
      "loss": 0.0139,
      "step": 1932
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00016005788712011577,
      "loss": 0.013,
      "step": 1933
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00016002572760894036,
      "loss": 0.7988,
      "step": 1934
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00015999356809776491,
      "loss": 0.2734,
      "step": 1935
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0001599614085865895,
      "loss": 0.1261,
      "step": 1936
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00015992924907541405,
      "loss": 0.019,
      "step": 1937
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00015989708956423864,
      "loss": 0.0402,
      "step": 1938
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0001598649300530632,
      "loss": 0.3096,
      "step": 1939
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015983277054188778,
      "loss": 0.0116,
      "step": 1940
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015980061103071234,
      "loss": 0.0453,
      "step": 1941
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015976845151953692,
      "loss": 0.2426,
      "step": 1942
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015973629200836148,
      "loss": 0.0579,
      "step": 1943
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015970413249718606,
      "loss": 0.2295,
      "step": 1944
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015967197298601062,
      "loss": 0.2825,
      "step": 1945
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00015963981347483518,
      "loss": 0.4873,
      "step": 1946
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00015960765396365976,
      "loss": 0.5137,
      "step": 1947
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00015957549445248432,
      "loss": 0.2691,
      "step": 1948
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.0001595433349413089,
      "loss": 0.0272,
      "step": 1949
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00015951117543013348,
      "loss": 0.113,
      "step": 1950
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00015947901591895804,
      "loss": 0.2876,
      "step": 1951
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.0001594468564077826,
      "loss": 0.3877,
      "step": 1952
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.00015941469689660718,
      "loss": 0.3262,
      "step": 1953
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00015938253738543174,
      "loss": 0.2032,
      "step": 1954
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00015935037787425632,
      "loss": 0.2339,
      "step": 1955
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0001593182183630809,
      "loss": 0.3164,
      "step": 1956
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00015928605885190546,
      "loss": 0.1805,
      "step": 1957
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00015925389934073002,
      "loss": 0.0377,
      "step": 1958
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0001592217398295546,
      "loss": 0.1407,
      "step": 1959
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00015918958031837916,
      "loss": 0.1158,
      "step": 1960
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00015915742080720372,
      "loss": 0.3525,
      "step": 1961
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0001591252612960283,
      "loss": 0.1221,
      "step": 1962
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0001590931017848529,
      "loss": 0.0285,
      "step": 1963
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00015906094227367744,
      "loss": 0.0173,
      "step": 1964
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00015902878276250203,
      "loss": 0.8516,
      "step": 1965
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00015899662325132659,
      "loss": 0.0987,
      "step": 1966
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00015896446374015114,
      "loss": 0.02,
      "step": 1967
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.00015893230422897573,
      "loss": 0.1711,
      "step": 1968
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.0001589001447178003,
      "loss": 0.3748,
      "step": 1969
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.00015886798520662487,
      "loss": 0.5713,
      "step": 1970
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.00015883582569544942,
      "loss": 0.1522,
      "step": 1971
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.000158803666184274,
      "loss": 0.2998,
      "step": 1972
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.00015877150667309857,
      "loss": 0.1785,
      "step": 1973
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.00015873934716192312,
      "loss": 0.0141,
      "step": 1974
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00015870718765074773,
      "loss": 0.4346,
      "step": 1975
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.0001586750281395723,
      "loss": 0.0159,
      "step": 1976
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00015864286862839685,
      "loss": 0.0084,
      "step": 1977
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00015861070911722143,
      "loss": 0.4219,
      "step": 1978
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.000158578549606046,
      "loss": 0.1133,
      "step": 1979
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00015854639009487055,
      "loss": 0.0104,
      "step": 1980
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00015851423058369516,
      "loss": 0.0461,
      "step": 1981
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00015848207107251971,
      "loss": 0.3641,
      "step": 1982
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00015844991156134427,
      "loss": 0.194,
      "step": 1983
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00015841775205016886,
      "loss": 0.0982,
      "step": 1984
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.0001583855925389934,
      "loss": 0.5241,
      "step": 1985
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00015835343302781797,
      "loss": 0.0928,
      "step": 1986
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00015832127351664255,
      "loss": 0.0137,
      "step": 1987
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.00015828911400546714,
      "loss": 0.2435,
      "step": 1988
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.0001582569544942917,
      "loss": 0.037,
      "step": 1989
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.00015822479498311625,
      "loss": 0.216,
      "step": 1990
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.00015819263547194084,
      "loss": 0.4902,
      "step": 1991
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.0001581604759607654,
      "loss": 0.0137,
      "step": 1992
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.00015812831644958998,
      "loss": 0.026,
      "step": 1993
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.00015809615693841456,
      "loss": 0.0464,
      "step": 1994
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015806399742723912,
      "loss": 0.0305,
      "step": 1995
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015803183791606367,
      "loss": 0.2808,
      "step": 1996
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015799967840488826,
      "loss": 0.3354,
      "step": 1997
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015796751889371282,
      "loss": 0.0187,
      "step": 1998
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015793535938253737,
      "loss": 0.0598,
      "step": 1999
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015790319987136198,
      "loss": 0.1438,
      "step": 2000
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00015787104036018654,
      "loss": 0.0119,
      "step": 2001
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0001578388808490111,
      "loss": 0.017,
      "step": 2002
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.00015780672133783568,
      "loss": 0.3076,
      "step": 2003
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.00015777456182666024,
      "loss": 0.0094,
      "step": 2004
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0001577424023154848,
      "loss": 0.3273,
      "step": 2005
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.00015771024280430938,
      "loss": 0.4968,
      "step": 2006
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.00015767808329313396,
      "loss": 0.2585,
      "step": 2007
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.00015764592378195852,
      "loss": 0.1998,
      "step": 2008
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0001576137642707831,
      "loss": 0.0549,
      "step": 2009
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.00015758160475960766,
      "loss": 0.0504,
      "step": 2010
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.00015754944524843222,
      "loss": 0.0888,
      "step": 2011
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0001575172857372568,
      "loss": 0.2949,
      "step": 2012
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0001574851262260814,
      "loss": 0.2532,
      "step": 2013
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.00015745296671490594,
      "loss": 0.2325,
      "step": 2014
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0001574208072037305,
      "loss": 0.0394,
      "step": 2015
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00015738864769255508,
      "loss": 0.252,
      "step": 2016
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00015735648818137964,
      "loss": 0.1149,
      "step": 2017
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00015732432867020423,
      "loss": 0.0218,
      "step": 2018
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00015729216915902878,
      "loss": 0.4497,
      "step": 2019
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00015726000964785337,
      "loss": 0.2221,
      "step": 2020
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00015722785013667792,
      "loss": 0.1936,
      "step": 2021
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0001571956906255025,
      "loss": 0.065,
      "step": 2022
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.00015716353111432706,
      "loss": 0.1867,
      "step": 2023
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.00015713137160315162,
      "loss": 0.1152,
      "step": 2024
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0001570992120919762,
      "loss": 0.0612,
      "step": 2025
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0001570670525808008,
      "loss": 0.0903,
      "step": 2026
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.00015703489306962535,
      "loss": 0.0605,
      "step": 2027
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.00015700273355844993,
      "loss": 0.0685,
      "step": 2028
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0001569705740472745,
      "loss": 0.2529,
      "step": 2029
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.00015693841453609904,
      "loss": 0.3046,
      "step": 2030
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.00015690625502492363,
      "loss": 0.1602,
      "step": 2031
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0001568740955137482,
      "loss": 0.4082,
      "step": 2032
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.00015684193600257277,
      "loss": 0.0274,
      "step": 2033
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.00015680977649139735,
      "loss": 0.0233,
      "step": 2034
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0001567776169802219,
      "loss": 0.4481,
      "step": 2035
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.00015674545746904647,
      "loss": 0.3701,
      "step": 2036
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.00015671329795787105,
      "loss": 0.0446,
      "step": 2037
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0001566811384466956,
      "loss": 0.5459,
      "step": 2038
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0001566489789355202,
      "loss": 0.4189,
      "step": 2039
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.00015661681942434475,
      "loss": 0.1179,
      "step": 2040
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.00015658465991316933,
      "loss": 0.205,
      "step": 2041
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0001565525004019939,
      "loss": 0.2819,
      "step": 2042
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.00015652034089081847,
      "loss": 0.2593,
      "step": 2043
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015648818137964303,
      "loss": 0.5488,
      "step": 2044
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015645602186846762,
      "loss": 0.2734,
      "step": 2045
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015642386235729217,
      "loss": 0.1224,
      "step": 2046
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015639170284611676,
      "loss": 0.5303,
      "step": 2047
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015635954333494131,
      "loss": 0.2375,
      "step": 2048
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015632738382376587,
      "loss": 0.2517,
      "step": 2049
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00015629522431259045,
      "loss": 0.2017,
      "step": 2050
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.000156263064801415,
      "loss": 0.2212,
      "step": 2051
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.0001562309052902396,
      "loss": 0.3268,
      "step": 2052
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.00015619874577906418,
      "loss": 0.0243,
      "step": 2053
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.00015616658626788874,
      "loss": 0.3549,
      "step": 2054
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.0001561344267567133,
      "loss": 0.1644,
      "step": 2055
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.00015610226724553788,
      "loss": 0.4922,
      "step": 2056
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.00015607010773436243,
      "loss": 0.583,
      "step": 2057
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00015603794822318702,
      "loss": 0.2036,
      "step": 2058
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0001560057887120116,
      "loss": 0.2704,
      "step": 2059
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00015597362920083616,
      "loss": 0.0529,
      "step": 2060
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00015594146968966072,
      "loss": 0.4307,
      "step": 2061
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0001559093101784853,
      "loss": 0.0286,
      "step": 2062
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00015587715066730986,
      "loss": 0.3091,
      "step": 2063
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.00015584499115613441,
      "loss": 0.2855,
      "step": 2064
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.000155812831644959,
      "loss": 0.5498,
      "step": 2065
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00015578067213378358,
      "loss": 0.0745,
      "step": 2066
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00015574851262260814,
      "loss": 0.015,
      "step": 2067
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00015571635311143272,
      "loss": 0.2207,
      "step": 2068
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00015568419360025728,
      "loss": 0.0756,
      "step": 2069
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00015565203408908184,
      "loss": 0.157,
      "step": 2070
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00015561987457790642,
      "loss": 0.0165,
      "step": 2071
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.000155587715066731,
      "loss": 0.3521,
      "step": 2072
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00015555555555555556,
      "loss": 0.0047,
      "step": 2073
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00015552339604438012,
      "loss": 0.6113,
      "step": 2074
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0001554912365332047,
      "loss": 0.02,
      "step": 2075
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00015545907702202926,
      "loss": 0.3555,
      "step": 2076
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00015542691751085385,
      "loss": 0.1415,
      "step": 2077
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015539475799967843,
      "loss": 0.183,
      "step": 2078
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015536259848850299,
      "loss": 0.12,
      "step": 2079
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015533043897732754,
      "loss": 0.0127,
      "step": 2080
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015529827946615213,
      "loss": 0.0058,
      "step": 2081
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015526611995497668,
      "loss": 0.1831,
      "step": 2082
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015523396044380124,
      "loss": 0.2087,
      "step": 2083
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.00015520180093262585,
      "loss": 0.0527,
      "step": 2084
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0001551696414214504,
      "loss": 0.0683,
      "step": 2085
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.00015513748191027497,
      "loss": 0.0721,
      "step": 2086
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.00015510532239909955,
      "loss": 0.1737,
      "step": 2087
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0001550731628879241,
      "loss": 0.061,
      "step": 2088
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.00015504100337674866,
      "loss": 0.0925,
      "step": 2089
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.00015500884386557325,
      "loss": 0.0186,
      "step": 2090
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.00015497668435439783,
      "loss": 0.1965,
      "step": 2091
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0001549445248432224,
      "loss": 0.0649,
      "step": 2092
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.00015491236533204697,
      "loss": 0.0434,
      "step": 2093
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.00015488020582087153,
      "loss": 0.0138,
      "step": 2094
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0001548480463096961,
      "loss": 0.1797,
      "step": 2095
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.00015481588679852067,
      "loss": 0.1164,
      "step": 2096
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.00015478372728734526,
      "loss": 0.0092,
      "step": 2097
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0001547515677761698,
      "loss": 0.0776,
      "step": 2098
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00015471940826499437,
      "loss": 0.1714,
      "step": 2099
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00015468724875381895,
      "loss": 0.1658,
      "step": 2100
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0001546550892426435,
      "loss": 0.0601,
      "step": 2101
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00015462292973146807,
      "loss": 0.0217,
      "step": 2102
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00015459077022029268,
      "loss": 0.0279,
      "step": 2103
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.00015455861070911724,
      "loss": 0.124,
      "step": 2104
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0001545264511979418,
      "loss": 0.0349,
      "step": 2105
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.00015449429168676638,
      "loss": 0.2842,
      "step": 2106
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.00015446213217559093,
      "loss": 0.085,
      "step": 2107
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0001544299726644155,
      "loss": 0.251,
      "step": 2108
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.00015439781315324007,
      "loss": 0.0368,
      "step": 2109
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.00015436565364206466,
      "loss": 0.271,
      "step": 2110
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.00015433349413088922,
      "loss": 0.0139,
      "step": 2111
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0001543013346197138,
      "loss": 0.1753,
      "step": 2112
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00015426917510853836,
      "loss": 0.1248,
      "step": 2113
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.0001542370155973629,
      "loss": 0.1971,
      "step": 2114
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.0001542048560861875,
      "loss": 0.2129,
      "step": 2115
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00015417269657501208,
      "loss": 0.0497,
      "step": 2116
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00015414053706383664,
      "loss": 0.1599,
      "step": 2117
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00015410837755266122,
      "loss": 0.2656,
      "step": 2118
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.00015407621804148578,
      "loss": 0.025,
      "step": 2119
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.00015404405853031034,
      "loss": 0.1426,
      "step": 2120
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.00015401189901913492,
      "loss": 0.012,
      "step": 2121
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0001539797395079595,
      "loss": 0.097,
      "step": 2122
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.00015394757999678406,
      "loss": 0.0129,
      "step": 2123
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.00015391542048560862,
      "loss": 0.0309,
      "step": 2124
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0001538832609744332,
      "loss": 0.0213,
      "step": 2125
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.00015385110146325776,
      "loss": 0.0036,
      "step": 2126
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00015381894195208232,
      "loss": 0.2964,
      "step": 2127
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.0001537867824409069,
      "loss": 0.2764,
      "step": 2128
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00015375462292973148,
      "loss": 0.0239,
      "step": 2129
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00015372246341855604,
      "loss": 0.0202,
      "step": 2130
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00015369030390738063,
      "loss": 0.0266,
      "step": 2131
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00015365814439620518,
      "loss": 0.1533,
      "step": 2132
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.00015362598488502974,
      "loss": 0.2759,
      "step": 2133
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00015359382537385432,
      "loss": 0.2373,
      "step": 2134
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.0001535616658626789,
      "loss": 0.0333,
      "step": 2135
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00015352950635150346,
      "loss": 0.0421,
      "step": 2136
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00015349734684032805,
      "loss": 0.097,
      "step": 2137
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.0001534651873291526,
      "loss": 0.0835,
      "step": 2138
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00015343302781797716,
      "loss": 0.1277,
      "step": 2139
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.00015340086830680175,
      "loss": 0.334,
      "step": 2140
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0001533687087956263,
      "loss": 0.2192,
      "step": 2141
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0001533365492844509,
      "loss": 0.056,
      "step": 2142
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.00015330438977327547,
      "loss": 0.073,
      "step": 2143
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.00015327223026210003,
      "loss": 0.1313,
      "step": 2144
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.00015324007075092459,
      "loss": 0.102,
      "step": 2145
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.00015320791123974917,
      "loss": 0.1021,
      "step": 2146
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.00015317575172857373,
      "loss": 0.0301,
      "step": 2147
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0001531435922173983,
      "loss": 0.1536,
      "step": 2148
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.00015311143270622287,
      "loss": 0.0132,
      "step": 2149
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.00015307927319504745,
      "loss": 0.319,
      "step": 2150
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.000153047113683872,
      "loss": 0.1758,
      "step": 2151
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.00015301495417269657,
      "loss": 0.1602,
      "step": 2152
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.00015298279466152115,
      "loss": 0.0742,
      "step": 2153
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0001529506351503457,
      "loss": 0.0083,
      "step": 2154
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.0001529184756391703,
      "loss": 0.0659,
      "step": 2155
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.00015288631612799487,
      "loss": 0.1758,
      "step": 2156
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.00015285415661681943,
      "loss": 0.1928,
      "step": 2157
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.000152821997105644,
      "loss": 0.0133,
      "step": 2158
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.00015278983759446857,
      "loss": 0.1567,
      "step": 2159
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.00015275767808329313,
      "loss": 0.2162,
      "step": 2160
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.00015272551857211771,
      "loss": 0.252,
      "step": 2161
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0001526933590609423,
      "loss": 0.0325,
      "step": 2162
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.00015266119954976685,
      "loss": 0.016,
      "step": 2163
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0001526290400385914,
      "loss": 0.4502,
      "step": 2164
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.000152596880527416,
      "loss": 0.0421,
      "step": 2165
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.00015256472101624055,
      "loss": 0.112,
      "step": 2166
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.00015253256150506514,
      "loss": 0.0369,
      "step": 2167
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015250040199388972,
      "loss": 0.3315,
      "step": 2168
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015246824248271428,
      "loss": 0.0422,
      "step": 2169
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015243608297153883,
      "loss": 0.0102,
      "step": 2170
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015240392346036342,
      "loss": 0.007,
      "step": 2171
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015237176394918798,
      "loss": 0.1807,
      "step": 2172
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015233960443801253,
      "loss": 0.2046,
      "step": 2173
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.00015230744492683712,
      "loss": 0.0137,
      "step": 2174
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0001522752854156617,
      "loss": 0.2129,
      "step": 2175
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00015224312590448626,
      "loss": 0.0042,
      "step": 2176
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00015221096639331081,
      "loss": 0.1875,
      "step": 2177
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0001521788068821354,
      "loss": 0.1357,
      "step": 2178
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00015214664737095996,
      "loss": 0.0496,
      "step": 2179
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00015211448785978454,
      "loss": 0.2524,
      "step": 2180
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.00015208232834860912,
      "loss": 0.3174,
      "step": 2181
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00015205016883743368,
      "loss": 0.2973,
      "step": 2182
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00015201800932625824,
      "loss": 0.2039,
      "step": 2183
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00015198584981508282,
      "loss": 0.0242,
      "step": 2184
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00015195369030390738,
      "loss": 0.1917,
      "step": 2185
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00015192153079273194,
      "loss": 0.0186,
      "step": 2186
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.00015188937128155655,
      "loss": 0.0708,
      "step": 2187
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.0001518572117703811,
      "loss": 0.0344,
      "step": 2188
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00015182505225920566,
      "loss": 0.0535,
      "step": 2189
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00015179289274803025,
      "loss": 0.1399,
      "step": 2190
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.0001517607332368548,
      "loss": 0.0433,
      "step": 2191
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00015172857372567936,
      "loss": 0.2455,
      "step": 2192
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00015169641421450397,
      "loss": 0.259,
      "step": 2193
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00015166425470332853,
      "loss": 0.0286,
      "step": 2194
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.00015163209519215308,
      "loss": 0.1496,
      "step": 2195
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00015159993568097767,
      "loss": 0.1963,
      "step": 2196
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00015156777616980223,
      "loss": 0.0679,
      "step": 2197
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00015153561665862678,
      "loss": 0.0514,
      "step": 2198
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00015150345714745137,
      "loss": 0.1929,
      "step": 2199
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00015147129763627595,
      "loss": 0.2537,
      "step": 2200
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.0001514391381251005,
      "loss": 0.0592,
      "step": 2201
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.00015140697861392506,
      "loss": 0.0889,
      "step": 2202
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.00015137481910274965,
      "loss": 0.0225,
      "step": 2203
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0001513426595915742,
      "loss": 0.1571,
      "step": 2204
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0001513105000803988,
      "loss": 0.0239,
      "step": 2205
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.00015127834056922337,
      "loss": 0.0629,
      "step": 2206
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.00015124618105804793,
      "loss": 0.0769,
      "step": 2207
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0001512140215468725,
      "loss": 0.1362,
      "step": 2208
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.00015118186203569707,
      "loss": 0.0548,
      "step": 2209
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.00015114970252452163,
      "loss": 0.0979,
      "step": 2210
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.00015111754301334619,
      "loss": 0.2207,
      "step": 2211
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0001510853835021708,
      "loss": 0.1075,
      "step": 2212
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.00015105322399099535,
      "loss": 0.1743,
      "step": 2213
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0001510210644798199,
      "loss": 0.1801,
      "step": 2214
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0001509889049686445,
      "loss": 0.2422,
      "step": 2215
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.00015095674545746905,
      "loss": 0.0664,
      "step": 2216
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0001509245859462936,
      "loss": 0.1568,
      "step": 2217
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0001508924264351182,
      "loss": 0.0115,
      "step": 2218
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.00015086026692394278,
      "loss": 0.1263,
      "step": 2219
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.00015082810741276733,
      "loss": 0.2305,
      "step": 2220
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.00015079594790159192,
      "loss": 0.0483,
      "step": 2221
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.00015076378839041647,
      "loss": 0.0881,
      "step": 2222
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.00015073162887924103,
      "loss": 0.2249,
      "step": 2223
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.00015069946936806562,
      "loss": 0.2402,
      "step": 2224
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0001506673098568902,
      "loss": 0.0283,
      "step": 2225
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.00015063515034571476,
      "loss": 0.252,
      "step": 2226
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0001506029908345393,
      "loss": 0.1022,
      "step": 2227
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0001505708313233639,
      "loss": 0.1093,
      "step": 2228
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.00015053867181218845,
      "loss": 0.1129,
      "step": 2229
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.00015050651230101304,
      "loss": 0.0269,
      "step": 2230
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.0001504743527898376,
      "loss": 0.1677,
      "step": 2231
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00015044219327866218,
      "loss": 0.0752,
      "step": 2232
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00015041003376748674,
      "loss": 0.0987,
      "step": 2233
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00015037787425631132,
      "loss": 0.3164,
      "step": 2234
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00015034571474513588,
      "loss": 0.147,
      "step": 2235
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00015031355523396043,
      "loss": 0.1199,
      "step": 2236
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.00015028139572278502,
      "loss": 0.1531,
      "step": 2237
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0001502492362116096,
      "loss": 0.0872,
      "step": 2238
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.00015021707670043416,
      "loss": 0.1255,
      "step": 2239
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.00015018491718925874,
      "loss": 0.0168,
      "step": 2240
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0001501527576780833,
      "loss": 0.1337,
      "step": 2241
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.00015012059816690786,
      "loss": 0.0651,
      "step": 2242
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.00015008843865573244,
      "loss": 0.0089,
      "step": 2243
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.000150056279144557,
      "loss": 0.1321,
      "step": 2244
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.00015002411963338158,
      "loss": 0.0795,
      "step": 2245
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.00014999196012220617,
      "loss": 0.12,
      "step": 2246
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.00014995980061103072,
      "loss": 0.0311,
      "step": 2247
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.00014992764109985528,
      "loss": 0.167,
      "step": 2248
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.00014989548158867986,
      "loss": 0.1797,
      "step": 2249
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.00014986332207750442,
      "loss": 0.0106,
      "step": 2250
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.000149831162566329,
      "loss": 0.1895,
      "step": 2251
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.00014979900305515356,
      "loss": 0.1045,
      "step": 2252
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.00014976684354397815,
      "loss": 0.1655,
      "step": 2253
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0001497346840328027,
      "loss": 0.0066,
      "step": 2254
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.00014970252452162726,
      "loss": 0.1371,
      "step": 2255
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.00014967036501045184,
      "loss": 0.3135,
      "step": 2256
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0001496382054992764,
      "loss": 0.1536,
      "step": 2257
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00014960604598810099,
      "loss": 0.2502,
      "step": 2258
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00014957388647692557,
      "loss": 0.127,
      "step": 2259
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00014954172696575013,
      "loss": 0.0455,
      "step": 2260
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00014950956745457468,
      "loss": 0.0283,
      "step": 2261
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00014947740794339927,
      "loss": 0.0894,
      "step": 2262
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.00014944524843222382,
      "loss": 0.1685,
      "step": 2263
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.0001494130889210484,
      "loss": 0.2177,
      "step": 2264
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.000149380929409873,
      "loss": 0.1078,
      "step": 2265
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.00014934876989869755,
      "loss": 0.2177,
      "step": 2266
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0001493166103875221,
      "loss": 0.0515,
      "step": 2267
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0001492844508763467,
      "loss": 0.0804,
      "step": 2268
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.00014925229136517125,
      "loss": 0.1871,
      "step": 2269
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.00014922013185399583,
      "loss": 0.1134,
      "step": 2270
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.00014918797234282042,
      "loss": 0.3008,
      "step": 2271
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.00014915581283164497,
      "loss": 0.0806,
      "step": 2272
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.00014912365332046953,
      "loss": 0.2502,
      "step": 2273
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.00014909149380929411,
      "loss": 0.0444,
      "step": 2274
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.00014905933429811867,
      "loss": 0.0165,
      "step": 2275
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.00014902717478694323,
      "loss": 0.1357,
      "step": 2276
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.0001489950152757678,
      "loss": 0.0804,
      "step": 2277
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.0001489628557645924,
      "loss": 0.083,
      "step": 2278
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.00014893069625341695,
      "loss": 0.2651,
      "step": 2279
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0001488985367422415,
      "loss": 0.0897,
      "step": 2280
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0001488663772310661,
      "loss": 0.2312,
      "step": 2281
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.00014883421771989065,
      "loss": 0.1382,
      "step": 2282
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.00014880205820871524,
      "loss": 0.1403,
      "step": 2283
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.00014876989869753982,
      "loss": 0.0115,
      "step": 2284
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.00014873773918636438,
      "loss": 0.1068,
      "step": 2285
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00014870557967518893,
      "loss": 0.2568,
      "step": 2286
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00014867342016401352,
      "loss": 0.1916,
      "step": 2287
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00014864126065283807,
      "loss": 0.0672,
      "step": 2288
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00014860910114166263,
      "loss": 0.2488,
      "step": 2289
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00014857694163048724,
      "loss": 0.1151,
      "step": 2290
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.0001485447821193118,
      "loss": 0.2559,
      "step": 2291
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.00014851262260813636,
      "loss": 0.0839,
      "step": 2292
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00014848046309696094,
      "loss": 0.2759,
      "step": 2293
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.0001484483035857855,
      "loss": 0.0214,
      "step": 2294
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00014841614407461005,
      "loss": 0.2426,
      "step": 2295
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00014838398456343467,
      "loss": 0.0507,
      "step": 2296
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00014835182505225922,
      "loss": 0.0613,
      "step": 2297
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00014831966554108378,
      "loss": 0.1235,
      "step": 2298
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.00014828750602990836,
      "loss": 0.0665,
      "step": 2299
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.00014825534651873292,
      "loss": 0.1319,
      "step": 2300
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.00014822318700755748,
      "loss": 0.0274,
      "step": 2301
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.00014819102749638206,
      "loss": 0.2178,
      "step": 2302
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.00014815886798520665,
      "loss": 0.239,
      "step": 2303
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0001481267084740312,
      "loss": 0.1039,
      "step": 2304
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.00014809454896285576,
      "loss": 0.0701,
      "step": 2305
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.00014806238945168034,
      "loss": 0.2457,
      "step": 2306
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.0001480302299405049,
      "loss": 0.017,
      "step": 2307
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00014799807042932948,
      "loss": 0.061,
      "step": 2308
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00014796591091815407,
      "loss": 0.25,
      "step": 2309
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00014793375140697863,
      "loss": 0.0047,
      "step": 2310
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00014790159189580318,
      "loss": 0.1229,
      "step": 2311
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00014786943238462777,
      "loss": 0.1846,
      "step": 2312
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.00014783727287345232,
      "loss": 0.144,
      "step": 2313
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.00014780511336227688,
      "loss": 0.3208,
      "step": 2314
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0001477729538511015,
      "loss": 0.3176,
      "step": 2315
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.00014774079433992605,
      "loss": 0.1576,
      "step": 2316
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0001477086348287506,
      "loss": 0.2107,
      "step": 2317
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0001476764753175752,
      "loss": 0.105,
      "step": 2318
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.00014764431580639975,
      "loss": 0.0339,
      "step": 2319
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0001476121562952243,
      "loss": 0.046,
      "step": 2320
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0001475799967840489,
      "loss": 0.0342,
      "step": 2321
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.00014754783727287347,
      "loss": 0.1483,
      "step": 2322
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.00014751567776169803,
      "loss": 0.2056,
      "step": 2323
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0001474835182505226,
      "loss": 0.215,
      "step": 2324
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.00014745135873934717,
      "loss": 0.0714,
      "step": 2325
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.00014741919922817173,
      "loss": 0.2939,
      "step": 2326
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0001473870397169963,
      "loss": 0.0181,
      "step": 2327
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0001473548802058209,
      "loss": 0.049,
      "step": 2328
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.00014732272069464545,
      "loss": 0.1611,
      "step": 2329
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.00014729056118347,
      "loss": 0.1524,
      "step": 2330
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0001472584016722946,
      "loss": 0.1208,
      "step": 2331
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.00014722624216111915,
      "loss": 0.0706,
      "step": 2332
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.00014719408264994373,
      "loss": 0.0977,
      "step": 2333
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.0001471619231387683,
      "loss": 0.2532,
      "step": 2334
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.00014712976362759287,
      "loss": 0.0044,
      "step": 2335
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.00014709760411641743,
      "loss": 0.0038,
      "step": 2336
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.00014706544460524202,
      "loss": 0.3496,
      "step": 2337
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.00014703328509406657,
      "loss": 0.1411,
      "step": 2338
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.00014700112558289113,
      "loss": 0.0596,
      "step": 2339
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.00014696896607171571,
      "loss": 0.1354,
      "step": 2340
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0001469368065605403,
      "loss": 0.0434,
      "step": 2341
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.00014690464704936485,
      "loss": 0.2293,
      "step": 2342
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.00014687248753818944,
      "loss": 0.103,
      "step": 2343
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.000146840328027014,
      "loss": 0.113,
      "step": 2344
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.00014680816851583855,
      "loss": 0.1689,
      "step": 2345
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.00014677600900466314,
      "loss": 0.095,
      "step": 2346
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0001467438494934877,
      "loss": 0.1616,
      "step": 2347
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00014671168998231228,
      "loss": 0.1036,
      "step": 2348
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00014667953047113686,
      "loss": 0.025,
      "step": 2349
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00014664737095996142,
      "loss": 0.1931,
      "step": 2350
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00014661521144878598,
      "loss": 0.0475,
      "step": 2351
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00014658305193761056,
      "loss": 0.0693,
      "step": 2352
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.00014655089242643512,
      "loss": 0.0176,
      "step": 2353
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.0001465187329152597,
      "loss": 0.0444,
      "step": 2354
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00014648657340408426,
      "loss": 0.1219,
      "step": 2355
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00014645441389290884,
      "loss": 0.083,
      "step": 2356
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.0001464222543817334,
      "loss": 0.048,
      "step": 2357
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00014639009487055798,
      "loss": 0.0598,
      "step": 2358
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00014635793535938254,
      "loss": 0.2192,
      "step": 2359
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00014632577584820712,
      "loss": 0.0372,
      "step": 2360
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.00014629361633703168,
      "loss": 0.0825,
      "step": 2361
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00014626145682585626,
      "loss": 0.2793,
      "step": 2362
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00014622929731468082,
      "loss": 0.0414,
      "step": 2363
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00014619713780350538,
      "loss": 0.2427,
      "step": 2364
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00014616497829232996,
      "loss": 0.0898,
      "step": 2365
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.00014613281878115452,
      "loss": 0.1904,
      "step": 2366
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0001461006592699791,
      "loss": 0.0643,
      "step": 2367
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0001460684997588037,
      "loss": 0.129,
      "step": 2368
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.00014603634024762824,
      "loss": 0.3325,
      "step": 2369
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0001460041807364528,
      "loss": 0.3403,
      "step": 2370
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.00014597202122527739,
      "loss": 0.0083,
      "step": 2371
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.00014593986171410194,
      "loss": 0.0469,
      "step": 2372
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.00014590770220292653,
      "loss": 0.153,
      "step": 2373
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0001458755426917511,
      "loss": 0.1052,
      "step": 2374
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.00014584338318057567,
      "loss": 0.2611,
      "step": 2375
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.00014581122366940022,
      "loss": 0.4355,
      "step": 2376
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0001457790641582248,
      "loss": 0.0771,
      "step": 2377
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.00014574690464704937,
      "loss": 0.0085,
      "step": 2378
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.00014571474513587392,
      "loss": 0.1448,
      "step": 2379
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0001456825856246985,
      "loss": 0.1914,
      "step": 2380
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0001456504261135231,
      "loss": 0.0477,
      "step": 2381
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.00014561826660234765,
      "loss": 0.1934,
      "step": 2382
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00014558610709117223,
      "loss": 0.1792,
      "step": 2383
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.0001455539475799968,
      "loss": 0.05,
      "step": 2384
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00014552178806882135,
      "loss": 0.1281,
      "step": 2385
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00014548962855764593,
      "loss": 0.2778,
      "step": 2386
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00014545746904647051,
      "loss": 0.0759,
      "step": 2387
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00014542530953529507,
      "loss": 0.0454,
      "step": 2388
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.00014539315002411963,
      "loss": 0.1728,
      "step": 2389
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0001453609905129442,
      "loss": 0.0707,
      "step": 2390
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00014532883100176877,
      "loss": 0.1086,
      "step": 2391
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00014529667149059333,
      "loss": 0.0909,
      "step": 2392
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00014526451197941794,
      "loss": 0.0963,
      "step": 2393
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0001452323524682425,
      "loss": 0.194,
      "step": 2394
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00014520019295706705,
      "loss": 0.012,
      "step": 2395
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.00014516803344589164,
      "loss": 0.0109,
      "step": 2396
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.0001451358739347162,
      "loss": 0.078,
      "step": 2397
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.00014510371442354075,
      "loss": 0.0479,
      "step": 2398
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.00014507155491236536,
      "loss": 0.1395,
      "step": 2399
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.00014503939540118992,
      "loss": 0.1588,
      "step": 2400
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.00014500723589001447,
      "loss": 0.0574,
      "step": 2401
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.00014497507637883906,
      "loss": 0.1456,
      "step": 2402
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00014494291686766362,
      "loss": 0.1139,
      "step": 2403
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00014491075735648817,
      "loss": 0.1553,
      "step": 2404
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00014487859784531276,
      "loss": 0.066,
      "step": 2405
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00014484643833413734,
      "loss": 0.0324,
      "step": 2406
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.0001448142788229619,
      "loss": 0.0213,
      "step": 2407
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00014478211931178648,
      "loss": 0.3984,
      "step": 2408
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.00014474995980061104,
      "loss": 0.0739,
      "step": 2409
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.0001447178002894356,
      "loss": 0.1646,
      "step": 2410
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00014468564077826018,
      "loss": 0.14,
      "step": 2411
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00014465348126708476,
      "loss": 0.1953,
      "step": 2412
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00014462132175590932,
      "loss": 0.0742,
      "step": 2413
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00014458916224473388,
      "loss": 0.165,
      "step": 2414
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00014455700273355846,
      "loss": 0.0695,
      "step": 2415
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.00014452484322238302,
      "loss": 0.0086,
      "step": 2416
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00014449268371120758,
      "loss": 0.0104,
      "step": 2417
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00014446052420003219,
      "loss": 0.2061,
      "step": 2418
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00014442836468885674,
      "loss": 0.0661,
      "step": 2419
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.0001443962051776813,
      "loss": 0.1223,
      "step": 2420
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00014436404566650588,
      "loss": 0.1023,
      "step": 2421
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.00014433188615533044,
      "loss": 0.2482,
      "step": 2422
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.000144299726644155,
      "loss": 0.0273,
      "step": 2423
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.00014426756713297958,
      "loss": 0.0416,
      "step": 2424
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.00014423540762180417,
      "loss": 0.109,
      "step": 2425
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.00014420324811062872,
      "loss": 0.0549,
      "step": 2426
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0001441710885994533,
      "loss": 0.117,
      "step": 2427
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.00014413892908827786,
      "loss": 0.2542,
      "step": 2428
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.00014410676957710242,
      "loss": 0.1802,
      "step": 2429
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.000144074610065927,
      "loss": 0.282,
      "step": 2430
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0001440424505547516,
      "loss": 0.1406,
      "step": 2431
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.00014401029104357615,
      "loss": 0.2694,
      "step": 2432
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.00014397813153240073,
      "loss": 0.1134,
      "step": 2433
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0001439459720212253,
      "loss": 0.1541,
      "step": 2434
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.00014391381251004984,
      "loss": 0.1364,
      "step": 2435
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.00014388165299887443,
      "loss": 0.014,
      "step": 2436
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.00014384949348769899,
      "loss": 0.129,
      "step": 2437
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.00014381733397652357,
      "loss": 0.0565,
      "step": 2438
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.00014378517446534813,
      "loss": 0.397,
      "step": 2439
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0001437530149541727,
      "loss": 0.1969,
      "step": 2440
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.00014372085544299727,
      "loss": 0.0239,
      "step": 2441
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.00014368869593182182,
      "loss": 0.2764,
      "step": 2442
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0001436565364206464,
      "loss": 0.024,
      "step": 2443
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.000143624376909471,
      "loss": 0.1076,
      "step": 2444
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00014359221739829555,
      "loss": 0.0045,
      "step": 2445
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00014356005788712013,
      "loss": 0.1488,
      "step": 2446
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.0001435278983759447,
      "loss": 0.1208,
      "step": 2447
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00014349573886476925,
      "loss": 0.225,
      "step": 2448
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00014346357935359383,
      "loss": 0.092,
      "step": 2449
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00014343141984241842,
      "loss": 0.1454,
      "step": 2450
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.00014339926033124297,
      "loss": 0.0631,
      "step": 2451
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.00014336710082006756,
      "loss": 0.2091,
      "step": 2452
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.00014333494130889211,
      "loss": 0.1223,
      "step": 2453
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.00014330278179771667,
      "loss": 0.0366,
      "step": 2454
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.00014327062228654125,
      "loss": 0.0574,
      "step": 2455
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0001432384627753658,
      "loss": 0.2998,
      "step": 2456
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0001432063032641904,
      "loss": 0.0232,
      "step": 2457
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.00014317414375301498,
      "loss": 0.0276,
      "step": 2458
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00014314198424183954,
      "loss": 0.025,
      "step": 2459
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.0001431098247306641,
      "loss": 0.0738,
      "step": 2460
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00014307766521948868,
      "loss": 0.0136,
      "step": 2461
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00014304550570831323,
      "loss": 0.0225,
      "step": 2462
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00014301334619713782,
      "loss": 0.2814,
      "step": 2463
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00014298118668596238,
      "loss": 0.1664,
      "step": 2464
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.00014294902717478696,
      "loss": 0.249,
      "step": 2465
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00014291686766361152,
      "loss": 0.0632,
      "step": 2466
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00014288470815243607,
      "loss": 0.0151,
      "step": 2467
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00014285254864126066,
      "loss": 0.1273,
      "step": 2468
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00014282038913008521,
      "loss": 0.1478,
      "step": 2469
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.0001427882296189098,
      "loss": 0.081,
      "step": 2470
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00014275607010773438,
      "loss": 0.2267,
      "step": 2471
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.00014272391059655894,
      "loss": 0.2105,
      "step": 2472
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.0001426917510853835,
      "loss": 0.1763,
      "step": 2473
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.00014265959157420808,
      "loss": 0.2033,
      "step": 2474
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.00014262743206303264,
      "loss": 0.2751,
      "step": 2475
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.00014259527255185722,
      "loss": 0.2607,
      "step": 2476
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.0001425631130406818,
      "loss": 0.0376,
      "step": 2477
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.00014253095352950636,
      "loss": 0.0496,
      "step": 2478
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.00014249879401833092,
      "loss": 0.3325,
      "step": 2479
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.0001424666345071555,
      "loss": 0.0139,
      "step": 2480
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00014243447499598006,
      "loss": 0.1357,
      "step": 2481
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00014240231548480462,
      "loss": 0.0439,
      "step": 2482
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00014237015597362923,
      "loss": 0.3174,
      "step": 2483
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00014233799646245379,
      "loss": 0.0307,
      "step": 2484
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.00014230583695127834,
      "loss": 0.1641,
      "step": 2485
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00014227367744010293,
      "loss": 0.1063,
      "step": 2486
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00014224151792892748,
      "loss": 0.1777,
      "step": 2487
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00014220935841775204,
      "loss": 0.0412,
      "step": 2488
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00014217719890657663,
      "loss": 0.1049,
      "step": 2489
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.0001421450393954012,
      "loss": 0.0356,
      "step": 2490
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00014211287988422577,
      "loss": 0.0391,
      "step": 2491
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.00014208072037305032,
      "loss": 0.0954,
      "step": 2492
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.0001420485608618749,
      "loss": 0.0277,
      "step": 2493
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00014201640135069946,
      "loss": 0.1258,
      "step": 2494
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00014198424183952405,
      "loss": 0.1572,
      "step": 2495
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00014195208232834863,
      "loss": 0.0323,
      "step": 2496
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.0001419199228171732,
      "loss": 0.0321,
      "step": 2497
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00014188776330599775,
      "loss": 0.1465,
      "step": 2498
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.00014185560379482233,
      "loss": 0.0353,
      "step": 2499
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0001418234442836469,
      "loss": 0.0626,
      "step": 2500
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.00014179128477247144,
      "loss": 0.0428,
      "step": 2501
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.00014175912526129606,
      "loss": 0.0217,
      "step": 2502
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0001417269657501206,
      "loss": 0.1519,
      "step": 2503
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.00014169480623894517,
      "loss": 0.1953,
      "step": 2504
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.00014166264672776975,
      "loss": 0.2417,
      "step": 2505
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0001416304872165943,
      "loss": 0.0337,
      "step": 2506
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.00014159832770541887,
      "loss": 0.1387,
      "step": 2507
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.00014156616819424345,
      "loss": 0.2074,
      "step": 2508
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.00014153400868306804,
      "loss": 0.0554,
      "step": 2509
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0001415018491718926,
      "loss": 0.3789,
      "step": 2510
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.00014146968966071718,
      "loss": 0.0323,
      "step": 2511
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.00014143753014954173,
      "loss": 0.5098,
      "step": 2512
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0001414053706383663,
      "loss": 0.1738,
      "step": 2513
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.00014137321112719087,
      "loss": 0.0403,
      "step": 2514
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.00014134105161601546,
      "loss": 0.141,
      "step": 2515
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.00014130889210484002,
      "loss": 0.1518,
      "step": 2516
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.00014127673259366457,
      "loss": 0.2151,
      "step": 2517
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.00014124457308248916,
      "loss": 0.015,
      "step": 2518
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0001412124135713137,
      "loss": 0.0858,
      "step": 2519
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0001411802540601383,
      "loss": 0.1735,
      "step": 2520
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.00014114809454896288,
      "loss": 0.1978,
      "step": 2521
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.00014111593503778744,
      "loss": 0.0113,
      "step": 2522
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.000141083775526612,
      "loss": 0.003,
      "step": 2523
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.00014105161601543658,
      "loss": 0.0547,
      "step": 2524
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.00014101945650426114,
      "loss": 0.0878,
      "step": 2525
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.0001409872969930857,
      "loss": 0.3672,
      "step": 2526
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.00014095513748191028,
      "loss": 0.2368,
      "step": 2527
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.00014092297797073486,
      "loss": 0.2212,
      "step": 2528
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.00014089081845955942,
      "loss": 0.007,
      "step": 2529
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.000140858658948384,
      "loss": 0.2922,
      "step": 2530
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.00014082649943720856,
      "loss": 0.0599,
      "step": 2531
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.00014079433992603312,
      "loss": 0.2712,
      "step": 2532
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0001407621804148577,
      "loss": 0.0172,
      "step": 2533
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.00014073002090368228,
      "loss": 0.2765,
      "step": 2534
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00014069786139250684,
      "loss": 0.0279,
      "step": 2535
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00014066570188133143,
      "loss": 0.0845,
      "step": 2536
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00014063354237015598,
      "loss": 0.2998,
      "step": 2537
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00014060138285898054,
      "loss": 0.2195,
      "step": 2538
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00014056922334780512,
      "loss": 0.3013,
      "step": 2539
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.0001405370638366297,
      "loss": 0.2228,
      "step": 2540
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.00014050490432545426,
      "loss": 0.1082,
      "step": 2541
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.00014047274481427882,
      "loss": 0.0708,
      "step": 2542
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0001404405853031034,
      "loss": 0.0754,
      "step": 2543
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.00014040842579192796,
      "loss": 0.1901,
      "step": 2544
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.00014037626628075252,
      "loss": 0.1733,
      "step": 2545
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0001403441067695771,
      "loss": 0.121,
      "step": 2546
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0001403119472584017,
      "loss": 0.2085,
      "step": 2547
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.00014027978774722624,
      "loss": 0.1233,
      "step": 2548
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00014024762823605083,
      "loss": 0.0298,
      "step": 2549
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00014021546872487539,
      "loss": 0.2759,
      "step": 2550
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00014018330921369994,
      "loss": 0.1448,
      "step": 2551
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00014015114970252453,
      "loss": 0.1897,
      "step": 2552
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.0001401189901913491,
      "loss": 0.0074,
      "step": 2553
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00014008683068017367,
      "loss": 0.0731,
      "step": 2554
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.00014005467116899825,
      "loss": 0.0331,
      "step": 2555
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0001400225116578228,
      "loss": 0.0887,
      "step": 2556
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.00013999035214664737,
      "loss": 0.3906,
      "step": 2557
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.00013995819263547195,
      "loss": 0.0381,
      "step": 2558
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0001399260331242965,
      "loss": 0.1776,
      "step": 2559
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0001398938736131211,
      "loss": 0.1151,
      "step": 2560
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.00013986171410194567,
      "loss": 0.0053,
      "step": 2561
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.00013982955459077023,
      "loss": 0.3486,
      "step": 2562
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0001397973950795948,
      "loss": 0.2056,
      "step": 2563
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.00013976523556841937,
      "loss": 0.103,
      "step": 2564
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.00013973307605724393,
      "loss": 0.0137,
      "step": 2565
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.00013970091654606851,
      "loss": 0.111,
      "step": 2566
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.00013966875703489307,
      "loss": 0.0178,
      "step": 2567
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.00013963659752371765,
      "loss": 0.0841,
      "step": 2568
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0001396044380125422,
      "loss": 0.293,
      "step": 2569
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.00013957227850136677,
      "loss": 0.2329,
      "step": 2570
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.00013954011899019135,
      "loss": 0.0773,
      "step": 2571
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0001395079594790159,
      "loss": 0.2578,
      "step": 2572
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0001394757999678405,
      "loss": 0.052,
      "step": 2573
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.00013944364045666508,
      "loss": 0.0087,
      "step": 2574
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.00013941148094548963,
      "loss": 0.3438,
      "step": 2575
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.0001393793214343142,
      "loss": 0.2544,
      "step": 2576
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00013934716192313878,
      "loss": 0.1897,
      "step": 2577
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00013931500241196333,
      "loss": 0.1146,
      "step": 2578
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00013928284290078792,
      "loss": 0.0984,
      "step": 2579
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.0001392506833896125,
      "loss": 0.074,
      "step": 2580
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00013921852387843706,
      "loss": 0.3896,
      "step": 2581
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.00013918636436726161,
      "loss": 0.0078,
      "step": 2582
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0001391542048560862,
      "loss": 0.0721,
      "step": 2583
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.00013912204534491076,
      "loss": 0.0859,
      "step": 2584
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.00013908988583373534,
      "loss": 0.1468,
      "step": 2585
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.00013905772632255992,
      "loss": 0.0035,
      "step": 2586
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.00013902556681138448,
      "loss": 0.0611,
      "step": 2587
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.00013899340730020904,
      "loss": 0.1444,
      "step": 2588
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.00013896124778903362,
      "loss": 0.1697,
      "step": 2589
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00013892908827785818,
      "loss": 0.1718,
      "step": 2590
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00013889692876668274,
      "loss": 0.0135,
      "step": 2591
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00013886476925550732,
      "loss": 0.0086,
      "step": 2592
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0001388326097443319,
      "loss": 0.2661,
      "step": 2593
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00013880045023315646,
      "loss": 0.0434,
      "step": 2594
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.00013876829072198102,
      "loss": 0.2144,
      "step": 2595
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0001387361312108056,
      "loss": 0.1537,
      "step": 2596
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013870397169963016,
      "loss": 0.0781,
      "step": 2597
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013867181218845474,
      "loss": 0.0903,
      "step": 2598
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013863965267727933,
      "loss": 0.0192,
      "step": 2599
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013860749316610388,
      "loss": 0.2544,
      "step": 2600
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013857533365492844,
      "loss": 0.3428,
      "step": 2601
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013854317414375303,
      "loss": 0.2545,
      "step": 2602
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.00013851101463257758,
      "loss": 0.0393,
      "step": 2603
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.00013847885512140214,
      "loss": 0.064,
      "step": 2604
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.00013844669561022675,
      "loss": 0.0376,
      "step": 2605
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0001384145360990513,
      "loss": 0.0293,
      "step": 2606
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.00013838237658787586,
      "loss": 0.4336,
      "step": 2607
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.00013835021707670045,
      "loss": 0.1915,
      "step": 2608
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.000138318057565525,
      "loss": 0.161,
      "step": 2609
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.00013828589805434956,
      "loss": 0.0375,
      "step": 2610
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.00013825373854317417,
      "loss": 0.1117,
      "step": 2611
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.00013822157903199873,
      "loss": 0.1504,
      "step": 2612
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0001381894195208233,
      "loss": 0.2334,
      "step": 2613
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.00013815726000964787,
      "loss": 0.2437,
      "step": 2614
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.00013812510049847243,
      "loss": 0.1284,
      "step": 2615
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.00013809294098729699,
      "loss": 0.1729,
      "step": 2616
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.00013806078147612157,
      "loss": 0.017,
      "step": 2617
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.00013802862196494615,
      "loss": 0.0602,
      "step": 2618
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0001379964624537707,
      "loss": 0.0143,
      "step": 2619
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.00013796430294259527,
      "loss": 0.1401,
      "step": 2620
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.00013793214343141985,
      "loss": 0.3111,
      "step": 2621
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0001378999839202444,
      "loss": 0.2465,
      "step": 2622
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.000137867824409069,
      "loss": 0.1692,
      "step": 2623
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.00013783566489789358,
      "loss": 0.1254,
      "step": 2624
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.00013780350538671813,
      "loss": 0.2031,
      "step": 2625
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0001377713458755427,
      "loss": 0.1836,
      "step": 2626
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.00013773918636436727,
      "loss": 0.0214,
      "step": 2627
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.00013770702685319183,
      "loss": 0.3643,
      "step": 2628
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0001376748673420164,
      "loss": 0.0388,
      "step": 2629
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.00013764270783084097,
      "loss": 0.135,
      "step": 2630
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.00013761054831966556,
      "loss": 0.1223,
      "step": 2631
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0001375783888084901,
      "loss": 0.0628,
      "step": 2632
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0001375462292973147,
      "loss": 0.0469,
      "step": 2633
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.00013751406978613925,
      "loss": 0.0178,
      "step": 2634
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0001374819102749638,
      "loss": 0.0163,
      "step": 2635
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0001374497507637884,
      "loss": 0.158,
      "step": 2636
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.00013741759125261298,
      "loss": 0.2888,
      "step": 2637
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.00013738543174143754,
      "loss": 0.0836,
      "step": 2638
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.00013735327223026212,
      "loss": 0.2397,
      "step": 2639
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.00013732111271908668,
      "loss": 0.4287,
      "step": 2640
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.00013728895320791123,
      "loss": 0.2916,
      "step": 2641
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.00013725679369673582,
      "loss": 0.0308,
      "step": 2642
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.0001372246341855604,
      "loss": 0.1259,
      "step": 2643
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.00013719247467438496,
      "loss": 0.3623,
      "step": 2644
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.00013716031516320952,
      "loss": 0.2725,
      "step": 2645
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0001371281556520341,
      "loss": 0.1414,
      "step": 2646
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.00013709599614085866,
      "loss": 0.1588,
      "step": 2647
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.00013706383662968324,
      "loss": 0.1205,
      "step": 2648
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0001370316771185078,
      "loss": 0.1737,
      "step": 2649
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.00013699951760733238,
      "loss": 0.0642,
      "step": 2650
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.00013696735809615694,
      "loss": 0.0712,
      "step": 2651
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00013693519858498152,
      "loss": 0.0101,
      "step": 2652
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00013690303907380608,
      "loss": 0.2517,
      "step": 2653
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00013687087956263064,
      "loss": 0.0915,
      "step": 2654
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00013683872005145522,
      "loss": 0.0109,
      "step": 2655
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0001368065605402798,
      "loss": 0.08,
      "step": 2656
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00013677440102910436,
      "loss": 0.1438,
      "step": 2657
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.00013674224151792895,
      "loss": 0.0546,
      "step": 2658
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.0001367100820067535,
      "loss": 0.2761,
      "step": 2659
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.00013667792249557806,
      "loss": 0.1412,
      "step": 2660
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.00013664576298440264,
      "loss": 0.3975,
      "step": 2661
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.0001366136034732272,
      "loss": 0.1272,
      "step": 2662
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.00013658144396205179,
      "loss": 0.0806,
      "step": 2663
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.00013654928445087637,
      "loss": 0.1996,
      "step": 2664
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.00013651712493970093,
      "loss": 0.1963,
      "step": 2665
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00013648496542852548,
      "loss": 0.043,
      "step": 2666
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00013645280591735007,
      "loss": 0.0738,
      "step": 2667
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00013642064640617462,
      "loss": 0.2234,
      "step": 2668
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0001363884868949992,
      "loss": 0.2429,
      "step": 2669
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00013635632738382377,
      "loss": 0.1432,
      "step": 2670
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.00013632416787264835,
      "loss": 0.1307,
      "step": 2671
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0001362920083614729,
      "loss": 0.0442,
      "step": 2672
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0001362598488502975,
      "loss": 0.1423,
      "step": 2673
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.00013622768933912205,
      "loss": 0.0301,
      "step": 2674
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0001361955298279466,
      "loss": 0.1074,
      "step": 2675
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0001361633703167712,
      "loss": 0.1718,
      "step": 2676
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.00013613121080559577,
      "loss": 0.1346,
      "step": 2677
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.00013609905129442033,
      "loss": 0.2354,
      "step": 2678
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0001360668917832449,
      "loss": 0.0293,
      "step": 2679
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.00013603473227206947,
      "loss": 0.2172,
      "step": 2680
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.00013600257276089403,
      "loss": 0.0549,
      "step": 2681
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0001359704132497186,
      "loss": 0.061,
      "step": 2682
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0001359382537385432,
      "loss": 0.2646,
      "step": 2683
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.00013590609422736775,
      "loss": 0.2695,
      "step": 2684
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0001358739347161923,
      "loss": 0.0085,
      "step": 2685
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0001358417752050169,
      "loss": 0.0308,
      "step": 2686
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013580961569384145,
      "loss": 0.0105,
      "step": 2687
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013577745618266604,
      "loss": 0.15,
      "step": 2688
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013574529667149062,
      "loss": 0.0389,
      "step": 2689
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013571313716031518,
      "loss": 0.1042,
      "step": 2690
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013568097764913973,
      "loss": 0.5225,
      "step": 2691
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013564881813796432,
      "loss": 0.3247,
      "step": 2692
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.00013561665862678887,
      "loss": 0.2188,
      "step": 2693
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.00013558449911561343,
      "loss": 0.1084,
      "step": 2694
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.00013555233960443802,
      "loss": 0.0742,
      "step": 2695
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0001355201800932626,
      "loss": 0.1572,
      "step": 2696
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.00013548802058208716,
      "loss": 0.022,
      "step": 2697
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.00013545586107091174,
      "loss": 0.0238,
      "step": 2698
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0001354237015597363,
      "loss": 0.0622,
      "step": 2699
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.00013539154204856085,
      "loss": 0.189,
      "step": 2700
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013535938253738544,
      "loss": 0.0329,
      "step": 2701
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013532722302621002,
      "loss": 0.1896,
      "step": 2702
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013529506351503458,
      "loss": 0.2559,
      "step": 2703
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013526290400385914,
      "loss": 0.0652,
      "step": 2704
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013523074449268372,
      "loss": 0.0107,
      "step": 2705
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013519858498150828,
      "loss": 0.0195,
      "step": 2706
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.00013516642547033283,
      "loss": 0.2856,
      "step": 2707
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.00013513426595915745,
      "loss": 0.313,
      "step": 2708
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.000135102106447982,
      "loss": 0.0421,
      "step": 2709
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.00013506994693680656,
      "loss": 0.1102,
      "step": 2710
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.00013503778742563114,
      "loss": 0.1168,
      "step": 2711
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0001350056279144557,
      "loss": 0.144,
      "step": 2712
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.00013497346840328026,
      "loss": 0.0374,
      "step": 2713
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.00013494130889210487,
      "loss": 0.0984,
      "step": 2714
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013490914938092943,
      "loss": 0.0265,
      "step": 2715
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013487698986975398,
      "loss": 0.0994,
      "step": 2716
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013484483035857857,
      "loss": 0.0246,
      "step": 2717
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013481267084740312,
      "loss": 0.1948,
      "step": 2718
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013478051133622768,
      "loss": 0.1774,
      "step": 2719
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013474835182505226,
      "loss": 0.0889,
      "step": 2720
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.00013471619231387685,
      "loss": 0.2098,
      "step": 2721
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0001346840328027014,
      "loss": 0.1232,
      "step": 2722
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.000134651873291526,
      "loss": 0.1697,
      "step": 2723
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.00013461971378035055,
      "loss": 0.0459,
      "step": 2724
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0001345875542691751,
      "loss": 0.1804,
      "step": 2725
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0001345553947579997,
      "loss": 0.1339,
      "step": 2726
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.00013452323524682427,
      "loss": 0.2424,
      "step": 2727
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.00013449107573564883,
      "loss": 0.0598,
      "step": 2728
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.00013445891622447339,
      "loss": 0.1699,
      "step": 2729
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.00013442675671329797,
      "loss": 0.1,
      "step": 2730
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.00013439459720212253,
      "loss": 0.2554,
      "step": 2731
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.00013436243769094708,
      "loss": 0.2988,
      "step": 2732
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0001343302781797717,
      "loss": 0.2961,
      "step": 2733
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.00013429811866859625,
      "loss": 0.0289,
      "step": 2734
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0001342659591574208,
      "loss": 0.0309,
      "step": 2735
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0001342337996462454,
      "loss": 0.0965,
      "step": 2736
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.00013420164013506995,
      "loss": 0.088,
      "step": 2737
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0001341694806238945,
      "loss": 0.1633,
      "step": 2738
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0001341373211127191,
      "loss": 0.0501,
      "step": 2739
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.00013410516160154367,
      "loss": 0.0489,
      "step": 2740
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.00013407300209036823,
      "loss": 0.2336,
      "step": 2741
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00013404084257919282,
      "loss": 0.1605,
      "step": 2742
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00013400868306801737,
      "loss": 0.0494,
      "step": 2743
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00013397652355684193,
      "loss": 0.1282,
      "step": 2744
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0001339443640456665,
      "loss": 0.0048,
      "step": 2745
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0001339122045344911,
      "loss": 0.0749,
      "step": 2746
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00013388004502331565,
      "loss": 0.0113,
      "step": 2747
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.00013384788551214024,
      "loss": 0.1257,
      "step": 2748
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0001338157260009648,
      "loss": 0.1504,
      "step": 2749
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.00013378356648978935,
      "loss": 0.1152,
      "step": 2750
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.00013375140697861394,
      "loss": 0.0057,
      "step": 2751
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0001337192474674385,
      "loss": 0.0202,
      "step": 2752
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.00013368708795626308,
      "loss": 0.1018,
      "step": 2753
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.00013365492844508763,
      "loss": 0.0255,
      "step": 2754
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.00013362276893391222,
      "loss": 0.0334,
      "step": 2755
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.00013359060942273678,
      "loss": 0.0118,
      "step": 2756
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.00013355844991156133,
      "loss": 0.1422,
      "step": 2757
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.00013352629040038592,
      "loss": 0.1011,
      "step": 2758
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0001334941308892105,
      "loss": 0.1716,
      "step": 2759
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.00013346197137803506,
      "loss": 0.1792,
      "step": 2760
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.00013342981186685964,
      "loss": 0.0273,
      "step": 2761
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0001333976523556842,
      "loss": 0.0128,
      "step": 2762
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00013336549284450876,
      "loss": 0.035,
      "step": 2763
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00013333333333333334,
      "loss": 0.023,
      "step": 2764
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0001333011738221579,
      "loss": 0.06,
      "step": 2765
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00013326901431098248,
      "loss": 0.1799,
      "step": 2766
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00013323685479980706,
      "loss": 0.0686,
      "step": 2767
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00013320469528863162,
      "loss": 0.1179,
      "step": 2768
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.00013317253577745618,
      "loss": 0.0988,
      "step": 2769
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.00013314037626628076,
      "loss": 0.0282,
      "step": 2770
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.00013310821675510532,
      "loss": 0.146,
      "step": 2771
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0001330760572439299,
      "loss": 0.0303,
      "step": 2772
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0001330438977327545,
      "loss": 0.0592,
      "step": 2773
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.00013301173822157904,
      "loss": 0.0385,
      "step": 2774
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0001329795787104036,
      "loss": 0.0346,
      "step": 2775
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.00013294741919922819,
      "loss": 0.0203,
      "step": 2776
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013291525968805274,
      "loss": 0.0518,
      "step": 2777
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013288310017687733,
      "loss": 0.0085,
      "step": 2778
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013285094066570188,
      "loss": 0.0712,
      "step": 2779
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013281878115452647,
      "loss": 0.0099,
      "step": 2780
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013278662164335102,
      "loss": 0.0529,
      "step": 2781
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013275446213217558,
      "loss": 0.0137,
      "step": 2782
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.00013272230262100017,
      "loss": 0.0247,
      "step": 2783
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.00013269014310982472,
      "loss": 0.157,
      "step": 2784
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0001326579835986493,
      "loss": 0.0727,
      "step": 2785
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0001326258240874739,
      "loss": 0.0894,
      "step": 2786
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.00013259366457629845,
      "loss": 0.0852,
      "step": 2787
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.000132561505065123,
      "loss": 0.0137,
      "step": 2788
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0001325293455539476,
      "loss": 0.0555,
      "step": 2789
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.00013249718604277215,
      "loss": 0.0683,
      "step": 2790
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00013246502653159673,
      "loss": 0.146,
      "step": 2791
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00013243286702042131,
      "loss": 0.1023,
      "step": 2792
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00013240070750924587,
      "loss": 0.1013,
      "step": 2793
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00013236854799807043,
      "loss": 0.0159,
      "step": 2794
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.000132336388486895,
      "loss": 0.1562,
      "step": 2795
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00013230422897571957,
      "loss": 0.0916,
      "step": 2796
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.00013227206946454413,
      "loss": 0.0656,
      "step": 2797
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.0001322399099533687,
      "loss": 0.065,
      "step": 2798
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.0001322077504421933,
      "loss": 0.0615,
      "step": 2799
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.00013217559093101785,
      "loss": 0.1175,
      "step": 2800
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.00013214343141984244,
      "loss": 0.1191,
      "step": 2801
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.000132111271908667,
      "loss": 0.2065,
      "step": 2802
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.00013207911239749155,
      "loss": 0.0631,
      "step": 2803
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.00013204695288631613,
      "loss": 0.0476,
      "step": 2804
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00013201479337514072,
      "loss": 0.1284,
      "step": 2805
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00013198263386396527,
      "loss": 0.1031,
      "step": 2806
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00013195047435278983,
      "loss": 0.0455,
      "step": 2807
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00013191831484161442,
      "loss": 0.0962,
      "step": 2808
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00013188615533043897,
      "loss": 0.0145,
      "step": 2809
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.00013185399581926356,
      "loss": 0.0927,
      "step": 2810
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.00013182183630808814,
      "loss": 0.1124,
      "step": 2811
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0001317896767969127,
      "loss": 0.0252,
      "step": 2812
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.00013175751728573725,
      "loss": 0.0813,
      "step": 2813
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.00013172535777456184,
      "loss": 0.0279,
      "step": 2814
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0001316931982633864,
      "loss": 0.1022,
      "step": 2815
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.00013166103875221095,
      "loss": 0.149,
      "step": 2816
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.00013162887924103556,
      "loss": 0.0593,
      "step": 2817
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013159671972986012,
      "loss": 0.026,
      "step": 2818
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013156456021868468,
      "loss": 0.0951,
      "step": 2819
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013153240070750926,
      "loss": 0.0235,
      "step": 2820
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013150024119633382,
      "loss": 0.0885,
      "step": 2821
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013146808168515838,
      "loss": 0.0973,
      "step": 2822
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013143592217398296,
      "loss": 0.011,
      "step": 2823
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.00013140376266280754,
      "loss": 0.1362,
      "step": 2824
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0001313716031516321,
      "loss": 0.1348,
      "step": 2825
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.00013133944364045668,
      "loss": 0.1865,
      "step": 2826
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.00013130728412928124,
      "loss": 0.0811,
      "step": 2827
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0001312751246181058,
      "loss": 0.2988,
      "step": 2828
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.00013124296510693038,
      "loss": 0.0668,
      "step": 2829
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.00013121080559575497,
      "loss": 0.1145,
      "step": 2830
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.00013117864608457952,
      "loss": 0.0825,
      "step": 2831
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.00013114648657340408,
      "loss": 0.0658,
      "step": 2832
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.00013111432706222866,
      "loss": 0.0279,
      "step": 2833
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.00013108216755105322,
      "loss": 0.1161,
      "step": 2834
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.00013105000803987778,
      "loss": 0.0884,
      "step": 2835
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0001310178485287024,
      "loss": 0.0322,
      "step": 2836
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.00013098568901752695,
      "loss": 0.0104,
      "step": 2837
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0001309535295063515,
      "loss": 0.1492,
      "step": 2838
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0001309213699951761,
      "loss": 0.1538,
      "step": 2839
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.00013088921048400064,
      "loss": 0.0629,
      "step": 2840
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0001308570509728252,
      "loss": 0.0055,
      "step": 2841
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.00013082489146164979,
      "loss": 0.0378,
      "step": 2842
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.00013079273195047437,
      "loss": 0.0759,
      "step": 2843
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.00013076057243929893,
      "loss": 0.0969,
      "step": 2844
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0001307284129281235,
      "loss": 0.0681,
      "step": 2845
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.00013069625341694807,
      "loss": 0.0856,
      "step": 2846
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.00013066409390577262,
      "loss": 0.0071,
      "step": 2847
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0001306319343945972,
      "loss": 0.0203,
      "step": 2848
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0001305997748834218,
      "loss": 0.0363,
      "step": 2849
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.00013056761537224635,
      "loss": 0.0092,
      "step": 2850
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.00013053545586107093,
      "loss": 0.0652,
      "step": 2851
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0001305032963498955,
      "loss": 0.0487,
      "step": 2852
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00013047113683872005,
      "loss": 0.0103,
      "step": 2853
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00013043897732754463,
      "loss": 0.0459,
      "step": 2854
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.0001304068178163692,
      "loss": 0.0503,
      "step": 2855
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00013037465830519377,
      "loss": 0.1184,
      "step": 2856
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00013034249879401833,
      "loss": 0.1365,
      "step": 2857
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00013031033928284291,
      "loss": 0.0912,
      "step": 2858
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.00013027817977166747,
      "loss": 0.1053,
      "step": 2859
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.00013024602026049203,
      "loss": 0.1394,
      "step": 2860
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0001302138607493166,
      "loss": 0.0623,
      "step": 2861
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0001301817012381412,
      "loss": 0.0495,
      "step": 2862
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.00013014954172696575,
      "loss": 0.0733,
      "step": 2863
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.00013011738221579034,
      "loss": 0.0553,
      "step": 2864
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0001300852227046149,
      "loss": 0.0267,
      "step": 2865
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.00013005306319343945,
      "loss": 0.2041,
      "step": 2866
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00013002090368226403,
      "loss": 0.0453,
      "step": 2867
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00012998874417108862,
      "loss": 0.1534,
      "step": 2868
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00012995658465991318,
      "loss": 0.0668,
      "step": 2869
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00012992442514873776,
      "loss": 0.1655,
      "step": 2870
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00012989226563756232,
      "loss": 0.1758,
      "step": 2871
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00012986010612638687,
      "loss": 0.0076,
      "step": 2872
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.00012982794661521146,
      "loss": 0.0241,
      "step": 2873
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00012979578710403601,
      "loss": 0.045,
      "step": 2874
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0001297636275928606,
      "loss": 0.125,
      "step": 2875
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00012973146808168518,
      "loss": 0.0752,
      "step": 2876
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00012969930857050974,
      "loss": 0.0187,
      "step": 2877
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0001296671490593343,
      "loss": 0.0217,
      "step": 2878
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00012963498954815888,
      "loss": 0.0326,
      "step": 2879
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.00012960283003698344,
      "loss": 0.0155,
      "step": 2880
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012957067052580802,
      "loss": 0.0717,
      "step": 2881
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012953851101463258,
      "loss": 0.0037,
      "step": 2882
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012950635150345716,
      "loss": 0.0782,
      "step": 2883
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012947419199228172,
      "loss": 0.155,
      "step": 2884
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012944203248110628,
      "loss": 0.0768,
      "step": 2885
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012940987296993086,
      "loss": 0.0446,
      "step": 2886
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.00012937771345875542,
      "loss": 0.0237,
      "step": 2887
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.00012934555394758,
      "loss": 0.0082,
      "step": 2888
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.00012931339443640459,
      "loss": 0.1724,
      "step": 2889
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.00012928123492522914,
      "loss": 0.1616,
      "step": 2890
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.0001292490754140537,
      "loss": 0.0657,
      "step": 2891
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.00012921691590287828,
      "loss": 0.2979,
      "step": 2892
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.00012918475639170284,
      "loss": 0.0326,
      "step": 2893
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00012915259688052742,
      "loss": 0.0265,
      "step": 2894
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.000129120437369352,
      "loss": 0.0746,
      "step": 2895
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00012908827785817657,
      "loss": 0.0663,
      "step": 2896
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00012905611834700112,
      "loss": 0.0587,
      "step": 2897
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.0001290239588358257,
      "loss": 0.0138,
      "step": 2898
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00012899179932465026,
      "loss": 0.067,
      "step": 2899
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.00012895963981347482,
      "loss": 0.0708,
      "step": 2900
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.00012892748030229943,
      "loss": 0.0207,
      "step": 2901
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.000128895320791124,
      "loss": 0.146,
      "step": 2902
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.00012886316127994855,
      "loss": 0.1227,
      "step": 2903
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.00012883100176877313,
      "loss": 0.1053,
      "step": 2904
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0001287988422575977,
      "loss": 0.0703,
      "step": 2905
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.00012876668274642224,
      "loss": 0.074,
      "step": 2906
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.00012873452323524683,
      "loss": 0.2832,
      "step": 2907
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0001287023637240714,
      "loss": 0.1479,
      "step": 2908
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.00012867020421289597,
      "loss": 0.1494,
      "step": 2909
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.00012863804470172053,
      "loss": 0.029,
      "step": 2910
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0001286058851905451,
      "loss": 0.012,
      "step": 2911
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.00012857372567936967,
      "loss": 0.0056,
      "step": 2912
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.00012854156616819425,
      "loss": 0.0258,
      "step": 2913
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.00012850940665701884,
      "loss": 0.1733,
      "step": 2914
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0001284772471458434,
      "loss": 0.0892,
      "step": 2915
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.00012844508763466795,
      "loss": 0.0356,
      "step": 2916
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.00012841292812349253,
      "loss": 0.0195,
      "step": 2917
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0001283807686123171,
      "loss": 0.0845,
      "step": 2918
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.00012834860910114165,
      "loss": 0.1233,
      "step": 2919
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.00012831644958996626,
      "loss": 0.2241,
      "step": 2920
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.00012828429007879082,
      "loss": 0.0223,
      "step": 2921
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.00012825213056761537,
      "loss": 0.0144,
      "step": 2922
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.00012821997105643996,
      "loss": 0.084,
      "step": 2923
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0001281878115452645,
      "loss": 0.0709,
      "step": 2924
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.00012815565203408907,
      "loss": 0.292,
      "step": 2925
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.00012812349252291368,
      "loss": 0.1943,
      "step": 2926
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.00012809133301173824,
      "loss": 0.1029,
      "step": 2927
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0001280591735005628,
      "loss": 0.0654,
      "step": 2928
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00012802701398938738,
      "loss": 0.0672,
      "step": 2929
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00012799485447821194,
      "loss": 0.0692,
      "step": 2930
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.0001279626949670365,
      "loss": 0.043,
      "step": 2931
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00012793053545586108,
      "loss": 0.0998,
      "step": 2932
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00012789837594468566,
      "loss": 0.0591,
      "step": 2933
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00012786621643351022,
      "loss": 0.1001,
      "step": 2934
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.00012783405692233478,
      "loss": 0.0824,
      "step": 2935
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.00012780189741115936,
      "loss": 0.0649,
      "step": 2936
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.00012776973789998392,
      "loss": 0.0632,
      "step": 2937
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0001277375783888085,
      "loss": 0.2466,
      "step": 2938
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.00012770541887763308,
      "loss": 0.2349,
      "step": 2939
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.00012767325936645764,
      "loss": 0.0661,
      "step": 2940
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0001276410998552822,
      "loss": 0.0994,
      "step": 2941
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.00012760894034410678,
      "loss": 0.0406,
      "step": 2942
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.00012757678083293134,
      "loss": 0.1845,
      "step": 2943
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.0001275446213217559,
      "loss": 0.3096,
      "step": 2944
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.00012751246181058048,
      "loss": 0.0142,
      "step": 2945
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.00012748030229940506,
      "loss": 0.0319,
      "step": 2946
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.00012744814278822962,
      "loss": 0.0804,
      "step": 2947
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.0001274159832770542,
      "loss": 0.0957,
      "step": 2948
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.00012738382376587876,
      "loss": 0.3096,
      "step": 2949
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.00012735166425470332,
      "loss": 0.1389,
      "step": 2950
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0001273195047435279,
      "loss": 0.0387,
      "step": 2951
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0001272873452323525,
      "loss": 0.168,
      "step": 2952
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.00012725518572117704,
      "loss": 0.1748,
      "step": 2953
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.00012722302621000163,
      "loss": 0.038,
      "step": 2954
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.00012719086669882619,
      "loss": 0.0212,
      "step": 2955
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.00012715870718765074,
      "loss": 0.024,
      "step": 2956
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00012712654767647533,
      "loss": 0.151,
      "step": 2957
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00012709438816529988,
      "loss": 0.1545,
      "step": 2958
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00012706222865412447,
      "loss": 0.1324,
      "step": 2959
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00012703006914294902,
      "loss": 0.0176,
      "step": 2960
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.0001269979096317736,
      "loss": 0.0132,
      "step": 2961
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00012696575012059817,
      "loss": 0.0636,
      "step": 2962
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.00012693359060942275,
      "loss": 0.1572,
      "step": 2963
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0001269014310982473,
      "loss": 0.0467,
      "step": 2964
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0001268692715870719,
      "loss": 0.1252,
      "step": 2965
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.00012683711207589645,
      "loss": 0.0265,
      "step": 2966
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.00012680495256472103,
      "loss": 0.0124,
      "step": 2967
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0001267727930535456,
      "loss": 0.0205,
      "step": 2968
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.00012674063354237015,
      "loss": 0.0665,
      "step": 2969
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00012670847403119473,
      "loss": 0.0825,
      "step": 2970
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00012667631452001931,
      "loss": 0.075,
      "step": 2971
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00012664415500884387,
      "loss": 0.0536,
      "step": 2972
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00012661199549766845,
      "loss": 0.0973,
      "step": 2973
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.000126579835986493,
      "loss": 0.0587,
      "step": 2974
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00012654767647531757,
      "loss": 0.1216,
      "step": 2975
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.00012651551696414215,
      "loss": 0.1453,
      "step": 2976
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.0001264833574529667,
      "loss": 0.0196,
      "step": 2977
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.0001264511979417913,
      "loss": 0.0232,
      "step": 2978
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.00012641903843061588,
      "loss": 0.0514,
      "step": 2979
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.00012638687891944043,
      "loss": 0.1265,
      "step": 2980
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.000126354719408265,
      "loss": 0.0283,
      "step": 2981
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.00012632255989708958,
      "loss": 0.2331,
      "step": 2982
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.00012629040038591413,
      "loss": 0.1095,
      "step": 2983
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.00012625824087473872,
      "loss": 0.0069,
      "step": 2984
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.00012622608136356327,
      "loss": 0.0309,
      "step": 2985
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.00012619392185238786,
      "loss": 0.1372,
      "step": 2986
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.00012616176234121241,
      "loss": 0.1316,
      "step": 2987
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.000126129602830037,
      "loss": 0.0494,
      "step": 2988
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.00012609744331886156,
      "loss": 0.1191,
      "step": 2989
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0001260652838076861,
      "loss": 0.1021,
      "step": 2990
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0001260331242965107,
      "loss": 0.0847,
      "step": 2991
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.00012600096478533528,
      "loss": 0.0675,
      "step": 2992
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.00012596880527415984,
      "loss": 0.1481,
      "step": 2993
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0001259366457629844,
      "loss": 0.0088,
      "step": 2994
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.00012590448625180898,
      "loss": 0.2358,
      "step": 2995
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.00012587232674063354,
      "loss": 0.011,
      "step": 2996
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.00012584016722945812,
      "loss": 0.0259,
      "step": 2997
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.0001258080077182827,
      "loss": 0.0108,
      "step": 2998
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00012577584820710726,
      "loss": 0.0401,
      "step": 2999
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00012574368869593182,
      "loss": 0.1401,
      "step": 3000
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.0001257115291847564,
      "loss": 0.0763,
      "step": 3001
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00012567936967358096,
      "loss": 0.1997,
      "step": 3002
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00012564721016240552,
      "loss": 0.0175,
      "step": 3003
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.00012561505065123013,
      "loss": 0.0631,
      "step": 3004
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00012558289114005468,
      "loss": 0.1621,
      "step": 3005
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00012555073162887924,
      "loss": 0.0155,
      "step": 3006
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00012551857211770383,
      "loss": 0.1589,
      "step": 3007
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00012548641260652838,
      "loss": 0.0344,
      "step": 3008
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00012545425309535294,
      "loss": 0.0843,
      "step": 3009
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.00012542209358417752,
      "loss": 0.041,
      "step": 3010
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.0001253899340730021,
      "loss": 0.0448,
      "step": 3011
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00012535777456182666,
      "loss": 0.0165,
      "step": 3012
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00012532561505065125,
      "loss": 0.0914,
      "step": 3013
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0001252934555394758,
      "loss": 0.059,
      "step": 3014
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00012526129602830036,
      "loss": 0.2056,
      "step": 3015
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00012522913651712495,
      "loss": 0.025,
      "step": 3016
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.00012519697700594953,
      "loss": 0.1689,
      "step": 3017
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0001251648174947741,
      "loss": 0.0738,
      "step": 3018
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00012513265798359864,
      "loss": 0.1239,
      "step": 3019
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00012510049847242323,
      "loss": 0.1052,
      "step": 3020
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00012506833896124779,
      "loss": 0.0781,
      "step": 3021
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00012503617945007234,
      "loss": 0.1748,
      "step": 3022
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00012500401993889695,
      "loss": 0.045,
      "step": 3023
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.0001249718604277215,
      "loss": 0.0972,
      "step": 3024
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.00012493970091654607,
      "loss": 0.106,
      "step": 3025
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.00012490754140537065,
      "loss": 0.0561,
      "step": 3026
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0001248753818941952,
      "loss": 0.1469,
      "step": 3027
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.00012484322238301977,
      "loss": 0.1316,
      "step": 3028
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.00012481106287184438,
      "loss": 0.0822,
      "step": 3029
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.00012477890336066893,
      "loss": 0.0131,
      "step": 3030
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0001247467438494935,
      "loss": 0.2401,
      "step": 3031
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.00012471458433831807,
      "loss": 0.0603,
      "step": 3032
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.00012468242482714263,
      "loss": 0.0343,
      "step": 3033
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0001246502653159672,
      "loss": 0.1592,
      "step": 3034
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.00012461810580479177,
      "loss": 0.1494,
      "step": 3035
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.00012458594629361636,
      "loss": 0.0659,
      "step": 3036
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0001245537867824409,
      "loss": 0.1936,
      "step": 3037
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0001245216272712655,
      "loss": 0.0666,
      "step": 3038
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.00012448946776009005,
      "loss": 0.1023,
      "step": 3039
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0001244573082489146,
      "loss": 0.1145,
      "step": 3040
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0001244251487377392,
      "loss": 0.0669,
      "step": 3041
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.00012439298922656378,
      "loss": 0.125,
      "step": 3042
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.00012436082971538834,
      "loss": 0.2695,
      "step": 3043
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0001243286702042129,
      "loss": 0.1206,
      "step": 3044
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.00012429651069303748,
      "loss": 0.0237,
      "step": 3045
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.00012426435118186203,
      "loss": 0.0243,
      "step": 3046
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0001242321916706866,
      "loss": 0.0237,
      "step": 3047
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.00012420003215951118,
      "loss": 0.1255,
      "step": 3048
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.00012416787264833576,
      "loss": 0.1887,
      "step": 3049
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.00012413571313716032,
      "loss": 0.009,
      "step": 3050
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0001241035536259849,
      "loss": 0.1335,
      "step": 3051
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.00012407139411480946,
      "loss": 0.1479,
      "step": 3052
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.00012403923460363401,
      "loss": 0.0038,
      "step": 3053
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0001240070750924586,
      "loss": 0.1455,
      "step": 3054
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.00012397491558128318,
      "loss": 0.0109,
      "step": 3055
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.00012394275607010774,
      "loss": 0.1047,
      "step": 3056
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.00012391059655893232,
      "loss": 0.0931,
      "step": 3057
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.00012387843704775688,
      "loss": 0.1758,
      "step": 3058
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.00012384627753658144,
      "loss": 0.0183,
      "step": 3059
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.00012381411802540602,
      "loss": 0.0834,
      "step": 3060
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0001237819585142306,
      "loss": 0.0075,
      "step": 3061
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.00012374979900305516,
      "loss": 0.106,
      "step": 3062
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.00012371763949187975,
      "loss": 0.0263,
      "step": 3063
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0001236854799807043,
      "loss": 0.081,
      "step": 3064
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.00012365332046952886,
      "loss": 0.0067,
      "step": 3065
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.00012362116095835344,
      "loss": 0.0057,
      "step": 3066
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.000123589001447178,
      "loss": 0.109,
      "step": 3067
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00012355684193600259,
      "loss": 0.0513,
      "step": 3068
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00012352468242482714,
      "loss": 0.1738,
      "step": 3069
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00012349252291365173,
      "loss": 0.084,
      "step": 3070
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00012346036340247628,
      "loss": 0.0084,
      "step": 3071
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00012342820389130084,
      "loss": 0.1714,
      "step": 3072
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.00012339604438012542,
      "loss": 0.1261,
      "step": 3073
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.00012336388486895,
      "loss": 0.0707,
      "step": 3074
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.00012333172535777457,
      "loss": 0.0038,
      "step": 3075
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.00012329956584659915,
      "loss": 0.082,
      "step": 3076
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0001232674063354237,
      "loss": 0.228,
      "step": 3077
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.00012323524682424826,
      "loss": 0.1146,
      "step": 3078
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.00012320308731307285,
      "loss": 0.0909,
      "step": 3079
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0001231709278018974,
      "loss": 0.1385,
      "step": 3080
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.000123138768290722,
      "loss": 0.1851,
      "step": 3081
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.00012310660877954657,
      "loss": 0.1282,
      "step": 3082
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.00012307444926837113,
      "loss": 0.0516,
      "step": 3083
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0001230422897571957,
      "loss": 0.087,
      "step": 3084
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.00012301013024602027,
      "loss": 0.0964,
      "step": 3085
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.00012297797073484483,
      "loss": 0.0458,
      "step": 3086
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0001229458112236694,
      "loss": 0.0728,
      "step": 3087
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.00012291365171249397,
      "loss": 0.0543,
      "step": 3088
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.00012288149220131855,
      "loss": 0.117,
      "step": 3089
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0001228493326901431,
      "loss": 0.089,
      "step": 3090
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0001228171731789677,
      "loss": 0.103,
      "step": 3091
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.00012278501366779225,
      "loss": 0.1855,
      "step": 3092
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0001227528541566168,
      "loss": 0.0225,
      "step": 3093
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0001227206946454414,
      "loss": 0.1587,
      "step": 3094
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.00012268853513426598,
      "loss": 0.046,
      "step": 3095
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.00012265637562309053,
      "loss": 0.1155,
      "step": 3096
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0001226242161119151,
      "loss": 0.0156,
      "step": 3097
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.00012259205660073967,
      "loss": 0.0701,
      "step": 3098
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.00012255989708956423,
      "loss": 0.0561,
      "step": 3099
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.00012252773757838881,
      "loss": 0.0229,
      "step": 3100
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0001224955780672134,
      "loss": 0.0362,
      "step": 3101
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.00012246341855603796,
      "loss": 0.2048,
      "step": 3102
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0001224312590448625,
      "loss": 0.0489,
      "step": 3103
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0001223990995336871,
      "loss": 0.057,
      "step": 3104
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.00012236694002251165,
      "loss": 0.0464,
      "step": 3105
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.00012233478051133624,
      "loss": 0.0452,
      "step": 3106
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.00012230262100016082,
      "loss": 0.0036,
      "step": 3107
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.00012227046148898538,
      "loss": 0.0276,
      "step": 3108
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00012223830197780994,
      "loss": 0.1462,
      "step": 3109
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00012220614246663452,
      "loss": 0.0097,
      "step": 3110
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00012217398295545908,
      "loss": 0.0901,
      "step": 3111
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00012214182344428363,
      "loss": 0.2214,
      "step": 3112
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00012210966393310822,
      "loss": 0.0298,
      "step": 3113
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.0001220775044219328,
      "loss": 0.1808,
      "step": 3114
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.00012204534491075736,
      "loss": 0.012,
      "step": 3115
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.00012201318539958193,
      "loss": 0.1572,
      "step": 3116
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0001219810258884065,
      "loss": 0.1045,
      "step": 3117
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.00012194886637723106,
      "loss": 0.0372,
      "step": 3118
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.00012191670686605565,
      "loss": 0.058,
      "step": 3119
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.00012188454735488023,
      "loss": 0.0297,
      "step": 3120
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.00012185238784370478,
      "loss": 0.0112,
      "step": 3121
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.00012182022833252935,
      "loss": 0.0329,
      "step": 3122
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00012178806882135392,
      "loss": 0.0151,
      "step": 3123
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00012175590931017848,
      "loss": 0.271,
      "step": 3124
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00012172374979900305,
      "loss": 0.1106,
      "step": 3125
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00012169159028782763,
      "loss": 0.0121,
      "step": 3126
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.0001216594307766522,
      "loss": 0.0177,
      "step": 3127
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00012162727126547678,
      "loss": 0.0875,
      "step": 3128
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.00012159511175430133,
      "loss": 0.016,
      "step": 3129
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0001215629522431259,
      "loss": 0.127,
      "step": 3130
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.00012153079273195047,
      "loss": 0.1212,
      "step": 3131
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.00012149863322077506,
      "loss": 0.2024,
      "step": 3132
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.00012146647370959963,
      "loss": 0.0374,
      "step": 3133
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0001214343141984242,
      "loss": 0.001,
      "step": 3134
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.00012140215468724876,
      "loss": 0.1068,
      "step": 3135
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00012136999517607333,
      "loss": 0.1196,
      "step": 3136
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.0001213378356648979,
      "loss": 0.0833,
      "step": 3137
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00012130567615372245,
      "loss": 0.0392,
      "step": 3138
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00012127351664254705,
      "loss": 0.0611,
      "step": 3139
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00012124135713137161,
      "loss": 0.1672,
      "step": 3140
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00012120919762019618,
      "loss": 0.0349,
      "step": 3141
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.00012117703810902075,
      "loss": 0.1315,
      "step": 3142
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.0001211448785978453,
      "loss": 0.1345,
      "step": 3143
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.00012111271908666988,
      "loss": 0.0016,
      "step": 3144
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.00012108055957549447,
      "loss": 0.0394,
      "step": 3145
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.00012104840006431903,
      "loss": 0.0068,
      "step": 3146
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.0001210162405531436,
      "loss": 0.0362,
      "step": 3147
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.00012098408104196817,
      "loss": 0.0562,
      "step": 3148
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.00012095192153079273,
      "loss": 0.0021,
      "step": 3149
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.0001209197620196173,
      "loss": 0.123,
      "step": 3150
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00012088760250844188,
      "loss": 0.0248,
      "step": 3151
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00012085544299726645,
      "loss": 0.1709,
      "step": 3152
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00012082328348609103,
      "loss": 0.2107,
      "step": 3153
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00012079112397491558,
      "loss": 0.0949,
      "step": 3154
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00012075896446374015,
      "loss": 0.0022,
      "step": 3155
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.00012072680495256472,
      "loss": 0.241,
      "step": 3156
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0001206946454413893,
      "loss": 0.0357,
      "step": 3157
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.00012066248593021388,
      "loss": 0.1113,
      "step": 3158
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.00012063032641903845,
      "loss": 0.0668,
      "step": 3159
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.000120598166907863,
      "loss": 0.0134,
      "step": 3160
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.00012056600739668758,
      "loss": 0.0934,
      "step": 3161
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.00012053384788551215,
      "loss": 0.114,
      "step": 3162
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0001205016883743367,
      "loss": 0.0467,
      "step": 3163
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0001204695288631613,
      "loss": 0.0459,
      "step": 3164
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.00012043736935198586,
      "loss": 0.0158,
      "step": 3165
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.00012040520984081043,
      "loss": 0.0537,
      "step": 3166
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.000120373050329635,
      "loss": 0.0104,
      "step": 3167
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.00012034089081845956,
      "loss": 0.1179,
      "step": 3168
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.00012030873130728413,
      "loss": 0.0654,
      "step": 3169
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0001202765717961087,
      "loss": 0.1226,
      "step": 3170
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.00012024441228493328,
      "loss": 0.0804,
      "step": 3171
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.00012021225277375785,
      "loss": 0.0593,
      "step": 3172
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.00012018009326258242,
      "loss": 0.0451,
      "step": 3173
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.00012014793375140698,
      "loss": 0.1097,
      "step": 3174
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.00012011577424023155,
      "loss": 0.1182,
      "step": 3175
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.00012008361472905612,
      "loss": 0.0321,
      "step": 3176
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0001200514552178807,
      "loss": 0.1045,
      "step": 3177
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.00012001929570670527,
      "loss": 0.147,
      "step": 3178
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.00011998713619552983,
      "loss": 0.0089,
      "step": 3179
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0001199549766843544,
      "loss": 0.2695,
      "step": 3180
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.00011992281717317897,
      "loss": 0.0244,
      "step": 3181
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.00011989065766200353,
      "loss": 0.0593,
      "step": 3182
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0001198584981508281,
      "loss": 0.0263,
      "step": 3183
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0001198263386396527,
      "loss": 0.2083,
      "step": 3184
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00011979417912847725,
      "loss": 0.0807,
      "step": 3185
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00011976201961730182,
      "loss": 0.1112,
      "step": 3186
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.0001197298601061264,
      "loss": 0.012,
      "step": 3187
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00011969770059495095,
      "loss": 0.1014,
      "step": 3188
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00011966554108377552,
      "loss": 0.0051,
      "step": 3189
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00011963338157260011,
      "loss": 0.0085,
      "step": 3190
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.00011960122206142468,
      "loss": 0.12,
      "step": 3191
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.00011956906255024925,
      "loss": 0.1223,
      "step": 3192
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0001195369030390738,
      "loss": 0.1169,
      "step": 3193
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.00011950474352789838,
      "loss": 0.0567,
      "step": 3194
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.00011947258401672295,
      "loss": 0.2517,
      "step": 3195
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.00011944042450554753,
      "loss": 0.0841,
      "step": 3196
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0001194082649943721,
      "loss": 0.0198,
      "step": 3197
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.00011937610548319667,
      "loss": 0.1187,
      "step": 3198
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00011934394597202123,
      "loss": 0.1026,
      "step": 3199
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0001193117864608458,
      "loss": 0.0437,
      "step": 3200
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00011927962694967037,
      "loss": 0.249,
      "step": 3201
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00011924746743849493,
      "loss": 0.0095,
      "step": 3202
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00011921530792731952,
      "loss": 0.0851,
      "step": 3203
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00011918314841614408,
      "loss": 0.1836,
      "step": 3204
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.00011915098890496865,
      "loss": 0.1199,
      "step": 3205
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00011911882939379322,
      "loss": 0.1057,
      "step": 3206
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00011908666988261778,
      "loss": 0.026,
      "step": 3207
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00011905451037144235,
      "loss": 0.1266,
      "step": 3208
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00011902235086026695,
      "loss": 0.1238,
      "step": 3209
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.0001189901913490915,
      "loss": 0.0536,
      "step": 3210
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00011895803183791607,
      "loss": 0.0358,
      "step": 3211
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.00011892587232674064,
      "loss": 0.0096,
      "step": 3212
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0001188937128155652,
      "loss": 0.0303,
      "step": 3213
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.00011886155330438977,
      "loss": 0.002,
      "step": 3214
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.00011882939379321434,
      "loss": 0.0452,
      "step": 3215
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.00011879723428203893,
      "loss": 0.0098,
      "step": 3216
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0001187650747708635,
      "loss": 0.0583,
      "step": 3217
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.00011873291525968805,
      "loss": 0.0849,
      "step": 3218
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.00011870075574851262,
      "loss": 0.1161,
      "step": 3219
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0001186685962373372,
      "loss": 0.0693,
      "step": 3220
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.00011863643672616177,
      "loss": 0.0009,
      "step": 3221
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.00011860427721498635,
      "loss": 0.1709,
      "step": 3222
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.00011857211770381092,
      "loss": 0.0888,
      "step": 3223
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.00011853995819263548,
      "loss": 0.2039,
      "step": 3224
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.00011850779868146005,
      "loss": 0.0234,
      "step": 3225
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00011847563917028462,
      "loss": 0.0831,
      "step": 3226
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00011844347965910918,
      "loss": 0.1071,
      "step": 3227
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00011841132014793375,
      "loss": 0.0547,
      "step": 3228
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00011837916063675833,
      "loss": 0.1079,
      "step": 3229
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.0001183470011255829,
      "loss": 0.1572,
      "step": 3230
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00011831484161440747,
      "loss": 0.1424,
      "step": 3231
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.00011828268210323203,
      "loss": 0.1217,
      "step": 3232
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0001182505225920566,
      "loss": 0.0394,
      "step": 3233
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.00011821836308088117,
      "loss": 0.0112,
      "step": 3234
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.00011818620356970575,
      "loss": 0.0126,
      "step": 3235
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.00011815404405853032,
      "loss": 0.0036,
      "step": 3236
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0001181218845473549,
      "loss": 0.0685,
      "step": 3237
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.00011808972503617945,
      "loss": 0.1091,
      "step": 3238
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.00011805756552500402,
      "loss": 0.0106,
      "step": 3239
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00011802540601382859,
      "loss": 0.1321,
      "step": 3240
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00011799324650265318,
      "loss": 0.0059,
      "step": 3241
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00011796108699147775,
      "loss": 0.0701,
      "step": 3242
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.0001179289274803023,
      "loss": 0.0402,
      "step": 3243
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00011789676796912687,
      "loss": 0.1011,
      "step": 3244
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00011786460845795144,
      "loss": 0.0107,
      "step": 3245
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.00011783244894677601,
      "loss": 0.2283,
      "step": 3246
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.00011780028943560057,
      "loss": 0.1047,
      "step": 3247
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.00011776812992442517,
      "loss": 0.0382,
      "step": 3248
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.00011773597041324973,
      "loss": 0.0144,
      "step": 3249
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.0001177038109020743,
      "loss": 0.0202,
      "step": 3250
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.00011767165139089887,
      "loss": 0.2681,
      "step": 3251
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.00011763949187972342,
      "loss": 0.1631,
      "step": 3252
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.000117607332368548,
      "loss": 0.2761,
      "step": 3253
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011757517285737258,
      "loss": 0.0179,
      "step": 3254
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011754301334619715,
      "loss": 0.0215,
      "step": 3255
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011751085383502172,
      "loss": 0.0018,
      "step": 3256
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011747869432384628,
      "loss": 0.0269,
      "step": 3257
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011744653481267085,
      "loss": 0.0287,
      "step": 3258
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011741437530149542,
      "loss": 0.0102,
      "step": 3259
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.00011738221579031999,
      "loss": 0.1592,
      "step": 3260
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.00011735005627914457,
      "loss": 0.0215,
      "step": 3261
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.00011731789676796914,
      "loss": 0.0673,
      "step": 3262
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.0001172857372567937,
      "loss": 0.134,
      "step": 3263
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.00011725357774561827,
      "loss": 0.167,
      "step": 3264
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.00011722141823444284,
      "loss": 0.1841,
      "step": 3265
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.0001171892587232674,
      "loss": 0.0464,
      "step": 3266
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.000117157099212092,
      "loss": 0.161,
      "step": 3267
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00011712493970091655,
      "loss": 0.1211,
      "step": 3268
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00011709278018974112,
      "loss": 0.0263,
      "step": 3269
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.0001170606206785657,
      "loss": 0.0316,
      "step": 3270
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00011702846116739026,
      "loss": 0.0596,
      "step": 3271
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00011699630165621482,
      "loss": 0.1389,
      "step": 3272
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00011696414214503939,
      "loss": 0.0166,
      "step": 3273
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.00011693198263386398,
      "loss": 0.1619,
      "step": 3274
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00011689982312268855,
      "loss": 0.0945,
      "step": 3275
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00011686766361151312,
      "loss": 0.1208,
      "step": 3276
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00011683550410033767,
      "loss": 0.0483,
      "step": 3277
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00011680334458916224,
      "loss": 0.0106,
      "step": 3278
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00011677118507798681,
      "loss": 0.0903,
      "step": 3279
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.0001167390255668114,
      "loss": 0.1836,
      "step": 3280
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.00011670686605563597,
      "loss": 0.0889,
      "step": 3281
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.00011667470654446053,
      "loss": 0.1365,
      "step": 3282
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.0001166425470332851,
      "loss": 0.2114,
      "step": 3283
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.00011661038752210967,
      "loss": 0.076,
      "step": 3284
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.00011657822801093424,
      "loss": 0.0027,
      "step": 3285
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.0001165460684997588,
      "loss": 0.012,
      "step": 3286
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.00011651390898858339,
      "loss": 0.0945,
      "step": 3287
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.00011648174947740795,
      "loss": 0.0659,
      "step": 3288
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.00011644958996623252,
      "loss": 0.0052,
      "step": 3289
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.00011641743045505709,
      "loss": 0.2563,
      "step": 3290
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.00011638527094388165,
      "loss": 0.1082,
      "step": 3291
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.00011635311143270622,
      "loss": 0.1584,
      "step": 3292
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0001163209519215308,
      "loss": 0.021,
      "step": 3293
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.00011628879241035537,
      "loss": 0.137,
      "step": 3294
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00011625663289917994,
      "loss": 0.0308,
      "step": 3295
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00011622447338800451,
      "loss": 0.1475,
      "step": 3296
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00011619231387682907,
      "loss": 0.1608,
      "step": 3297
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00011616015436565364,
      "loss": 0.0366,
      "step": 3298
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00011612799485447822,
      "loss": 0.1346,
      "step": 3299
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.0001160958353433028,
      "loss": 0.1233,
      "step": 3300
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.00011606367583212737,
      "loss": 0.0077,
      "step": 3301
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00011603151632095192,
      "loss": 0.0181,
      "step": 3302
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.0001159993568097765,
      "loss": 0.032,
      "step": 3303
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00011596719729860106,
      "loss": 0.0104,
      "step": 3304
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00011593503778742562,
      "loss": 0.1782,
      "step": 3305
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00011590287827625022,
      "loss": 0.1862,
      "step": 3306
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00011587071876507478,
      "loss": 0.0157,
      "step": 3307
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.00011583855925389935,
      "loss": 0.197,
      "step": 3308
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00011580639974272392,
      "loss": 0.0056,
      "step": 3309
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00011577424023154849,
      "loss": 0.1137,
      "step": 3310
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00011574208072037304,
      "loss": 0.1819,
      "step": 3311
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00011570992120919764,
      "loss": 0.0659,
      "step": 3312
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.0001156777616980222,
      "loss": 0.165,
      "step": 3313
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00011564560218684677,
      "loss": 0.0401,
      "step": 3314
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.00011561344267567134,
      "loss": 0.035,
      "step": 3315
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.0001155812831644959,
      "loss": 0.1002,
      "step": 3316
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00011554912365332047,
      "loss": 0.1362,
      "step": 3317
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00011551696414214504,
      "loss": 0.0072,
      "step": 3318
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00011548480463096962,
      "loss": 0.0773,
      "step": 3319
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00011545264511979419,
      "loss": 0.0059,
      "step": 3320
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00011542048560861875,
      "loss": 0.037,
      "step": 3321
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.00011538832609744332,
      "loss": 0.0677,
      "step": 3322
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011535616658626789,
      "loss": 0.059,
      "step": 3323
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011532400707509246,
      "loss": 0.0963,
      "step": 3324
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011529184756391704,
      "loss": 0.0848,
      "step": 3325
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011525968805274162,
      "loss": 0.0204,
      "step": 3326
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011522752854156617,
      "loss": 0.0409,
      "step": 3327
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011519536903039074,
      "loss": 0.0273,
      "step": 3328
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.00011516320951921531,
      "loss": 0.0704,
      "step": 3329
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.00011513105000803987,
      "loss": 0.0218,
      "step": 3330
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.00011509889049686444,
      "loss": 0.1216,
      "step": 3331
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.00011506673098568902,
      "loss": 0.0228,
      "step": 3332
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0001150345714745136,
      "loss": 0.0019,
      "step": 3333
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.00011500241196333817,
      "loss": 0.0899,
      "step": 3334
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.00011497025245216274,
      "loss": 0.1477,
      "step": 3335
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0001149380929409873,
      "loss": 0.0353,
      "step": 3336
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011490593342981186,
      "loss": 0.1909,
      "step": 3337
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011487377391863645,
      "loss": 0.0306,
      "step": 3338
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011484161440746102,
      "loss": 0.165,
      "step": 3339
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011480945489628559,
      "loss": 0.1074,
      "step": 3340
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011477729538511015,
      "loss": 0.0317,
      "step": 3341
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011474513587393472,
      "loss": 0.0163,
      "step": 3342
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.00011471297636275929,
      "loss": 0.0335,
      "step": 3343
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00011468081685158387,
      "loss": 0.1794,
      "step": 3344
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00011464865734040844,
      "loss": 0.0469,
      "step": 3345
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.000114616497829233,
      "loss": 0.0563,
      "step": 3346
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00011458433831805757,
      "loss": 0.024,
      "step": 3347
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00011455217880688214,
      "loss": 0.127,
      "step": 3348
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00011452001929570671,
      "loss": 0.0812,
      "step": 3349
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.00011448785978453127,
      "loss": 0.0544,
      "step": 3350
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011445570027335586,
      "loss": 0.0784,
      "step": 3351
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011442354076218042,
      "loss": 0.1711,
      "step": 3352
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011439138125100499,
      "loss": 0.1221,
      "step": 3353
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011435922173982956,
      "loss": 0.0096,
      "step": 3354
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011432706222865412,
      "loss": 0.0137,
      "step": 3355
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011429490271747869,
      "loss": 0.0561,
      "step": 3356
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.00011426274320630327,
      "loss": 0.1108,
      "step": 3357
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011423058369512784,
      "loss": 0.2173,
      "step": 3358
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011419842418395242,
      "loss": 0.1934,
      "step": 3359
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011416626467277699,
      "loss": 0.0114,
      "step": 3360
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011413410516160154,
      "loss": 0.0213,
      "step": 3361
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011410194565042611,
      "loss": 0.0022,
      "step": 3362
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011406978613925068,
      "loss": 0.1525,
      "step": 3363
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.00011403762662807527,
      "loss": 0.0498,
      "step": 3364
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00011400546711689984,
      "loss": 0.1116,
      "step": 3365
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.0001139733076057244,
      "loss": 0.0052,
      "step": 3366
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00011394114809454897,
      "loss": 0.0176,
      "step": 3367
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00011390898858337354,
      "loss": 0.0986,
      "step": 3368
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00011387682907219809,
      "loss": 0.1611,
      "step": 3369
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00011384466956102269,
      "loss": 0.0601,
      "step": 3370
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.00011381251004984725,
      "loss": 0.0545,
      "step": 3371
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.00011378035053867182,
      "loss": 0.2225,
      "step": 3372
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.00011374819102749639,
      "loss": 0.2065,
      "step": 3373
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.00011371603151632096,
      "loss": 0.0424,
      "step": 3374
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.00011368387200514552,
      "loss": 0.0454,
      "step": 3375
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.00011365171249397009,
      "loss": 0.1577,
      "step": 3376
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.00011361955298279467,
      "loss": 0.0411,
      "step": 3377
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00011358739347161924,
      "loss": 0.0115,
      "step": 3378
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00011355523396044381,
      "loss": 0.1567,
      "step": 3379
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00011352307444926837,
      "loss": 0.1769,
      "step": 3380
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00011349091493809294,
      "loss": 0.1533,
      "step": 3381
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00011345875542691751,
      "loss": 0.0079,
      "step": 3382
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.0001134265959157421,
      "loss": 0.1816,
      "step": 3383
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.00011339443640456666,
      "loss": 0.1786,
      "step": 3384
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011336227689339123,
      "loss": 0.0121,
      "step": 3385
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011333011738221579,
      "loss": 0.0194,
      "step": 3386
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011329795787104036,
      "loss": 0.0862,
      "step": 3387
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011326579835986493,
      "loss": 0.023,
      "step": 3388
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011323363884868952,
      "loss": 0.1265,
      "step": 3389
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011320147933751409,
      "loss": 0.0904,
      "step": 3390
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.00011316931982633864,
      "loss": 0.0191,
      "step": 3391
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00011313716031516321,
      "loss": 0.1133,
      "step": 3392
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00011310500080398779,
      "loss": 0.0426,
      "step": 3393
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00011307284129281234,
      "loss": 0.0226,
      "step": 3394
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00011304068178163691,
      "loss": 0.1357,
      "step": 3395
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.0001130085222704615,
      "loss": 0.0037,
      "step": 3396
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00011297636275928607,
      "loss": 0.0677,
      "step": 3397
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.00011294420324811064,
      "loss": 0.0869,
      "step": 3398
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011291204373693521,
      "loss": 0.0239,
      "step": 3399
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011287988422575977,
      "loss": 0.1428,
      "step": 3400
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011284772471458434,
      "loss": 0.0458,
      "step": 3401
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011281556520340892,
      "loss": 0.1279,
      "step": 3402
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011278340569223349,
      "loss": 0.0145,
      "step": 3403
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011275124618105806,
      "loss": 0.0284,
      "step": 3404
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.00011271908666988262,
      "loss": 0.2061,
      "step": 3405
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011268692715870719,
      "loss": 0.2148,
      "step": 3406
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011265476764753176,
      "loss": 0.0876,
      "step": 3407
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011262260813635632,
      "loss": 0.2207,
      "step": 3408
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011259044862518091,
      "loss": 0.093,
      "step": 3409
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011255828911400548,
      "loss": 0.1113,
      "step": 3410
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011252612960283004,
      "loss": 0.0114,
      "step": 3411
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.00011249397009165461,
      "loss": 0.1272,
      "step": 3412
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00011246181058047918,
      "loss": 0.158,
      "step": 3413
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00011242965106930374,
      "loss": 0.0859,
      "step": 3414
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00011239749155812834,
      "loss": 0.1205,
      "step": 3415
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.0001123653320469529,
      "loss": 0.0398,
      "step": 3416
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00011233317253577746,
      "loss": 0.0214,
      "step": 3417
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00011230101302460203,
      "loss": 0.0473,
      "step": 3418
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.00011226885351342659,
      "loss": 0.059,
      "step": 3419
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011223669400225116,
      "loss": 0.1176,
      "step": 3420
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011220453449107573,
      "loss": 0.1257,
      "step": 3421
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011217237497990032,
      "loss": 0.0118,
      "step": 3422
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011214021546872489,
      "loss": 0.0152,
      "step": 3423
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011210805595754946,
      "loss": 0.0901,
      "step": 3424
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011207589644637401,
      "loss": 0.0112,
      "step": 3425
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.00011204373693519859,
      "loss": 0.0093,
      "step": 3426
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011201157742402316,
      "loss": 0.087,
      "step": 3427
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011197941791284774,
      "loss": 0.1232,
      "step": 3428
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011194725840167231,
      "loss": 0.1061,
      "step": 3429
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011191509889049687,
      "loss": 0.024,
      "step": 3430
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011188293937932144,
      "loss": 0.0859,
      "step": 3431
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011185077986814601,
      "loss": 0.0286,
      "step": 3432
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.00011181862035697057,
      "loss": 0.1498,
      "step": 3433
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011178646084579516,
      "loss": 0.0741,
      "step": 3434
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011175430133461972,
      "loss": 0.0199,
      "step": 3435
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011172214182344429,
      "loss": 0.0536,
      "step": 3436
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011168998231226886,
      "loss": 0.0418,
      "step": 3437
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011165782280109343,
      "loss": 0.0745,
      "step": 3438
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011162566328991799,
      "loss": 0.1021,
      "step": 3439
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.00011159350377874256,
      "loss": 0.0029,
      "step": 3440
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011156134426756714,
      "loss": 0.0058,
      "step": 3441
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011152918475639171,
      "loss": 0.0101,
      "step": 3442
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011149702524521628,
      "loss": 0.0162,
      "step": 3443
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011146486573404084,
      "loss": 0.0014,
      "step": 3444
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011143270622286541,
      "loss": 0.0255,
      "step": 3445
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011140054671168998,
      "loss": 0.0141,
      "step": 3446
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.00011136838720051457,
      "loss": 0.0119,
      "step": 3447
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.00011133622768933914,
      "loss": 0.0092,
      "step": 3448
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.00011130406817816371,
      "loss": 0.0736,
      "step": 3449
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.00011127190866698826,
      "loss": 0.1072,
      "step": 3450
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.00011123974915581283,
      "loss": 0.0139,
      "step": 3451
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.0001112075896446374,
      "loss": 0.0187,
      "step": 3452
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.00011117543013346196,
      "loss": 0.0154,
      "step": 3453
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.00011114327062228656,
      "loss": 0.001,
      "step": 3454
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00011111111111111112,
      "loss": 0.0146,
      "step": 3455
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00011107895159993569,
      "loss": 0.1416,
      "step": 3456
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00011104679208876026,
      "loss": 0.0823,
      "step": 3457
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00011101463257758481,
      "loss": 0.0003,
      "step": 3458
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00011098247306640938,
      "loss": 0.0021,
      "step": 3459
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.00011095031355523397,
      "loss": 0.0128,
      "step": 3460
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011091815404405854,
      "loss": 0.0748,
      "step": 3461
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011088599453288311,
      "loss": 0.0283,
      "step": 3462
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011085383502170768,
      "loss": 0.0089,
      "step": 3463
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011082167551053224,
      "loss": 0.0033,
      "step": 3464
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011078951599935681,
      "loss": 0.0216,
      "step": 3465
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011075735648818138,
      "loss": 0.0742,
      "step": 3466
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.00011072519697700596,
      "loss": 0.0464,
      "step": 3467
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011069303746583053,
      "loss": 0.0029,
      "step": 3468
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011066087795465509,
      "loss": 0.0112,
      "step": 3469
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011062871844347966,
      "loss": 0.0459,
      "step": 3470
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011059655893230423,
      "loss": 0.1023,
      "step": 3471
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011056439942112879,
      "loss": 0.0056,
      "step": 3472
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011053223990995339,
      "loss": 0.0555,
      "step": 3473
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.00011050008039877796,
      "loss": 0.0039,
      "step": 3474
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011046792088760251,
      "loss": 0.0802,
      "step": 3475
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011043576137642708,
      "loss": 0.0959,
      "step": 3476
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011040360186525165,
      "loss": 0.0032,
      "step": 3477
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011037144235407621,
      "loss": 0.0699,
      "step": 3478
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011033928284290081,
      "loss": 0.0049,
      "step": 3479
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011030712333172537,
      "loss": 0.1417,
      "step": 3480
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.00011027496382054994,
      "loss": 0.0177,
      "step": 3481
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.0001102428043093745,
      "loss": 0.0202,
      "step": 3482
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00011021064479819906,
      "loss": 0.1257,
      "step": 3483
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00011017848528702363,
      "loss": 0.0087,
      "step": 3484
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.0001101463257758482,
      "loss": 0.0484,
      "step": 3485
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00011011416626467279,
      "loss": 0.0444,
      "step": 3486
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00011008200675349736,
      "loss": 0.118,
      "step": 3487
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.00011004984724232193,
      "loss": 0.0774,
      "step": 3488
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00011001768773114649,
      "loss": 0.0127,
      "step": 3489
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00010998552821997106,
      "loss": 0.0308,
      "step": 3490
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00010995336870879563,
      "loss": 0.026,
      "step": 3491
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00010992120919762021,
      "loss": 0.0363,
      "step": 3492
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00010988904968644478,
      "loss": 0.0382,
      "step": 3493
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00010985689017526934,
      "loss": 0.0124,
      "step": 3494
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.00010982473066409391,
      "loss": 0.0321,
      "step": 3495
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.00010979257115291848,
      "loss": 0.0271,
      "step": 3496
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.00010976041164174304,
      "loss": 0.0049,
      "step": 3497
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.00010972825213056761,
      "loss": 0.1907,
      "step": 3498
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0001096960926193922,
      "loss": 0.0617,
      "step": 3499
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.00010966393310821676,
      "loss": 0.0015,
      "step": 3500
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.00010963177359704133,
      "loss": 0.0085,
      "step": 3501
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0001095996140858659,
      "loss": 0.0366,
      "step": 3502
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010956745457469046,
      "loss": 0.0118,
      "step": 3503
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010953529506351503,
      "loss": 0.063,
      "step": 3504
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010950313555233961,
      "loss": 0.0116,
      "step": 3505
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010947097604116419,
      "loss": 0.0113,
      "step": 3506
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010943881652998876,
      "loss": 0.062,
      "step": 3507
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010940665701881331,
      "loss": 0.0054,
      "step": 3508
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.00010937449750763788,
      "loss": 0.113,
      "step": 3509
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00010934233799646245,
      "loss": 0.0588,
      "step": 3510
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00010931017848528702,
      "loss": 0.1572,
      "step": 3511
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00010927801897411161,
      "loss": 0.0124,
      "step": 3512
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00010924585946293618,
      "loss": 0.1279,
      "step": 3513
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00010921369995176074,
      "loss": 0.007,
      "step": 3514
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.0001091815404405853,
      "loss": 0.0463,
      "step": 3515
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.00010914938092940988,
      "loss": 0.0131,
      "step": 3516
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010911722141823443,
      "loss": 0.1027,
      "step": 3517
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010908506190705903,
      "loss": 0.0484,
      "step": 3518
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010905290239588359,
      "loss": 0.0016,
      "step": 3519
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010902074288470816,
      "loss": 0.0063,
      "step": 3520
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010898858337353273,
      "loss": 0.0544,
      "step": 3521
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010895642386235729,
      "loss": 0.0453,
      "step": 3522
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.00010892426435118186,
      "loss": 0.0369,
      "step": 3523
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010889210484000645,
      "loss": 0.0325,
      "step": 3524
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010885994532883101,
      "loss": 0.0135,
      "step": 3525
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010882778581765558,
      "loss": 0.0237,
      "step": 3526
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010879562630648015,
      "loss": 0.1377,
      "step": 3527
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010876346679530471,
      "loss": 0.0163,
      "step": 3528
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010873130728412928,
      "loss": 0.0663,
      "step": 3529
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.00010869914777295385,
      "loss": 0.0075,
      "step": 3530
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.00010866698826177843,
      "loss": 0.0183,
      "step": 3531
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.000108634828750603,
      "loss": 0.1311,
      "step": 3532
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.00010860266923942756,
      "loss": 0.0321,
      "step": 3533
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.00010857050972825213,
      "loss": 0.0177,
      "step": 3534
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0001085383502170767,
      "loss": 0.0413,
      "step": 3535
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.00010850619070590127,
      "loss": 0.0012,
      "step": 3536
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.00010847403119472586,
      "loss": 0.027,
      "step": 3537
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00010844187168355043,
      "loss": 0.0486,
      "step": 3538
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00010840971217237499,
      "loss": 0.0388,
      "step": 3539
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00010837755266119956,
      "loss": 0.0203,
      "step": 3540
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00010834539315002413,
      "loss": 0.1855,
      "step": 3541
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00010831323363884868,
      "loss": 0.1627,
      "step": 3542
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.00010828107412767325,
      "loss": 0.1836,
      "step": 3543
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.00010824891461649784,
      "loss": 0.0114,
      "step": 3544
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.00010821675510532241,
      "loss": 0.0381,
      "step": 3545
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.00010818459559414698,
      "loss": 0.0021,
      "step": 3546
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.00010815243608297154,
      "loss": 0.0488,
      "step": 3547
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0001081202765717961,
      "loss": 0.084,
      "step": 3548
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.00010808811706062068,
      "loss": 0.0801,
      "step": 3549
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.00010805595754944526,
      "loss": 0.009,
      "step": 3550
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.00010802379803826983,
      "loss": 0.0217,
      "step": 3551
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0001079916385270944,
      "loss": 0.1034,
      "step": 3552
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.00010795947901591896,
      "loss": 0.0074,
      "step": 3553
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.00010792731950474353,
      "loss": 0.0073,
      "step": 3554
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0001078951599935681,
      "loss": 0.0347,
      "step": 3555
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.00010786300048239266,
      "loss": 0.0096,
      "step": 3556
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.00010783084097121725,
      "loss": 0.0926,
      "step": 3557
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010779868146004181,
      "loss": 0.0865,
      "step": 3558
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010776652194886638,
      "loss": 0.0015,
      "step": 3559
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010773436243769095,
      "loss": 0.0687,
      "step": 3560
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010770220292651552,
      "loss": 0.0201,
      "step": 3561
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010767004341534008,
      "loss": 0.0847,
      "step": 3562
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010763788390416468,
      "loss": 0.118,
      "step": 3563
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.00010760572439298923,
      "loss": 0.0553,
      "step": 3564
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.0001075735648818138,
      "loss": 0.0298,
      "step": 3565
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.00010754140537063838,
      "loss": 0.0544,
      "step": 3566
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.00010750924585946293,
      "loss": 0.1294,
      "step": 3567
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.0001074770863482875,
      "loss": 0.0237,
      "step": 3568
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.00010744492683711209,
      "loss": 0.0419,
      "step": 3569
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.00010741276732593666,
      "loss": 0.0197,
      "step": 3570
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.00010738060781476123,
      "loss": 0.1235,
      "step": 3571
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00010734844830358579,
      "loss": 0.0557,
      "step": 3572
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00010731628879241036,
      "loss": 0.0676,
      "step": 3573
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00010728412928123493,
      "loss": 0.0238,
      "step": 3574
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.0001072519697700595,
      "loss": 0.0654,
      "step": 3575
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00010721981025888408,
      "loss": 0.1082,
      "step": 3576
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00010718765074770865,
      "loss": 0.0219,
      "step": 3577
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.00010715549123653321,
      "loss": 0.0705,
      "step": 3578
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.00010712333172535778,
      "loss": 0.004,
      "step": 3579
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.00010709117221418235,
      "loss": 0.0497,
      "step": 3580
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0001070590127030069,
      "loss": 0.0288,
      "step": 3581
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0001070268531918315,
      "loss": 0.0069,
      "step": 3582
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.00010699469368065606,
      "loss": 0.0952,
      "step": 3583
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.00010696253416948063,
      "loss": 0.0187,
      "step": 3584
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0001069303746583052,
      "loss": 0.0955,
      "step": 3585
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00010689821514712977,
      "loss": 0.103,
      "step": 3586
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00010686605563595433,
      "loss": 0.0037,
      "step": 3587
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.0001068338961247789,
      "loss": 0.0026,
      "step": 3588
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00010680173661360348,
      "loss": 0.0206,
      "step": 3589
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00010676957710242805,
      "loss": 0.0178,
      "step": 3590
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00010673741759125262,
      "loss": 0.0173,
      "step": 3591
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.00010670525808007718,
      "loss": 0.0281,
      "step": 3592
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00010667309856890175,
      "loss": 0.0075,
      "step": 3593
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00010664093905772632,
      "loss": 0.0119,
      "step": 3594
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00010660877954655091,
      "loss": 0.0583,
      "step": 3595
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00010657662003537548,
      "loss": 0.1653,
      "step": 3596
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00010654446052420003,
      "loss": 0.0016,
      "step": 3597
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.0001065123010130246,
      "loss": 0.0715,
      "step": 3598
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.00010648014150184918,
      "loss": 0.0664,
      "step": 3599
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.00010644798199067375,
      "loss": 0.0658,
      "step": 3600
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0001064158224794983,
      "loss": 0.0613,
      "step": 3601
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0001063836629683229,
      "loss": 0.0577,
      "step": 3602
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.00010635150345714746,
      "loss": 0.0018,
      "step": 3603
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.00010631934394597203,
      "loss": 0.0065,
      "step": 3604
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0001062871844347966,
      "loss": 0.0005,
      "step": 3605
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.00010625502492362116,
      "loss": 0.013,
      "step": 3606
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010622286541244573,
      "loss": 0.0228,
      "step": 3607
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010619070590127031,
      "loss": 0.0197,
      "step": 3608
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010615854639009488,
      "loss": 0.0833,
      "step": 3609
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010612638687891945,
      "loss": 0.0176,
      "step": 3610
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010609422736774401,
      "loss": 0.0699,
      "step": 3611
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010606206785656858,
      "loss": 0.004,
      "step": 3612
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.00010602990834539315,
      "loss": 0.1128,
      "step": 3613
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.00010599774883421773,
      "loss": 0.0725,
      "step": 3614
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.0001059655893230423,
      "loss": 0.0675,
      "step": 3615
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.00010593342981186687,
      "loss": 0.015,
      "step": 3616
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.00010590127030069143,
      "loss": 0.0089,
      "step": 3617
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.000105869110789516,
      "loss": 0.0435,
      "step": 3618
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.00010583695127834057,
      "loss": 0.0175,
      "step": 3619
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.00010580479176716513,
      "loss": 0.0738,
      "step": 3620
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.00010577263225598973,
      "loss": 0.0852,
      "step": 3621
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.00010574047274481428,
      "loss": 0.0322,
      "step": 3622
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.00010570831323363885,
      "loss": 0.1085,
      "step": 3623
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.00010567615372246342,
      "loss": 0.0935,
      "step": 3624
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.000105643994211288,
      "loss": 0.046,
      "step": 3625
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.00010561183470011255,
      "loss": 0.0067,
      "step": 3626
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.00010557967518893715,
      "loss": 0.0695,
      "step": 3627
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.0001055475156777617,
      "loss": 0.0543,
      "step": 3628
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.00010551535616658628,
      "loss": 0.0783,
      "step": 3629
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.00010548319665541085,
      "loss": 0.1373,
      "step": 3630
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.0001054510371442354,
      "loss": 0.0126,
      "step": 3631
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.00010541887763305998,
      "loss": 0.0039,
      "step": 3632
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.00010538671812188455,
      "loss": 0.1631,
      "step": 3633
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00010535455861070913,
      "loss": 0.0278,
      "step": 3634
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0001053223990995337,
      "loss": 0.0014,
      "step": 3635
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00010529023958835826,
      "loss": 0.0963,
      "step": 3636
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00010525808007718283,
      "loss": 0.0739,
      "step": 3637
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0001052259205660074,
      "loss": 0.1128,
      "step": 3638
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00010519376105483197,
      "loss": 0.0658,
      "step": 3639
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.00010516160154365655,
      "loss": 0.0011,
      "step": 3640
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010512944203248112,
      "loss": 0.0293,
      "step": 3641
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010509728252130568,
      "loss": 0.0743,
      "step": 3642
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010506512301013025,
      "loss": 0.0627,
      "step": 3643
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010503296349895482,
      "loss": 0.0025,
      "step": 3644
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010500080398777938,
      "loss": 0.0079,
      "step": 3645
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010496864447660395,
      "loss": 0.0242,
      "step": 3646
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.00010493648496542853,
      "loss": 0.0104,
      "step": 3647
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.0001049043254542531,
      "loss": 0.0138,
      "step": 3648
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00010487216594307767,
      "loss": 0.0366,
      "step": 3649
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00010484000643190224,
      "loss": 0.0379,
      "step": 3650
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.0001048078469207268,
      "loss": 0.0279,
      "step": 3651
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00010477568740955137,
      "loss": 0.041,
      "step": 3652
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00010474352789837596,
      "loss": 0.0381,
      "step": 3653
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.00010471136838720053,
      "loss": 0.0122,
      "step": 3654
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0001046792088760251,
      "loss": 0.0011,
      "step": 3655
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.00010464704936484965,
      "loss": 0.0344,
      "step": 3656
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.00010461488985367422,
      "loss": 0.0058,
      "step": 3657
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0001045827303424988,
      "loss": 0.044,
      "step": 3658
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.00010455057083132335,
      "loss": 0.0014,
      "step": 3659
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.00010451841132014795,
      "loss": 0.0141,
      "step": 3660
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0001044862518089725,
      "loss": 0.0003,
      "step": 3661
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.00010445409229779708,
      "loss": 0.0139,
      "step": 3662
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.00010442193278662165,
      "loss": 0.1423,
      "step": 3663
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.00010438977327544622,
      "loss": 0.013,
      "step": 3664
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.00010435761376427077,
      "loss": 0.0166,
      "step": 3665
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.00010432545425309537,
      "loss": 0.0243,
      "step": 3666
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.00010429329474191993,
      "loss": 0.0272,
      "step": 3667
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0001042611352307445,
      "loss": 0.049,
      "step": 3668
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00010422897571956907,
      "loss": 0.0623,
      "step": 3669
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00010419681620839363,
      "loss": 0.0977,
      "step": 3670
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.0001041646566972182,
      "loss": 0.0042,
      "step": 3671
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00010413249718604278,
      "loss": 0.0055,
      "step": 3672
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00010410033767486735,
      "loss": 0.0009,
      "step": 3673
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.00010406817816369192,
      "loss": 0.0038,
      "step": 3674
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.0001040360186525165,
      "loss": 0.0582,
      "step": 3675
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00010400385914134105,
      "loss": 0.175,
      "step": 3676
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00010397169963016562,
      "loss": 0.0062,
      "step": 3677
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00010393954011899019,
      "loss": 0.0052,
      "step": 3678
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00010390738060781478,
      "loss": 0.0734,
      "step": 3679
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00010387522109663935,
      "loss": 0.1667,
      "step": 3680
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.0001038430615854639,
      "loss": 0.0698,
      "step": 3681
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.00010381090207428847,
      "loss": 0.0544,
      "step": 3682
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.00010377874256311304,
      "loss": 0.0926,
      "step": 3683
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0001037465830519376,
      "loss": 0.0192,
      "step": 3684
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0001037144235407622,
      "loss": 0.0342,
      "step": 3685
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.00010368226402958676,
      "loss": 0.0476,
      "step": 3686
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.00010365010451841133,
      "loss": 0.0613,
      "step": 3687
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0001036179450072359,
      "loss": 0.0038,
      "step": 3688
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.00010358578549606047,
      "loss": 0.066,
      "step": 3689
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00010355362598488502,
      "loss": 0.0728,
      "step": 3690
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.0001035214664737096,
      "loss": 0.0632,
      "step": 3691
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00010348930696253418,
      "loss": 0.016,
      "step": 3692
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00010345714745135875,
      "loss": 0.0028,
      "step": 3693
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00010342498794018332,
      "loss": 0.0226,
      "step": 3694
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00010339282842900788,
      "loss": 0.015,
      "step": 3695
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.00010336066891783245,
      "loss": 0.1165,
      "step": 3696
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.00010332850940665702,
      "loss": 0.1357,
      "step": 3697
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0001032963498954816,
      "loss": 0.0324,
      "step": 3698
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.00010326419038430617,
      "loss": 0.0636,
      "step": 3699
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.00010323203087313074,
      "loss": 0.1609,
      "step": 3700
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0001031998713619553,
      "loss": 0.0018,
      "step": 3701
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.00010316771185077987,
      "loss": 0.0273,
      "step": 3702
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.00010313555233960444,
      "loss": 0.0206,
      "step": 3703
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.000103103392828429,
      "loss": 0.0774,
      "step": 3704
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0001030712333172536,
      "loss": 0.0391,
      "step": 3705
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.00010303907380607815,
      "loss": 0.1055,
      "step": 3706
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.00010300691429490272,
      "loss": 0.049,
      "step": 3707
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0001029747547837273,
      "loss": 0.0161,
      "step": 3708
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.00010294259527255185,
      "loss": 0.0056,
      "step": 3709
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.00010291043576137642,
      "loss": 0.0835,
      "step": 3710
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.000102878276250201,
      "loss": 0.0106,
      "step": 3711
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.00010284611673902558,
      "loss": 0.0347,
      "step": 3712
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.00010281395722785015,
      "loss": 0.0724,
      "step": 3713
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.00010278179771667472,
      "loss": 0.0172,
      "step": 3714
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.00010274963820549927,
      "loss": 0.0116,
      "step": 3715
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.00010271747869432384,
      "loss": 0.05,
      "step": 3716
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.00010268531918314843,
      "loss": 0.0186,
      "step": 3717
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.000102653159671973,
      "loss": 0.0073,
      "step": 3718
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.00010262100016079757,
      "loss": 0.0396,
      "step": 3719
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.00010258884064962213,
      "loss": 0.0004,
      "step": 3720
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.0001025566811384467,
      "loss": 0.0648,
      "step": 3721
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.00010252452162727127,
      "loss": 0.0886,
      "step": 3722
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.00010249236211609582,
      "loss": 0.0335,
      "step": 3723
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010246020260492042,
      "loss": 0.0204,
      "step": 3724
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010242804309374498,
      "loss": 0.1017,
      "step": 3725
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010239588358256955,
      "loss": 0.0573,
      "step": 3726
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010236372407139412,
      "loss": 0.0275,
      "step": 3727
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010233156456021869,
      "loss": 0.0464,
      "step": 3728
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010229940504904325,
      "loss": 0.0062,
      "step": 3729
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.00010226724553786784,
      "loss": 0.1086,
      "step": 3730
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0001022350860266924,
      "loss": 0.006,
      "step": 3731
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.00010220292651551697,
      "loss": 0.0089,
      "step": 3732
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.00010217076700434154,
      "loss": 0.0508,
      "step": 3733
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0001021386074931661,
      "loss": 0.0007,
      "step": 3734
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.00010210644798199067,
      "loss": 0.0469,
      "step": 3735
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.00010207428847081524,
      "loss": 0.0042,
      "step": 3736
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.00010204212895963982,
      "loss": 0.001,
      "step": 3737
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.0001020099694484644,
      "loss": 0.094,
      "step": 3738
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00010197780993728897,
      "loss": 0.0014,
      "step": 3739
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00010194565042611352,
      "loss": 0.011,
      "step": 3740
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00010191349091493809,
      "loss": 0.0685,
      "step": 3741
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00010188133140376266,
      "loss": 0.0596,
      "step": 3742
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00010184917189258725,
      "loss": 0.0454,
      "step": 3743
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.00010181701238141182,
      "loss": 0.0569,
      "step": 3744
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00010178485287023638,
      "loss": 0.0476,
      "step": 3745
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00010175269335906095,
      "loss": 0.0287,
      "step": 3746
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00010172053384788552,
      "loss": 0.0228,
      "step": 3747
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00010168837433671007,
      "loss": 0.0137,
      "step": 3748
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00010165621482553464,
      "loss": 0.1128,
      "step": 3749
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.00010162405531435923,
      "loss": 0.0004,
      "step": 3750
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0001015918958031838,
      "loss": 0.0817,
      "step": 3751
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00010155973629200837,
      "loss": 0.0684,
      "step": 3752
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00010152757678083294,
      "loss": 0.0209,
      "step": 3753
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.0001014954172696575,
      "loss": 0.09,
      "step": 3754
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00010146325775848207,
      "loss": 0.1208,
      "step": 3755
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00010143109824730665,
      "loss": 0.0384,
      "step": 3756
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00010139893873613122,
      "loss": 0.0333,
      "step": 3757
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.00010136677922495579,
      "loss": 0.0222,
      "step": 3758
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010133461971378035,
      "loss": 0.0796,
      "step": 3759
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010130246020260492,
      "loss": 0.0061,
      "step": 3760
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010127030069142949,
      "loss": 0.0298,
      "step": 3761
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010123814118025407,
      "loss": 0.0225,
      "step": 3762
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010120598166907864,
      "loss": 0.0448,
      "step": 3763
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010117382215790321,
      "loss": 0.0302,
      "step": 3764
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.00010114166264672777,
      "loss": 0.1109,
      "step": 3765
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.00010110950313555234,
      "loss": 0.1968,
      "step": 3766
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.00010107734362437691,
      "loss": 0.0386,
      "step": 3767
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.00010104518411320147,
      "loss": 0.0024,
      "step": 3768
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.00010101302460202607,
      "loss": 0.0103,
      "step": 3769
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.00010098086509085062,
      "loss": 0.0546,
      "step": 3770
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.0001009487055796752,
      "loss": 0.0258,
      "step": 3771
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.00010091654606849977,
      "loss": 0.0323,
      "step": 3772
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010088438655732432,
      "loss": 0.0331,
      "step": 3773
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010085222704614889,
      "loss": 0.0568,
      "step": 3774
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010082006753497348,
      "loss": 0.0183,
      "step": 3775
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010078790802379805,
      "loss": 0.155,
      "step": 3776
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010075574851262262,
      "loss": 0.0251,
      "step": 3777
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010072358900144719,
      "loss": 0.042,
      "step": 3778
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.00010069142949027175,
      "loss": 0.0383,
      "step": 3779
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.00010065926997909632,
      "loss": 0.0033,
      "step": 3780
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.00010062711046792089,
      "loss": 0.0316,
      "step": 3781
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.00010059495095674547,
      "loss": 0.065,
      "step": 3782
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.00010056279144557004,
      "loss": 0.0416,
      "step": 3783
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.0001005306319343946,
      "loss": 0.0268,
      "step": 3784
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.00010049847242321917,
      "loss": 0.0988,
      "step": 3785
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00010046631291204374,
      "loss": 0.0322,
      "step": 3786
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.0001004341534008683,
      "loss": 0.0439,
      "step": 3787
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.0001004019938896929,
      "loss": 0.0184,
      "step": 3788
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00010036983437851746,
      "loss": 0.0398,
      "step": 3789
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00010033767486734202,
      "loss": 0.0322,
      "step": 3790
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00010030551535616659,
      "loss": 0.0725,
      "step": 3791
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.00010027335584499116,
      "loss": 0.0023,
      "step": 3792
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010024119633381572,
      "loss": 0.0107,
      "step": 3793
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010020903682264029,
      "loss": 0.1041,
      "step": 3794
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010017687731146487,
      "loss": 0.1348,
      "step": 3795
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010014471780028944,
      "loss": 0.0354,
      "step": 3796
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010011255828911401,
      "loss": 0.0713,
      "step": 3797
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010008039877793857,
      "loss": 0.0472,
      "step": 3798
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.00010004823926676314,
      "loss": 0.0009,
      "step": 3799
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.00010001607975558771,
      "loss": 0.042,
      "step": 3800
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.998392024441228e-05,
      "loss": 0.026,
      "step": 3801
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.995176073323685e-05,
      "loss": 0.0794,
      "step": 3802
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.991960122206144e-05,
      "loss": 0.0039,
      "step": 3803
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.9887441710886e-05,
      "loss": 0.1085,
      "step": 3804
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.985528219971057e-05,
      "loss": 0.0104,
      "step": 3805
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.982312268853514e-05,
      "loss": 0.0299,
      "step": 3806
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.97909631773597e-05,
      "loss": 0.064,
      "step": 3807
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.975880366618428e-05,
      "loss": 0.0356,
      "step": 3808
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.972664415500885e-05,
      "loss": 0.0124,
      "step": 3809
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.969448464383342e-05,
      "loss": 0.0448,
      "step": 3810
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.966232513265799e-05,
      "loss": 0.005,
      "step": 3811
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.963016562148256e-05,
      "loss": 0.028,
      "step": 3812
    },
    {
      "epoch": 5.51,
      "learning_rate": 9.959800611030713e-05,
      "loss": 0.0088,
      "step": 3813
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.95658465991317e-05,
      "loss": 0.0635,
      "step": 3814
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.953368708795627e-05,
      "loss": 0.0482,
      "step": 3815
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.950152757678084e-05,
      "loss": 0.0423,
      "step": 3816
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.946936806560541e-05,
      "loss": 0.0104,
      "step": 3817
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.943720855442997e-05,
      "loss": 0.0939,
      "step": 3818
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.940504904325455e-05,
      "loss": 0.1426,
      "step": 3819
    },
    {
      "epoch": 5.52,
      "learning_rate": 9.937288953207912e-05,
      "loss": 0.0778,
      "step": 3820
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.934073002090368e-05,
      "loss": 0.0005,
      "step": 3821
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.930857050972826e-05,
      "loss": 0.0166,
      "step": 3822
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.927641099855282e-05,
      "loss": 0.1753,
      "step": 3823
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.924425148737739e-05,
      "loss": 0.0354,
      "step": 3824
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.921209197620198e-05,
      "loss": 0.0719,
      "step": 3825
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.917993246502653e-05,
      "loss": 0.1179,
      "step": 3826
    },
    {
      "epoch": 5.53,
      "learning_rate": 9.91477729538511e-05,
      "loss": 0.036,
      "step": 3827
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.911561344267569e-05,
      "loss": 0.0023,
      "step": 3828
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.908345393150024e-05,
      "loss": 0.0628,
      "step": 3829
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.905129442032481e-05,
      "loss": 0.1398,
      "step": 3830
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.901913490914939e-05,
      "loss": 0.0011,
      "step": 3831
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.898697539797396e-05,
      "loss": 0.0019,
      "step": 3832
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.895481588679853e-05,
      "loss": 0.0627,
      "step": 3833
    },
    {
      "epoch": 5.54,
      "learning_rate": 9.89226563756231e-05,
      "loss": 0.0923,
      "step": 3834
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.889049686444767e-05,
      "loss": 0.1775,
      "step": 3835
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.885833735327224e-05,
      "loss": 0.0267,
      "step": 3836
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.88261778420968e-05,
      "loss": 0.0767,
      "step": 3837
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.879401833092138e-05,
      "loss": 0.0206,
      "step": 3838
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.876185881974595e-05,
      "loss": 0.0162,
      "step": 3839
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.87296993085705e-05,
      "loss": 0.1787,
      "step": 3840
    },
    {
      "epoch": 5.55,
      "learning_rate": 9.869753979739509e-05,
      "loss": 0.0177,
      "step": 3841
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.866538028621966e-05,
      "loss": 0.0903,
      "step": 3842
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.863322077504422e-05,
      "loss": 0.1145,
      "step": 3843
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.86010612638688e-05,
      "loss": 0.0725,
      "step": 3844
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.856890175269337e-05,
      "loss": 0.0339,
      "step": 3845
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.853674224151793e-05,
      "loss": 0.1057,
      "step": 3846
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.85045827303425e-05,
      "loss": 0.0336,
      "step": 3847
    },
    {
      "epoch": 5.56,
      "learning_rate": 9.847242321916707e-05,
      "loss": 0.0545,
      "step": 3848
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.844026370799164e-05,
      "loss": 0.096,
      "step": 3849
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.840810419681621e-05,
      "loss": 0.1375,
      "step": 3850
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.837594468564078e-05,
      "loss": 0.1203,
      "step": 3851
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.834378517446535e-05,
      "loss": 0.0724,
      "step": 3852
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.831162566328992e-05,
      "loss": 0.0542,
      "step": 3853
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.82794661521145e-05,
      "loss": 0.1094,
      "step": 3854
    },
    {
      "epoch": 5.57,
      "learning_rate": 9.824730664093906e-05,
      "loss": 0.0134,
      "step": 3855
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.821514712976363e-05,
      "loss": 0.0789,
      "step": 3856
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.81829876185882e-05,
      "loss": 0.0124,
      "step": 3857
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.815082810741278e-05,
      "loss": 0.0107,
      "step": 3858
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.811866859623735e-05,
      "loss": 0.063,
      "step": 3859
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.808650908506192e-05,
      "loss": 0.0774,
      "step": 3860
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.805434957388649e-05,
      "loss": 0.0945,
      "step": 3861
    },
    {
      "epoch": 5.58,
      "learning_rate": 9.802219006271104e-05,
      "loss": 0.0331,
      "step": 3862
    },
    {
      "epoch": 5.59,
      "learning_rate": 9.799003055153561e-05,
      "loss": 0.0111,
      "step": 3863
    },
    {
      "epoch": 5.59,
      "learning_rate": 9.79578710403602e-05,
      "loss": 0.0023,
      "step": 3864
    },
    {
      "epoch": 5.59,
      "learning_rate": 9.792571152918476e-05,
      "loss": 0.1092,
      "step": 3865
    },
    {
      "epoch": 5.59,
      "learning_rate": 9.789355201800933e-05,
      "loss": 0.021,
      "step": 3866
    },
    {
      "epoch": 5.59,
      "learning_rate": 9.786139250683391e-05,
      "loss": 0.0427,
      "step": 3867
    },
    {
      "epoch": 5.59,
      "learning_rate": 9.782923299565847e-05,
      "loss": 0.0223,
      "step": 3868
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.779707348448304e-05,
      "loss": 0.0439,
      "step": 3869
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.776491397330761e-05,
      "loss": 0.0004,
      "step": 3870
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.773275446213218e-05,
      "loss": 0.0463,
      "step": 3871
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.770059495095675e-05,
      "loss": 0.0428,
      "step": 3872
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.766843543978132e-05,
      "loss": 0.0007,
      "step": 3873
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.763627592860589e-05,
      "loss": 0.009,
      "step": 3874
    },
    {
      "epoch": 5.6,
      "learning_rate": 9.760411641743046e-05,
      "loss": 0.1479,
      "step": 3875
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.757195690625503e-05,
      "loss": 0.0605,
      "step": 3876
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.75397973950796e-05,
      "loss": 0.0679,
      "step": 3877
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.750763788390417e-05,
      "loss": 0.0319,
      "step": 3878
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.747547837272873e-05,
      "loss": 0.0075,
      "step": 3879
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.744331886155331e-05,
      "loss": 0.1591,
      "step": 3880
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.741115935037788e-05,
      "loss": 0.0093,
      "step": 3881
    },
    {
      "epoch": 5.61,
      "learning_rate": 9.737899983920244e-05,
      "loss": 0.005,
      "step": 3882
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.734684032802702e-05,
      "loss": 0.0053,
      "step": 3883
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.73146808168516e-05,
      "loss": 0.1351,
      "step": 3884
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.728252130567615e-05,
      "loss": 0.0707,
      "step": 3885
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.725036179450074e-05,
      "loss": 0.0108,
      "step": 3886
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.721820228332529e-05,
      "loss": 0.0157,
      "step": 3887
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.718604277214986e-05,
      "loss": 0.0488,
      "step": 3888
    },
    {
      "epoch": 5.62,
      "learning_rate": 9.715388326097445e-05,
      "loss": 0.051,
      "step": 3889
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.7121723749799e-05,
      "loss": 0.0181,
      "step": 3890
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.708956423862358e-05,
      "loss": 0.046,
      "step": 3891
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.705740472744815e-05,
      "loss": 0.0869,
      "step": 3892
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.702524521627272e-05,
      "loss": 0.0817,
      "step": 3893
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.699308570509729e-05,
      "loss": 0.0027,
      "step": 3894
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.696092619392186e-05,
      "loss": 0.0194,
      "step": 3895
    },
    {
      "epoch": 5.63,
      "learning_rate": 9.692876668274643e-05,
      "loss": 0.0079,
      "step": 3896
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.6896607171571e-05,
      "loss": 0.006,
      "step": 3897
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.686444766039557e-05,
      "loss": 0.0731,
      "step": 3898
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.683228814922014e-05,
      "loss": 0.0573,
      "step": 3899
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.680012863804471e-05,
      "loss": 0.1375,
      "step": 3900
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.676796912686927e-05,
      "loss": 0.0343,
      "step": 3901
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.673580961569385e-05,
      "loss": 0.1231,
      "step": 3902
    },
    {
      "epoch": 5.64,
      "learning_rate": 9.670365010451842e-05,
      "loss": 0.0695,
      "step": 3903
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.667149059334298e-05,
      "loss": 0.0255,
      "step": 3904
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.663933108216755e-05,
      "loss": 0.0656,
      "step": 3905
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.660717157099213e-05,
      "loss": 0.0264,
      "step": 3906
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.657501205981669e-05,
      "loss": 0.0685,
      "step": 3907
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.654285254864126e-05,
      "loss": 0.0936,
      "step": 3908
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.651069303746584e-05,
      "loss": 0.0148,
      "step": 3909
    },
    {
      "epoch": 5.65,
      "learning_rate": 9.64785335262904e-05,
      "loss": 0.0212,
      "step": 3910
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.644637401511497e-05,
      "loss": 0.0103,
      "step": 3911
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.641421450393954e-05,
      "loss": 0.001,
      "step": 3912
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.638205499276411e-05,
      "loss": 0.0766,
      "step": 3913
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.634989548158868e-05,
      "loss": 0.0121,
      "step": 3914
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.631773597041325e-05,
      "loss": 0.0839,
      "step": 3915
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.628557645923782e-05,
      "loss": 0.146,
      "step": 3916
    },
    {
      "epoch": 5.66,
      "learning_rate": 9.62534169480624e-05,
      "loss": 0.02,
      "step": 3917
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.622125743688697e-05,
      "loss": 0.0059,
      "step": 3918
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.618909792571154e-05,
      "loss": 0.0008,
      "step": 3919
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.61569384145361e-05,
      "loss": 0.1636,
      "step": 3920
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.612477890336066e-05,
      "loss": 0.0062,
      "step": 3921
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.609261939218525e-05,
      "loss": 0.0413,
      "step": 3922
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.606045988100982e-05,
      "loss": 0.0076,
      "step": 3923
    },
    {
      "epoch": 5.67,
      "learning_rate": 9.602830036983437e-05,
      "loss": 0.0714,
      "step": 3924
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.599614085865896e-05,
      "loss": 0.1099,
      "step": 3925
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.596398134748352e-05,
      "loss": 0.0954,
      "step": 3926
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.593182183630809e-05,
      "loss": 0.001,
      "step": 3927
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.589966232513267e-05,
      "loss": 0.0509,
      "step": 3928
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.586750281395723e-05,
      "loss": 0.0172,
      "step": 3929
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.58353433027818e-05,
      "loss": 0.0486,
      "step": 3930
    },
    {
      "epoch": 5.68,
      "learning_rate": 9.580318379160638e-05,
      "loss": 0.0169,
      "step": 3931
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.577102428043094e-05,
      "loss": 0.0935,
      "step": 3932
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.573886476925551e-05,
      "loss": 0.0384,
      "step": 3933
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.57067052580801e-05,
      "loss": 0.1775,
      "step": 3934
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.567454574690465e-05,
      "loss": 0.093,
      "step": 3935
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.564238623572922e-05,
      "loss": 0.0078,
      "step": 3936
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.561022672455379e-05,
      "loss": 0.0322,
      "step": 3937
    },
    {
      "epoch": 5.69,
      "learning_rate": 9.557806721337836e-05,
      "loss": 0.0169,
      "step": 3938
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.554590770220293e-05,
      "loss": 0.0044,
      "step": 3939
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.55137481910275e-05,
      "loss": 0.0286,
      "step": 3940
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.548158867985207e-05,
      "loss": 0.0123,
      "step": 3941
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.544942916867664e-05,
      "loss": 0.0192,
      "step": 3942
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.54172696575012e-05,
      "loss": 0.0638,
      "step": 3943
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.538511014632579e-05,
      "loss": 0.045,
      "step": 3944
    },
    {
      "epoch": 5.7,
      "learning_rate": 9.535295063515036e-05,
      "loss": 0.0082,
      "step": 3945
    },
    {
      "epoch": 5.71,
      "learning_rate": 9.532079112397491e-05,
      "loss": 0.0329,
      "step": 3946
    },
    {
      "epoch": 5.71,
      "learning_rate": 9.52886316127995e-05,
      "loss": 0.02,
      "step": 3947
    },
    {
      "epoch": 5.71,
      "learning_rate": 9.525647210162407e-05,
      "loss": 0.0005,
      "step": 3948
    },
    {
      "epoch": 5.71,
      "learning_rate": 9.522431259044862e-05,
      "loss": 0.0131,
      "step": 3949
    },
    {
      "epoch": 5.71,
      "learning_rate": 9.51921530792732e-05,
      "loss": 0.087,
      "step": 3950
    },
    {
      "epoch": 5.71,
      "learning_rate": 9.515999356809777e-05,
      "loss": 0.1202,
      "step": 3951
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.512783405692234e-05,
      "loss": 0.0343,
      "step": 3952
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.50956745457469e-05,
      "loss": 0.0952,
      "step": 3953
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.506351503457148e-05,
      "loss": 0.0333,
      "step": 3954
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.503135552339605e-05,
      "loss": 0.1196,
      "step": 3955
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.499919601222062e-05,
      "loss": 0.0555,
      "step": 3956
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.496703650104519e-05,
      "loss": 0.0209,
      "step": 3957
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.493487698986976e-05,
      "loss": 0.0521,
      "step": 3958
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.490271747869433e-05,
      "loss": 0.1033,
      "step": 3959
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.48705579675189e-05,
      "loss": 0.0678,
      "step": 3960
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.483839845634347e-05,
      "loss": 0.0147,
      "step": 3961
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.480623894516804e-05,
      "loss": 0.3096,
      "step": 3962
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.477407943399261e-05,
      "loss": 0.1233,
      "step": 3963
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.474191992281718e-05,
      "loss": 0.0362,
      "step": 3964
    },
    {
      "epoch": 5.73,
      "learning_rate": 9.470976041164175e-05,
      "loss": 0.0868,
      "step": 3965
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.467760090046631e-05,
      "loss": 0.0198,
      "step": 3966
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.46454413892909e-05,
      "loss": 0.0512,
      "step": 3967
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.461328187811545e-05,
      "loss": 0.0659,
      "step": 3968
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.458112236694002e-05,
      "loss": 0.0583,
      "step": 3969
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.45489628557646e-05,
      "loss": 0.0457,
      "step": 3970
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.451680334458916e-05,
      "loss": 0.0453,
      "step": 3971
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.448464383341373e-05,
      "loss": 0.1051,
      "step": 3972
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.445248432223832e-05,
      "loss": 0.0337,
      "step": 3973
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.442032481106287e-05,
      "loss": 0.0132,
      "step": 3974
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.438816529988744e-05,
      "loss": 0.0255,
      "step": 3975
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.435600578871201e-05,
      "loss": 0.0305,
      "step": 3976
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.432384627753659e-05,
      "loss": 0.0135,
      "step": 3977
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.429168676636116e-05,
      "loss": 0.0244,
      "step": 3978
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.425952725518573e-05,
      "loss": 0.0435,
      "step": 3979
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.42273677440103e-05,
      "loss": 0.0299,
      "step": 3980
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.419520823283487e-05,
      "loss": 0.0247,
      "step": 3981
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.416304872165942e-05,
      "loss": 0.0401,
      "step": 3982
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.413088921048401e-05,
      "loss": 0.092,
      "step": 3983
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.409872969930858e-05,
      "loss": 0.1001,
      "step": 3984
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.406657018813314e-05,
      "loss": 0.0417,
      "step": 3985
    },
    {
      "epoch": 5.76,
      "learning_rate": 9.403441067695772e-05,
      "loss": 0.1111,
      "step": 3986
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.400225116578229e-05,
      "loss": 0.0703,
      "step": 3987
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.397009165460685e-05,
      "loss": 0.0939,
      "step": 3988
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.393793214343143e-05,
      "loss": 0.1139,
      "step": 3989
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.3905772632256e-05,
      "loss": 0.0901,
      "step": 3990
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.387361312108056e-05,
      "loss": 0.0144,
      "step": 3991
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.384145360990514e-05,
      "loss": 0.012,
      "step": 3992
    },
    {
      "epoch": 5.77,
      "learning_rate": 9.38092940987297e-05,
      "loss": 0.017,
      "step": 3993
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.377713458755427e-05,
      "loss": 0.0543,
      "step": 3994
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.374497507637884e-05,
      "loss": 0.0774,
      "step": 3995
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.371281556520341e-05,
      "loss": 0.0413,
      "step": 3996
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.368065605402798e-05,
      "loss": 0.0659,
      "step": 3997
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.364849654285255e-05,
      "loss": 0.0488,
      "step": 3998
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.361633703167712e-05,
      "loss": 0.0495,
      "step": 3999
    },
    {
      "epoch": 5.78,
      "learning_rate": 9.35841775205017e-05,
      "loss": 0.0439,
      "step": 4000
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.355201800932626e-05,
      "loss": 0.0541,
      "step": 4001
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.351985849815083e-05,
      "loss": 0.1075,
      "step": 4002
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.34876989869754e-05,
      "loss": 0.0176,
      "step": 4003
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.345553947579998e-05,
      "loss": 0.038,
      "step": 4004
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.342337996462455e-05,
      "loss": 0.0551,
      "step": 4005
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.339122045344912e-05,
      "loss": 0.0279,
      "step": 4006
    },
    {
      "epoch": 5.79,
      "learning_rate": 9.335906094227367e-05,
      "loss": 0.0266,
      "step": 4007
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.332690143109826e-05,
      "loss": 0.0013,
      "step": 4008
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.329474191992283e-05,
      "loss": 0.0042,
      "step": 4009
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.326258240874738e-05,
      "loss": 0.0453,
      "step": 4010
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.323042289757196e-05,
      "loss": 0.0405,
      "step": 4011
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.319826338639654e-05,
      "loss": 0.0674,
      "step": 4012
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.31661038752211e-05,
      "loss": 0.0075,
      "step": 4013
    },
    {
      "epoch": 5.8,
      "learning_rate": 9.313394436404567e-05,
      "loss": 0.0465,
      "step": 4014
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.310178485287024e-05,
      "loss": 0.0237,
      "step": 4015
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.306962534169481e-05,
      "loss": 0.0253,
      "step": 4016
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.303746583051938e-05,
      "loss": 0.0052,
      "step": 4017
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.300530631934395e-05,
      "loss": 0.0436,
      "step": 4018
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.297314680816852e-05,
      "loss": 0.099,
      "step": 4019
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.294098729699309e-05,
      "loss": 0.0073,
      "step": 4020
    },
    {
      "epoch": 5.81,
      "learning_rate": 9.290882778581766e-05,
      "loss": 0.0417,
      "step": 4021
    },
    {
      "epoch": 5.82,
      "learning_rate": 9.287666827464223e-05,
      "loss": 0.0123,
      "step": 4022
    },
    {
      "epoch": 5.82,
      "learning_rate": 9.28445087634668e-05,
      "loss": 0.0198,
      "step": 4023
    },
    {
      "epoch": 5.82,
      "learning_rate": 9.281234925229137e-05,
      "loss": 0.0675,
      "step": 4024
    },
    {
      "epoch": 5.82,
      "learning_rate": 9.278018974111594e-05,
      "loss": 0.0874,
      "step": 4025
    },
    {
      "epoch": 5.82,
      "learning_rate": 9.274803022994051e-05,
      "loss": 0.103,
      "step": 4026
    },
    {
      "epoch": 5.82,
      "learning_rate": 9.271587071876507e-05,
      "loss": 0.003,
      "step": 4027
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.268371120758965e-05,
      "loss": 0.0515,
      "step": 4028
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.265155169641422e-05,
      "loss": 0.0028,
      "step": 4029
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.261939218523878e-05,
      "loss": 0.0136,
      "step": 4030
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.258723267406337e-05,
      "loss": 0.0036,
      "step": 4031
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.255507316288792e-05,
      "loss": 0.013,
      "step": 4032
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.252291365171249e-05,
      "loss": 0.0633,
      "step": 4033
    },
    {
      "epoch": 5.83,
      "learning_rate": 9.249075414053708e-05,
      "loss": 0.0088,
      "step": 4034
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.245859462936163e-05,
      "loss": 0.0145,
      "step": 4035
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.24264351181862e-05,
      "loss": 0.0431,
      "step": 4036
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.239427560701079e-05,
      "loss": 0.0009,
      "step": 4037
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.236211609583535e-05,
      "loss": 0.0764,
      "step": 4038
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.232995658465992e-05,
      "loss": 0.0778,
      "step": 4039
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.229779707348449e-05,
      "loss": 0.0015,
      "step": 4040
    },
    {
      "epoch": 5.84,
      "learning_rate": 9.226563756230906e-05,
      "loss": 0.0788,
      "step": 4041
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.223347805113363e-05,
      "loss": 0.0288,
      "step": 4042
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.22013185399582e-05,
      "loss": 0.0029,
      "step": 4043
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.216915902878277e-05,
      "loss": 0.0494,
      "step": 4044
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.213699951760734e-05,
      "loss": 0.0545,
      "step": 4045
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.21048400064319e-05,
      "loss": 0.0052,
      "step": 4046
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.207268049525648e-05,
      "loss": 0.0007,
      "step": 4047
    },
    {
      "epoch": 5.85,
      "learning_rate": 9.204052098408105e-05,
      "loss": 0.0311,
      "step": 4048
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.200836147290561e-05,
      "loss": 0.0019,
      "step": 4049
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.197620196173019e-05,
      "loss": 0.0493,
      "step": 4050
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.194404245055476e-05,
      "loss": 0.0469,
      "step": 4051
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.191188293937932e-05,
      "loss": 0.0005,
      "step": 4052
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.18797234282039e-05,
      "loss": 0.0133,
      "step": 4053
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.184756391702847e-05,
      "loss": 0.0082,
      "step": 4054
    },
    {
      "epoch": 5.86,
      "learning_rate": 9.181540440585303e-05,
      "loss": 0.1074,
      "step": 4055
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.17832448946776e-05,
      "loss": 0.0155,
      "step": 4056
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.175108538350217e-05,
      "loss": 0.0316,
      "step": 4057
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.171892587232674e-05,
      "loss": 0.0285,
      "step": 4058
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.168676636115131e-05,
      "loss": 0.0782,
      "step": 4059
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.165460684997588e-05,
      "loss": 0.1307,
      "step": 4060
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.162244733880045e-05,
      "loss": 0.0977,
      "step": 4061
    },
    {
      "epoch": 5.87,
      "learning_rate": 9.159028782762502e-05,
      "loss": 0.0294,
      "step": 4062
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.15581283164496e-05,
      "loss": 0.0175,
      "step": 4063
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.152596880527417e-05,
      "loss": 0.0576,
      "step": 4064
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.149380929409874e-05,
      "loss": 0.0012,
      "step": 4065
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.14616497829233e-05,
      "loss": 0.0708,
      "step": 4066
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.142949027174788e-05,
      "loss": 0.0037,
      "step": 4067
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.139733076057245e-05,
      "loss": 0.0989,
      "step": 4068
    },
    {
      "epoch": 5.88,
      "learning_rate": 9.1365171249397e-05,
      "loss": 0.1206,
      "step": 4069
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.133301173822159e-05,
      "loss": 0.0879,
      "step": 4070
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.130085222704615e-05,
      "loss": 0.0039,
      "step": 4071
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.126869271587072e-05,
      "loss": 0.018,
      "step": 4072
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.12365332046953e-05,
      "loss": 0.0397,
      "step": 4073
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.120437369351986e-05,
      "loss": 0.0534,
      "step": 4074
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.117221418234443e-05,
      "loss": 0.0229,
      "step": 4075
    },
    {
      "epoch": 5.89,
      "learning_rate": 9.114005467116901e-05,
      "loss": 0.0095,
      "step": 4076
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.110789515999357e-05,
      "loss": 0.0817,
      "step": 4077
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.107573564881814e-05,
      "loss": 0.0783,
      "step": 4078
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.104357613764272e-05,
      "loss": 0.0346,
      "step": 4079
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.101141662646728e-05,
      "loss": 0.0791,
      "step": 4080
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.097925711529185e-05,
      "loss": 0.0457,
      "step": 4081
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.094709760411642e-05,
      "loss": 0.0386,
      "step": 4082
    },
    {
      "epoch": 5.9,
      "learning_rate": 9.091493809294099e-05,
      "loss": 0.0472,
      "step": 4083
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.088277858176556e-05,
      "loss": 0.0752,
      "step": 4084
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.085061907059013e-05,
      "loss": 0.0874,
      "step": 4085
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.08184595594147e-05,
      "loss": 0.0231,
      "step": 4086
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.078630004823927e-05,
      "loss": 0.0179,
      "step": 4087
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.075414053706383e-05,
      "loss": 0.0738,
      "step": 4088
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.072198102588841e-05,
      "loss": 0.0272,
      "step": 4089
    },
    {
      "epoch": 5.91,
      "learning_rate": 9.068982151471299e-05,
      "loss": 0.0848,
      "step": 4090
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.065766200353754e-05,
      "loss": 0.0175,
      "step": 4091
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.062550249236213e-05,
      "loss": 0.0896,
      "step": 4092
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.05933429811867e-05,
      "loss": 0.0064,
      "step": 4093
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.056118347001125e-05,
      "loss": 0.1038,
      "step": 4094
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.052902395883584e-05,
      "loss": 0.1006,
      "step": 4095
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.04968644476604e-05,
      "loss": 0.1094,
      "step": 4096
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.046470493648497e-05,
      "loss": 0.015,
      "step": 4097
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.043254542530955e-05,
      "loss": 0.0579,
      "step": 4098
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.04003859141341e-05,
      "loss": 0.02,
      "step": 4099
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.036822640295868e-05,
      "loss": 0.0903,
      "step": 4100
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.033606689178325e-05,
      "loss": 0.0065,
      "step": 4101
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.030390738060782e-05,
      "loss": 0.085,
      "step": 4102
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.027174786943239e-05,
      "loss": 0.0085,
      "step": 4103
    },
    {
      "epoch": 5.93,
      "learning_rate": 9.023958835825696e-05,
      "loss": 0.0375,
      "step": 4104
    },
    {
      "epoch": 5.94,
      "learning_rate": 9.020742884708153e-05,
      "loss": 0.032,
      "step": 4105
    },
    {
      "epoch": 5.94,
      "learning_rate": 9.01752693359061e-05,
      "loss": 0.1562,
      "step": 4106
    },
    {
      "epoch": 5.94,
      "learning_rate": 9.014310982473067e-05,
      "loss": 0.0633,
      "step": 4107
    },
    {
      "epoch": 5.94,
      "learning_rate": 9.011095031355524e-05,
      "loss": 0.0037,
      "step": 4108
    },
    {
      "epoch": 5.94,
      "learning_rate": 9.007879080237981e-05,
      "loss": 0.0053,
      "step": 4109
    },
    {
      "epoch": 5.94,
      "learning_rate": 9.004663129120438e-05,
      "loss": 0.1316,
      "step": 4110
    },
    {
      "epoch": 5.95,
      "learning_rate": 9.001447178002895e-05,
      "loss": 0.0345,
      "step": 4111
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.998231226885352e-05,
      "loss": 0.0282,
      "step": 4112
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.995015275767808e-05,
      "loss": 0.0178,
      "step": 4113
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.991799324650265e-05,
      "loss": 0.1552,
      "step": 4114
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.988583373532723e-05,
      "loss": 0.0069,
      "step": 4115
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.985367422415179e-05,
      "loss": 0.0261,
      "step": 4116
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.982151471297636e-05,
      "loss": 0.0876,
      "step": 4117
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.978935520180095e-05,
      "loss": 0.0541,
      "step": 4118
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.97571956906255e-05,
      "loss": 0.0661,
      "step": 4119
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.972503617945007e-05,
      "loss": 0.0376,
      "step": 4120
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.969287666827464e-05,
      "loss": 0.0121,
      "step": 4121
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.966071715709921e-05,
      "loss": 0.0403,
      "step": 4122
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.962855764592378e-05,
      "loss": 0.0569,
      "step": 4123
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.959639813474836e-05,
      "loss": 0.0113,
      "step": 4124
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.956423862357293e-05,
      "loss": 0.1667,
      "step": 4125
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.95320791123975e-05,
      "loss": 0.0543,
      "step": 4126
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.949991960122207e-05,
      "loss": 0.0787,
      "step": 4127
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.946776009004664e-05,
      "loss": 0.0945,
      "step": 4128
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.943560057887121e-05,
      "loss": 0.0522,
      "step": 4129
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.940344106769576e-05,
      "loss": 0.071,
      "step": 4130
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.937128155652035e-05,
      "loss": 0.0023,
      "step": 4131
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.933912204534492e-05,
      "loss": 0.0134,
      "step": 4132
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.930696253416948e-05,
      "loss": 0.1431,
      "step": 4133
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.927480302299406e-05,
      "loss": 0.0663,
      "step": 4134
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.924264351181863e-05,
      "loss": 0.1484,
      "step": 4135
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.921048400064319e-05,
      "loss": 0.0338,
      "step": 4136
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.917832448946777e-05,
      "loss": 0.0256,
      "step": 4137
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.914616497829233e-05,
      "loss": 0.0184,
      "step": 4138
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.91140054671169e-05,
      "loss": 0.1091,
      "step": 4139
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.908184595594148e-05,
      "loss": 0.0308,
      "step": 4140
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.904968644476604e-05,
      "loss": 0.115,
      "step": 4141
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.901752693359061e-05,
      "loss": 0.0253,
      "step": 4142
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.89853674224152e-05,
      "loss": 0.0158,
      "step": 4143
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.895320791123975e-05,
      "loss": 0.0548,
      "step": 4144
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.892104840006432e-05,
      "loss": 0.0258,
      "step": 4145
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.888888888888889e-05,
      "loss": 0.0777,
      "step": 4146
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.885672937771346e-05,
      "loss": 0.0923,
      "step": 4147
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.882456986653803e-05,
      "loss": 0.0565,
      "step": 4148
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.87924103553626e-05,
      "loss": 0.0017,
      "step": 4149
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.876025084418718e-05,
      "loss": 0.0086,
      "step": 4150
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.872809133301175e-05,
      "loss": 0.0111,
      "step": 4151
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.86959318218363e-05,
      "loss": 0.0059,
      "step": 4152
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.866377231066089e-05,
      "loss": 0.0173,
      "step": 4153
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.863161279948546e-05,
      "loss": 0.0042,
      "step": 4154
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.859945328831001e-05,
      "loss": 0.0256,
      "step": 4155
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.85672937771346e-05,
      "loss": 0.0371,
      "step": 4156
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.853513426595917e-05,
      "loss": 0.004,
      "step": 4157
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.850297475478373e-05,
      "loss": 0.0002,
      "step": 4158
    },
    {
      "epoch": 6.01,
      "learning_rate": 8.84708152436083e-05,
      "loss": 0.0013,
      "step": 4159
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.843865573243287e-05,
      "loss": 0.0146,
      "step": 4160
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.840649622125744e-05,
      "loss": 0.0254,
      "step": 4161
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.837433671008201e-05,
      "loss": 0.0113,
      "step": 4162
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.834217719890658e-05,
      "loss": 0.0507,
      "step": 4163
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.831001768773115e-05,
      "loss": 0.0594,
      "step": 4164
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.827785817655572e-05,
      "loss": 0.0524,
      "step": 4165
    },
    {
      "epoch": 6.02,
      "learning_rate": 8.824569866538029e-05,
      "loss": 0.0481,
      "step": 4166
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.821353915420486e-05,
      "loss": 0.0177,
      "step": 4167
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.818137964302943e-05,
      "loss": 0.0452,
      "step": 4168
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.8149220131854e-05,
      "loss": 0.0389,
      "step": 4169
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.811706062067857e-05,
      "loss": 0.057,
      "step": 4170
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.808490110950314e-05,
      "loss": 0.0012,
      "step": 4171
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.805274159832771e-05,
      "loss": 0.0026,
      "step": 4172
    },
    {
      "epoch": 6.03,
      "learning_rate": 8.802058208715228e-05,
      "loss": 0.0062,
      "step": 4173
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.798842257597685e-05,
      "loss": 0.0345,
      "step": 4174
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.795626306480141e-05,
      "loss": 0.0843,
      "step": 4175
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.7924103553626e-05,
      "loss": 0.0494,
      "step": 4176
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.789194404245055e-05,
      "loss": 0.0333,
      "step": 4177
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.785978453127512e-05,
      "loss": 0.0086,
      "step": 4178
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.78276250200997e-05,
      "loss": 0.0021,
      "step": 4179
    },
    {
      "epoch": 6.04,
      "learning_rate": 8.779546550892426e-05,
      "loss": 0.004,
      "step": 4180
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.776330599774883e-05,
      "loss": 0.0093,
      "step": 4181
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.773114648657342e-05,
      "loss": 0.0341,
      "step": 4182
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.769898697539797e-05,
      "loss": 0.0622,
      "step": 4183
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.766682746422255e-05,
      "loss": 0.0159,
      "step": 4184
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.763466795304712e-05,
      "loss": 0.0172,
      "step": 4185
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.760250844187169e-05,
      "loss": 0.0492,
      "step": 4186
    },
    {
      "epoch": 6.05,
      "learning_rate": 8.757034893069626e-05,
      "loss": 0.0633,
      "step": 4187
    },
    {
      "epoch": 6.06,
      "learning_rate": 8.753818941952083e-05,
      "loss": 0.001,
      "step": 4188
    },
    {
      "epoch": 6.06,
      "learning_rate": 8.75060299083454e-05,
      "loss": 0.0452,
      "step": 4189
    },
    {
      "epoch": 6.06,
      "learning_rate": 8.747387039716997e-05,
      "loss": 0.0029,
      "step": 4190
    },
    {
      "epoch": 6.06,
      "learning_rate": 8.744171088599453e-05,
      "loss": 0.013,
      "step": 4191
    },
    {
      "epoch": 6.06,
      "learning_rate": 8.740955137481911e-05,
      "loss": 0.028,
      "step": 4192
    },
    {
      "epoch": 6.06,
      "learning_rate": 8.737739186364368e-05,
      "loss": 0.0095,
      "step": 4193
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.734523235246824e-05,
      "loss": 0.0014,
      "step": 4194
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.731307284129282e-05,
      "loss": 0.0235,
      "step": 4195
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.728091333011739e-05,
      "loss": 0.0003,
      "step": 4196
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.724875381894195e-05,
      "loss": 0.0123,
      "step": 4197
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.721659430776653e-05,
      "loss": 0.0081,
      "step": 4198
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.71844347965911e-05,
      "loss": 0.0053,
      "step": 4199
    },
    {
      "epoch": 6.07,
      "learning_rate": 8.715227528541566e-05,
      "loss": 0.05,
      "step": 4200
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.712011577424024e-05,
      "loss": 0.0035,
      "step": 4201
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.70879562630648e-05,
      "loss": 0.0033,
      "step": 4202
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.705579675188937e-05,
      "loss": 0.0796,
      "step": 4203
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.702363724071394e-05,
      "loss": 0.0023,
      "step": 4204
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.699147772953851e-05,
      "loss": 0.0051,
      "step": 4205
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.695931821836308e-05,
      "loss": 0.0036,
      "step": 4206
    },
    {
      "epoch": 6.08,
      "learning_rate": 8.692715870718765e-05,
      "loss": 0.0483,
      "step": 4207
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.689499919601222e-05,
      "loss": 0.044,
      "step": 4208
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.68628396848368e-05,
      "loss": 0.0312,
      "step": 4209
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.683068017366137e-05,
      "loss": 0.0537,
      "step": 4210
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.679852066248594e-05,
      "loss": 0.0945,
      "step": 4211
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.67663611513105e-05,
      "loss": 0.0078,
      "step": 4212
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.673420164013508e-05,
      "loss": 0.0635,
      "step": 4213
    },
    {
      "epoch": 6.09,
      "learning_rate": 8.670204212895965e-05,
      "loss": 0.0618,
      "step": 4214
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.666988261778422e-05,
      "loss": 0.0465,
      "step": 4215
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.663772310660877e-05,
      "loss": 0.0021,
      "step": 4216
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.660556359543336e-05,
      "loss": 0.0224,
      "step": 4217
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.657340408425793e-05,
      "loss": 0.033,
      "step": 4218
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.654124457308249e-05,
      "loss": 0.0623,
      "step": 4219
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.650908506190706e-05,
      "loss": 0.0273,
      "step": 4220
    },
    {
      "epoch": 6.1,
      "learning_rate": 8.647692555073164e-05,
      "loss": 0.0501,
      "step": 4221
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.64447660395562e-05,
      "loss": 0.0297,
      "step": 4222
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.641260652838077e-05,
      "loss": 0.0002,
      "step": 4223
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.638044701720535e-05,
      "loss": 0.0103,
      "step": 4224
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.634828750602991e-05,
      "loss": 0.0701,
      "step": 4225
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.631612799485448e-05,
      "loss": 0.0225,
      "step": 4226
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.628396848367905e-05,
      "loss": 0.0564,
      "step": 4227
    },
    {
      "epoch": 6.11,
      "learning_rate": 8.625180897250362e-05,
      "loss": 0.0698,
      "step": 4228
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.621964946132819e-05,
      "loss": 0.0581,
      "step": 4229
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.618748995015276e-05,
      "loss": 0.0164,
      "step": 4230
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.615533043897733e-05,
      "loss": 0.1338,
      "step": 4231
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.61231709278019e-05,
      "loss": 0.1035,
      "step": 4232
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.609101141662646e-05,
      "loss": 0.0101,
      "step": 4233
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.605885190545104e-05,
      "loss": 0.0017,
      "step": 4234
    },
    {
      "epoch": 6.12,
      "learning_rate": 8.602669239427561e-05,
      "loss": 0.0366,
      "step": 4235
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.599453288310017e-05,
      "loss": 0.0696,
      "step": 4236
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.596237337192476e-05,
      "loss": 0.0594,
      "step": 4237
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.593021386074933e-05,
      "loss": 0.0298,
      "step": 4238
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.589805434957388e-05,
      "loss": 0.018,
      "step": 4239
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.586589483839847e-05,
      "loss": 0.0037,
      "step": 4240
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.583373532722302e-05,
      "loss": 0.0101,
      "step": 4241
    },
    {
      "epoch": 6.13,
      "learning_rate": 8.58015758160476e-05,
      "loss": 0.0196,
      "step": 4242
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.576941630487218e-05,
      "loss": 0.0289,
      "step": 4243
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.573725679369674e-05,
      "loss": 0.0147,
      "step": 4244
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.57050972825213e-05,
      "loss": 0.0352,
      "step": 4245
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.567293777134589e-05,
      "loss": 0.041,
      "step": 4246
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.564077826017045e-05,
      "loss": 0.1059,
      "step": 4247
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.560861874899502e-05,
      "loss": 0.006,
      "step": 4248
    },
    {
      "epoch": 6.14,
      "learning_rate": 8.557645923781959e-05,
      "loss": 0.0121,
      "step": 4249
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.554429972664416e-05,
      "loss": 0.0084,
      "step": 4250
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.551214021546873e-05,
      "loss": 0.0441,
      "step": 4251
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.54799807042933e-05,
      "loss": 0.0009,
      "step": 4252
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.544782119311787e-05,
      "loss": 0.0025,
      "step": 4253
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.541566168194244e-05,
      "loss": 0.0052,
      "step": 4254
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.538350217076701e-05,
      "loss": 0.0588,
      "step": 4255
    },
    {
      "epoch": 6.15,
      "learning_rate": 8.535134265959158e-05,
      "loss": 0.0263,
      "step": 4256
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.531918314841615e-05,
      "loss": 0.0477,
      "step": 4257
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.528702363724071e-05,
      "loss": 0.0952,
      "step": 4258
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.52548641260653e-05,
      "loss": 0.0071,
      "step": 4259
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.522270461488986e-05,
      "loss": 0.0804,
      "step": 4260
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.519054510371442e-05,
      "loss": 0.0199,
      "step": 4261
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.5158385592539e-05,
      "loss": 0.0429,
      "step": 4262
    },
    {
      "epoch": 6.16,
      "learning_rate": 8.512622608136358e-05,
      "loss": 0.0523,
      "step": 4263
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.509406657018813e-05,
      "loss": 0.0538,
      "step": 4264
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.50619070590127e-05,
      "loss": 0.0286,
      "step": 4265
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.502974754783727e-05,
      "loss": 0.0084,
      "step": 4266
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.499758803666184e-05,
      "loss": 0.0813,
      "step": 4267
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.496542852548641e-05,
      "loss": 0.0345,
      "step": 4268
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.493326901431098e-05,
      "loss": 0.1299,
      "step": 4269
    },
    {
      "epoch": 6.17,
      "learning_rate": 8.490110950313556e-05,
      "loss": 0.0033,
      "step": 4270
    },
    {
      "epoch": 6.18,
      "learning_rate": 8.486894999196013e-05,
      "loss": 0.0126,
      "step": 4271
    },
    {
      "epoch": 6.18,
      "learning_rate": 8.48367904807847e-05,
      "loss": 0.0123,
      "step": 4272
    },
    {
      "epoch": 6.18,
      "learning_rate": 8.480463096960927e-05,
      "loss": 0.0803,
      "step": 4273
    },
    {
      "epoch": 6.18,
      "learning_rate": 8.477247145843384e-05,
      "loss": 0.032,
      "step": 4274
    },
    {
      "epoch": 6.18,
      "learning_rate": 8.474031194725841e-05,
      "loss": 0.0294,
      "step": 4275
    },
    {
      "epoch": 6.18,
      "learning_rate": 8.470815243608298e-05,
      "loss": 0.0107,
      "step": 4276
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.467599292490755e-05,
      "loss": 0.0504,
      "step": 4277
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.46438334137321e-05,
      "loss": 0.0654,
      "step": 4278
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.461167390255669e-05,
      "loss": 0.0236,
      "step": 4279
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.457951439138126e-05,
      "loss": 0.1341,
      "step": 4280
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.454735488020582e-05,
      "loss": 0.0062,
      "step": 4281
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.45151953690304e-05,
      "loss": 0.0582,
      "step": 4282
    },
    {
      "epoch": 6.19,
      "learning_rate": 8.448303585785496e-05,
      "loss": 0.0496,
      "step": 4283
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.445087634667953e-05,
      "loss": 0.0925,
      "step": 4284
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.441871683550411e-05,
      "loss": 0.0791,
      "step": 4285
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.438655732432867e-05,
      "loss": 0.0048,
      "step": 4286
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.435439781315324e-05,
      "loss": 0.014,
      "step": 4287
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.432223830197782e-05,
      "loss": 0.0074,
      "step": 4288
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.429007879080238e-05,
      "loss": 0.0079,
      "step": 4289
    },
    {
      "epoch": 6.2,
      "learning_rate": 8.425791927962695e-05,
      "loss": 0.0145,
      "step": 4290
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.422575976845152e-05,
      "loss": 0.0217,
      "step": 4291
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.419360025727609e-05,
      "loss": 0.0053,
      "step": 4292
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.416144074610066e-05,
      "loss": 0.0504,
      "step": 4293
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.412928123492523e-05,
      "loss": 0.0085,
      "step": 4294
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.40971217237498e-05,
      "loss": 0.0401,
      "step": 4295
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.406496221257438e-05,
      "loss": 0.0022,
      "step": 4296
    },
    {
      "epoch": 6.21,
      "learning_rate": 8.403280270139893e-05,
      "loss": 0.0296,
      "step": 4297
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.400064319022352e-05,
      "loss": 0.0032,
      "step": 4298
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.396848367904809e-05,
      "loss": 0.0001,
      "step": 4299
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.393632416787264e-05,
      "loss": 0.0699,
      "step": 4300
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.390416465669723e-05,
      "loss": 0.0702,
      "step": 4301
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.38720051455218e-05,
      "loss": 0.0135,
      "step": 4302
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.383984563434636e-05,
      "loss": 0.0008,
      "step": 4303
    },
    {
      "epoch": 6.22,
      "learning_rate": 8.380768612317094e-05,
      "loss": 0.0453,
      "step": 4304
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.37755266119955e-05,
      "loss": 0.0017,
      "step": 4305
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.374336710082007e-05,
      "loss": 0.0894,
      "step": 4306
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.371120758964465e-05,
      "loss": 0.0066,
      "step": 4307
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.367904807846921e-05,
      "loss": 0.0378,
      "step": 4308
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.364688856729378e-05,
      "loss": 0.1074,
      "step": 4309
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.361472905611835e-05,
      "loss": 0.0005,
      "step": 4310
    },
    {
      "epoch": 6.23,
      "learning_rate": 8.358256954494292e-05,
      "loss": 0.0576,
      "step": 4311
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.355041003376749e-05,
      "loss": 0.0828,
      "step": 4312
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.351825052259206e-05,
      "loss": 0.0005,
      "step": 4313
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.348609101141663e-05,
      "loss": 0.0073,
      "step": 4314
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.34539315002412e-05,
      "loss": 0.1302,
      "step": 4315
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.342177198906577e-05,
      "loss": 0.0004,
      "step": 4316
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.338961247789034e-05,
      "loss": 0.0071,
      "step": 4317
    },
    {
      "epoch": 6.24,
      "learning_rate": 8.335745296671491e-05,
      "loss": 0.0519,
      "step": 4318
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.332529345553948e-05,
      "loss": 0.0285,
      "step": 4319
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.329313394436405e-05,
      "loss": 0.0077,
      "step": 4320
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.326097443318862e-05,
      "loss": 0.0914,
      "step": 4321
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.322881492201318e-05,
      "loss": 0.0055,
      "step": 4322
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.319665541083775e-05,
      "loss": 0.063,
      "step": 4323
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.316449589966234e-05,
      "loss": 0.023,
      "step": 4324
    },
    {
      "epoch": 6.25,
      "learning_rate": 8.313233638848689e-05,
      "loss": 0.0024,
      "step": 4325
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.310017687731146e-05,
      "loss": 0.0941,
      "step": 4326
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.306801736613605e-05,
      "loss": 0.0193,
      "step": 4327
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.30358578549606e-05,
      "loss": 0.0004,
      "step": 4328
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.300369834378517e-05,
      "loss": 0.0638,
      "step": 4329
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.297153883260975e-05,
      "loss": 0.0648,
      "step": 4330
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.293937932143432e-05,
      "loss": 0.0677,
      "step": 4331
    },
    {
      "epoch": 6.26,
      "learning_rate": 8.290721981025889e-05,
      "loss": 0.0474,
      "step": 4332
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.287506029908346e-05,
      "loss": 0.0046,
      "step": 4333
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.284290078790803e-05,
      "loss": 0.0063,
      "step": 4334
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.28107412767326e-05,
      "loss": 0.045,
      "step": 4335
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.277858176555717e-05,
      "loss": 0.0062,
      "step": 4336
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.274642225438174e-05,
      "loss": 0.0335,
      "step": 4337
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.271426274320631e-05,
      "loss": 0.0229,
      "step": 4338
    },
    {
      "epoch": 6.27,
      "learning_rate": 8.268210323203087e-05,
      "loss": 0.0336,
      "step": 4339
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.264994372085545e-05,
      "loss": 0.0768,
      "step": 4340
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.261778420968002e-05,
      "loss": 0.0133,
      "step": 4341
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.258562469850458e-05,
      "loss": 0.001,
      "step": 4342
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.255346518732916e-05,
      "loss": 0.0367,
      "step": 4343
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.252130567615373e-05,
      "loss": 0.0186,
      "step": 4344
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.248914616497829e-05,
      "loss": 0.0868,
      "step": 4345
    },
    {
      "epoch": 6.28,
      "learning_rate": 8.245698665380287e-05,
      "loss": 0.1079,
      "step": 4346
    },
    {
      "epoch": 6.29,
      "learning_rate": 8.242482714262743e-05,
      "loss": 0.1953,
      "step": 4347
    },
    {
      "epoch": 6.29,
      "learning_rate": 8.2392667631452e-05,
      "loss": 0.0463,
      "step": 4348
    },
    {
      "epoch": 6.29,
      "learning_rate": 8.236050812027659e-05,
      "loss": 0.0155,
      "step": 4349
    },
    {
      "epoch": 6.29,
      "learning_rate": 8.232834860910114e-05,
      "loss": 0.001,
      "step": 4350
    },
    {
      "epoch": 6.29,
      "learning_rate": 8.229618909792571e-05,
      "loss": 0.04,
      "step": 4351
    },
    {
      "epoch": 6.29,
      "learning_rate": 8.22640295867503e-05,
      "loss": 0.0059,
      "step": 4352
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.223187007557485e-05,
      "loss": 0.0024,
      "step": 4353
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.219971056439942e-05,
      "loss": 0.0022,
      "step": 4354
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.2167551053224e-05,
      "loss": 0.0793,
      "step": 4355
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.213539154204857e-05,
      "loss": 0.0113,
      "step": 4356
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.210323203087314e-05,
      "loss": 0.0513,
      "step": 4357
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.20710725196977e-05,
      "loss": 0.0318,
      "step": 4358
    },
    {
      "epoch": 6.3,
      "learning_rate": 8.203891300852228e-05,
      "loss": 0.0026,
      "step": 4359
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.200675349734685e-05,
      "loss": 0.001,
      "step": 4360
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.19745939861714e-05,
      "loss": 0.0275,
      "step": 4361
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.194243447499599e-05,
      "loss": 0.0701,
      "step": 4362
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.191027496382056e-05,
      "loss": 0.0083,
      "step": 4363
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.187811545264512e-05,
      "loss": 0.0927,
      "step": 4364
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.18459559414697e-05,
      "loss": 0.0031,
      "step": 4365
    },
    {
      "epoch": 6.31,
      "learning_rate": 8.181379643029427e-05,
      "loss": 0.0457,
      "step": 4366
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.178163691911883e-05,
      "loss": 0.0011,
      "step": 4367
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.17494774079434e-05,
      "loss": 0.0103,
      "step": 4368
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.171731789676798e-05,
      "loss": 0.0547,
      "step": 4369
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.168515838559254e-05,
      "loss": 0.0073,
      "step": 4370
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.165299887441711e-05,
      "loss": 0.0151,
      "step": 4371
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.162083936324168e-05,
      "loss": 0.0089,
      "step": 4372
    },
    {
      "epoch": 6.32,
      "learning_rate": 8.158867985206625e-05,
      "loss": 0.006,
      "step": 4373
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.155652034089082e-05,
      "loss": 0.0598,
      "step": 4374
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.152436082971539e-05,
      "loss": 0.0733,
      "step": 4375
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.149220131853996e-05,
      "loss": 0.0135,
      "step": 4376
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.146004180736453e-05,
      "loss": 0.0493,
      "step": 4377
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.14278822961891e-05,
      "loss": 0.0075,
      "step": 4378
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.139572278501367e-05,
      "loss": 0.0164,
      "step": 4379
    },
    {
      "epoch": 6.33,
      "learning_rate": 8.136356327383824e-05,
      "loss": 0.0086,
      "step": 4380
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.133140376266281e-05,
      "loss": 0.0016,
      "step": 4381
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.129924425148738e-05,
      "loss": 0.0186,
      "step": 4382
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.126708474031196e-05,
      "loss": 0.01,
      "step": 4383
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.123492522913651e-05,
      "loss": 0.0791,
      "step": 4384
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.12027657179611e-05,
      "loss": 0.0284,
      "step": 4385
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.117060620678565e-05,
      "loss": 0.0194,
      "step": 4386
    },
    {
      "epoch": 6.34,
      "learning_rate": 8.113844669561022e-05,
      "loss": 0.0123,
      "step": 4387
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.110628718443481e-05,
      "loss": 0.0802,
      "step": 4388
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.107412767325936e-05,
      "loss": 0.0127,
      "step": 4389
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.104196816208394e-05,
      "loss": 0.0262,
      "step": 4390
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.100980865090852e-05,
      "loss": 0.0229,
      "step": 4391
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.097764913973308e-05,
      "loss": 0.0743,
      "step": 4392
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.094548962855765e-05,
      "loss": 0.0089,
      "step": 4393
    },
    {
      "epoch": 6.35,
      "learning_rate": 8.091333011738223e-05,
      "loss": 0.0002,
      "step": 4394
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.088117060620679e-05,
      "loss": 0.0024,
      "step": 4395
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.084901109503136e-05,
      "loss": 0.0009,
      "step": 4396
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.081685158385593e-05,
      "loss": 0.0021,
      "step": 4397
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.07846920726805e-05,
      "loss": 0.0074,
      "step": 4398
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.075253256150507e-05,
      "loss": 0.0005,
      "step": 4399
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.072037305032964e-05,
      "loss": 0.0131,
      "step": 4400
    },
    {
      "epoch": 6.36,
      "learning_rate": 8.068821353915421e-05,
      "loss": 0.1079,
      "step": 4401
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.065605402797878e-05,
      "loss": 0.0132,
      "step": 4402
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.062389451680334e-05,
      "loss": 0.0193,
      "step": 4403
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.059173500562792e-05,
      "loss": 0.0477,
      "step": 4404
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.055957549445249e-05,
      "loss": 0.0054,
      "step": 4405
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.052741598327705e-05,
      "loss": 0.0217,
      "step": 4406
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.049525647210163e-05,
      "loss": 0.0745,
      "step": 4407
    },
    {
      "epoch": 6.37,
      "learning_rate": 8.04630969609262e-05,
      "loss": 0.0816,
      "step": 4408
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.043093744975076e-05,
      "loss": 0.0045,
      "step": 4409
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.039877793857535e-05,
      "loss": 0.0106,
      "step": 4410
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.03666184273999e-05,
      "loss": 0.0001,
      "step": 4411
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.033445891622447e-05,
      "loss": 0.0349,
      "step": 4412
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.030229940504904e-05,
      "loss": 0.0467,
      "step": 4413
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.027013989387361e-05,
      "loss": 0.0001,
      "step": 4414
    },
    {
      "epoch": 6.38,
      "learning_rate": 8.023798038269818e-05,
      "loss": 0.0014,
      "step": 4415
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.020582087152276e-05,
      "loss": 0.0371,
      "step": 4416
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.017366136034733e-05,
      "loss": 0.0438,
      "step": 4417
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.01415018491719e-05,
      "loss": 0.0073,
      "step": 4418
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.010934233799647e-05,
      "loss": 0.0071,
      "step": 4419
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.007718282682104e-05,
      "loss": 0.1196,
      "step": 4420
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.004502331564561e-05,
      "loss": 0.0181,
      "step": 4421
    },
    {
      "epoch": 6.39,
      "learning_rate": 8.001286380447018e-05,
      "loss": 0.0035,
      "step": 4422
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.998070429329475e-05,
      "loss": 0.0035,
      "step": 4423
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.994854478211932e-05,
      "loss": 0.046,
      "step": 4424
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.991638527094389e-05,
      "loss": 0.0065,
      "step": 4425
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.988422575976846e-05,
      "loss": 0.0001,
      "step": 4426
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.985206624859303e-05,
      "loss": 0.0752,
      "step": 4427
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.981990673741759e-05,
      "loss": 0.0214,
      "step": 4428
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.978774722624216e-05,
      "loss": 0.0234,
      "step": 4429
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.975558771506674e-05,
      "loss": 0.0464,
      "step": 4430
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.97234282038913e-05,
      "loss": 0.0011,
      "step": 4431
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.969126869271587e-05,
      "loss": 0.1292,
      "step": 4432
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.965910918154045e-05,
      "loss": 0.0952,
      "step": 4433
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.962694967036501e-05,
      "loss": 0.0064,
      "step": 4434
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.959479015918958e-05,
      "loss": 0.0521,
      "step": 4435
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.956263064801415e-05,
      "loss": 0.0292,
      "step": 4436
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.953047113683872e-05,
      "loss": 0.0275,
      "step": 4437
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.949831162566329e-05,
      "loss": 0.0,
      "step": 4438
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.946615211448786e-05,
      "loss": 0.0027,
      "step": 4439
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.943399260331243e-05,
      "loss": 0.0023,
      "step": 4440
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.9401833092137e-05,
      "loss": 0.001,
      "step": 4441
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.936967358096156e-05,
      "loss": 0.0147,
      "step": 4442
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.933751406978615e-05,
      "loss": 0.0153,
      "step": 4443
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.930535455861072e-05,
      "loss": 0.0002,
      "step": 4444
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.927319504743527e-05,
      "loss": 0.0118,
      "step": 4445
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.924103553625986e-05,
      "loss": 0.0002,
      "step": 4446
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.920887602508443e-05,
      "loss": 0.0009,
      "step": 4447
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.917671651390898e-05,
      "loss": 0.0015,
      "step": 4448
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.914455700273357e-05,
      "loss": 0.0018,
      "step": 4449
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.911239749155813e-05,
      "loss": 0.0061,
      "step": 4450
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.90802379803827e-05,
      "loss": 0.0474,
      "step": 4451
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.904807846920728e-05,
      "loss": 0.0292,
      "step": 4452
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.901591895803184e-05,
      "loss": 0.0728,
      "step": 4453
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.898375944685641e-05,
      "loss": 0.0003,
      "step": 4454
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.895159993568099e-05,
      "loss": 0.0226,
      "step": 4455
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.891944042450555e-05,
      "loss": 0.0645,
      "step": 4456
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.888728091333012e-05,
      "loss": 0.0034,
      "step": 4457
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.885512140215469e-05,
      "loss": 0.0038,
      "step": 4458
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.882296189097926e-05,
      "loss": 0.1216,
      "step": 4459
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.879080237980383e-05,
      "loss": 0.0427,
      "step": 4460
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.87586428686284e-05,
      "loss": 0.0014,
      "step": 4461
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.872648335745297e-05,
      "loss": 0.0203,
      "step": 4462
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.869432384627754e-05,
      "loss": 0.0763,
      "step": 4463
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.866216433510211e-05,
      "loss": 0.0491,
      "step": 4464
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.863000482392668e-05,
      "loss": 0.0252,
      "step": 4465
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.859784531275125e-05,
      "loss": 0.0003,
      "step": 4466
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.856568580157581e-05,
      "loss": 0.0079,
      "step": 4467
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.85335262904004e-05,
      "loss": 0.167,
      "step": 4468
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.850136677922497e-05,
      "loss": 0.0,
      "step": 4469
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.846920726804952e-05,
      "loss": 0.1235,
      "step": 4470
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.84370477568741e-05,
      "loss": 0.0901,
      "step": 4471
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.840488824569868e-05,
      "loss": 0.0669,
      "step": 4472
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.837272873452323e-05,
      "loss": 0.0008,
      "step": 4473
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.83405692233478e-05,
      "loss": 0.0197,
      "step": 4474
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.830840971217237e-05,
      "loss": 0.0582,
      "step": 4475
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.827625020099695e-05,
      "loss": 0.0068,
      "step": 4476
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.824409068982152e-05,
      "loss": 0.0044,
      "step": 4477
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.821193117864609e-05,
      "loss": 0.0002,
      "step": 4478
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.817977166747066e-05,
      "loss": 0.0459,
      "step": 4479
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.814761215629523e-05,
      "loss": 0.0461,
      "step": 4480
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.81154526451198e-05,
      "loss": 0.0094,
      "step": 4481
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.808329313394437e-05,
      "loss": 0.0052,
      "step": 4482
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.805113362276894e-05,
      "loss": 0.0027,
      "step": 4483
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.801897411159351e-05,
      "loss": 0.0523,
      "step": 4484
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.798681460041808e-05,
      "loss": 0.0139,
      "step": 4485
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.795465508924265e-05,
      "loss": 0.1003,
      "step": 4486
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.792249557806721e-05,
      "loss": 0.0558,
      "step": 4487
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.789033606689179e-05,
      "loss": 0.0277,
      "step": 4488
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.785817655571636e-05,
      "loss": 0.0651,
      "step": 4489
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.782601704454092e-05,
      "loss": 0.045,
      "step": 4490
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.77938575333655e-05,
      "loss": 0.0139,
      "step": 4491
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.776169802219006e-05,
      "loss": 0.012,
      "step": 4492
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.772953851101463e-05,
      "loss": 0.0135,
      "step": 4493
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.769737899983921e-05,
      "loss": 0.0748,
      "step": 4494
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.766521948866377e-05,
      "loss": 0.0501,
      "step": 4495
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.763305997748834e-05,
      "loss": 0.1151,
      "step": 4496
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.760090046631293e-05,
      "loss": 0.0025,
      "step": 4497
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.756874095513748e-05,
      "loss": 0.0103,
      "step": 4498
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.753658144396205e-05,
      "loss": 0.0481,
      "step": 4499
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.750442193278662e-05,
      "loss": 0.0282,
      "step": 4500
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.74722624216112e-05,
      "loss": 0.0801,
      "step": 4501
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.744010291043577e-05,
      "loss": 0.0289,
      "step": 4502
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.740794339926034e-05,
      "loss": 0.019,
      "step": 4503
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.73757838880849e-05,
      "loss": 0.023,
      "step": 4504
    },
    {
      "epoch": 6.51,
      "learning_rate": 7.734362437690948e-05,
      "loss": 0.1052,
      "step": 4505
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.731146486573403e-05,
      "loss": 0.0067,
      "step": 4506
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.727930535455862e-05,
      "loss": 0.0665,
      "step": 4507
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.724714584338319e-05,
      "loss": 0.0075,
      "step": 4508
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.721498633220775e-05,
      "loss": 0.0453,
      "step": 4509
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.718282682103233e-05,
      "loss": 0.0812,
      "step": 4510
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.71506673098569e-05,
      "loss": 0.0689,
      "step": 4511
    },
    {
      "epoch": 6.52,
      "learning_rate": 7.711850779868146e-05,
      "loss": 0.0284,
      "step": 4512
    },
    {
      "epoch": 6.53,
      "learning_rate": 7.708634828750604e-05,
      "loss": 0.0954,
      "step": 4513
    },
    {
      "epoch": 6.53,
      "learning_rate": 7.705418877633061e-05,
      "loss": 0.0034,
      "step": 4514
    },
    {
      "epoch": 6.53,
      "learning_rate": 7.702202926515517e-05,
      "loss": 0.0424,
      "step": 4515
    },
    {
      "epoch": 6.53,
      "learning_rate": 7.698986975397975e-05,
      "loss": 0.0119,
      "step": 4516
    },
    {
      "epoch": 6.53,
      "learning_rate": 7.695771024280431e-05,
      "loss": 0.0236,
      "step": 4517
    },
    {
      "epoch": 6.53,
      "learning_rate": 7.692555073162888e-05,
      "loss": 0.0135,
      "step": 4518
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.689339122045345e-05,
      "loss": 0.0138,
      "step": 4519
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.686123170927802e-05,
      "loss": 0.064,
      "step": 4520
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.682907219810259e-05,
      "loss": 0.0139,
      "step": 4521
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.679691268692716e-05,
      "loss": 0.0964,
      "step": 4522
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.676475317575173e-05,
      "loss": 0.0704,
      "step": 4523
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.67325936645763e-05,
      "loss": 0.0048,
      "step": 4524
    },
    {
      "epoch": 6.54,
      "learning_rate": 7.670043415340087e-05,
      "loss": 0.0054,
      "step": 4525
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.666827464222544e-05,
      "loss": 0.0083,
      "step": 4526
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.663611513105001e-05,
      "loss": 0.0113,
      "step": 4527
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.660395561987458e-05,
      "loss": 0.0542,
      "step": 4528
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.657179610869916e-05,
      "loss": 0.0802,
      "step": 4529
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.653963659752373e-05,
      "loss": 0.0642,
      "step": 4530
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.650747708634828e-05,
      "loss": 0.0169,
      "step": 4531
    },
    {
      "epoch": 6.55,
      "learning_rate": 7.647531757517285e-05,
      "loss": 0.0106,
      "step": 4532
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.644315806399744e-05,
      "loss": 0.0151,
      "step": 4533
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.6410998552822e-05,
      "loss": 0.0406,
      "step": 4534
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.637883904164656e-05,
      "loss": 0.1053,
      "step": 4535
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.634667953047115e-05,
      "loss": 0.0271,
      "step": 4536
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.63145200192957e-05,
      "loss": 0.065,
      "step": 4537
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.628236050812028e-05,
      "loss": 0.0002,
      "step": 4538
    },
    {
      "epoch": 6.56,
      "learning_rate": 7.625020099694486e-05,
      "loss": 0.0608,
      "step": 4539
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.621804148576942e-05,
      "loss": 0.0161,
      "step": 4540
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.618588197459399e-05,
      "loss": 0.0107,
      "step": 4541
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.615372246341856e-05,
      "loss": 0.0091,
      "step": 4542
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.612156295224313e-05,
      "loss": 0.0464,
      "step": 4543
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.60894034410677e-05,
      "loss": 0.0938,
      "step": 4544
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.605724392989227e-05,
      "loss": 0.051,
      "step": 4545
    },
    {
      "epoch": 6.57,
      "learning_rate": 7.602508441871684e-05,
      "loss": 0.044,
      "step": 4546
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.599292490754141e-05,
      "loss": 0.0054,
      "step": 4547
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.596076539636597e-05,
      "loss": 0.051,
      "step": 4548
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.592860588519055e-05,
      "loss": 0.0141,
      "step": 4549
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.589644637401512e-05,
      "loss": 0.0857,
      "step": 4550
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.586428686283968e-05,
      "loss": 0.1123,
      "step": 4551
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.583212735166426e-05,
      "loss": 0.0249,
      "step": 4552
    },
    {
      "epoch": 6.58,
      "learning_rate": 7.579996784048883e-05,
      "loss": 0.0944,
      "step": 4553
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.576780832931339e-05,
      "loss": 0.0021,
      "step": 4554
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.573564881813798e-05,
      "loss": 0.0195,
      "step": 4555
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.570348930696253e-05,
      "loss": 0.0061,
      "step": 4556
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.56713297957871e-05,
      "loss": 0.0201,
      "step": 4557
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.563917028461169e-05,
      "loss": 0.0694,
      "step": 4558
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.560701077343624e-05,
      "loss": 0.0281,
      "step": 4559
    },
    {
      "epoch": 6.59,
      "learning_rate": 7.557485126226081e-05,
      "loss": 0.0273,
      "step": 4560
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.55426917510854e-05,
      "loss": 0.0342,
      "step": 4561
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.551053223990996e-05,
      "loss": 0.0615,
      "step": 4562
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.547837272873453e-05,
      "loss": 0.0567,
      "step": 4563
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.54462132175591e-05,
      "loss": 0.0157,
      "step": 4564
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.541405370638367e-05,
      "loss": 0.0563,
      "step": 4565
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.538189419520824e-05,
      "loss": 0.0012,
      "step": 4566
    },
    {
      "epoch": 6.6,
      "learning_rate": 7.534973468403281e-05,
      "loss": 0.0142,
      "step": 4567
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.531757517285738e-05,
      "loss": 0.0049,
      "step": 4568
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.528541566168195e-05,
      "loss": 0.0049,
      "step": 4569
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.525325615050652e-05,
      "loss": 0.0144,
      "step": 4570
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.522109663933109e-05,
      "loss": 0.0506,
      "step": 4571
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.518893712815566e-05,
      "loss": 0.0471,
      "step": 4572
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.515677761698022e-05,
      "loss": 0.0414,
      "step": 4573
    },
    {
      "epoch": 6.61,
      "learning_rate": 7.51246181058048e-05,
      "loss": 0.0284,
      "step": 4574
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.509245859462937e-05,
      "loss": 0.0346,
      "step": 4575
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.506029908345393e-05,
      "loss": 0.0914,
      "step": 4576
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.50281395722785e-05,
      "loss": 0.0075,
      "step": 4577
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.499598006110308e-05,
      "loss": 0.016,
      "step": 4578
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.496382054992764e-05,
      "loss": 0.0263,
      "step": 4579
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.493166103875221e-05,
      "loss": 0.0229,
      "step": 4580
    },
    {
      "epoch": 6.62,
      "learning_rate": 7.489950152757678e-05,
      "loss": 0.0088,
      "step": 4581
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.486734201640135e-05,
      "loss": 0.0308,
      "step": 4582
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.483518250522592e-05,
      "loss": 0.0169,
      "step": 4583
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.480302299405049e-05,
      "loss": 0.0702,
      "step": 4584
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.477086348287506e-05,
      "loss": 0.0084,
      "step": 4585
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.473870397169963e-05,
      "loss": 0.0199,
      "step": 4586
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.47065444605242e-05,
      "loss": 0.0009,
      "step": 4587
    },
    {
      "epoch": 6.63,
      "learning_rate": 7.467438494934877e-05,
      "loss": 0.0122,
      "step": 4588
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.464222543817335e-05,
      "loss": 0.0718,
      "step": 4589
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.461006592699792e-05,
      "loss": 0.0271,
      "step": 4590
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.457790641582249e-05,
      "loss": 0.0028,
      "step": 4591
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.454574690464706e-05,
      "loss": 0.0333,
      "step": 4592
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.451358739347161e-05,
      "loss": 0.036,
      "step": 4593
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.44814278822962e-05,
      "loss": 0.072,
      "step": 4594
    },
    {
      "epoch": 6.64,
      "learning_rate": 7.444926837112075e-05,
      "loss": 0.046,
      "step": 4595
    },
    {
      "epoch": 6.65,
      "learning_rate": 7.441710885994533e-05,
      "loss": 0.0512,
      "step": 4596
    },
    {
      "epoch": 6.65,
      "learning_rate": 7.438494934876991e-05,
      "loss": 0.0503,
      "step": 4597
    },
    {
      "epoch": 6.65,
      "learning_rate": 7.435278983759447e-05,
      "loss": 0.0047,
      "step": 4598
    },
    {
      "epoch": 6.65,
      "learning_rate": 7.432063032641904e-05,
      "loss": 0.0022,
      "step": 4599
    },
    {
      "epoch": 6.65,
      "learning_rate": 7.428847081524362e-05,
      "loss": 0.0164,
      "step": 4600
    },
    {
      "epoch": 6.65,
      "learning_rate": 7.425631130406818e-05,
      "loss": 0.0479,
      "step": 4601
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.422415179289275e-05,
      "loss": 0.0333,
      "step": 4602
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.419199228171733e-05,
      "loss": 0.1096,
      "step": 4603
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.415983277054189e-05,
      "loss": 0.0249,
      "step": 4604
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.412767325936646e-05,
      "loss": 0.0168,
      "step": 4605
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.409551374819103e-05,
      "loss": 0.0212,
      "step": 4606
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.40633542370156e-05,
      "loss": 0.1777,
      "step": 4607
    },
    {
      "epoch": 6.66,
      "learning_rate": 7.403119472584017e-05,
      "loss": 0.008,
      "step": 4608
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.399903521466474e-05,
      "loss": 0.013,
      "step": 4609
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.396687570348931e-05,
      "loss": 0.0514,
      "step": 4610
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.393471619231388e-05,
      "loss": 0.0133,
      "step": 4611
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.390255668113844e-05,
      "loss": 0.0227,
      "step": 4612
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.387039716996302e-05,
      "loss": 0.0102,
      "step": 4613
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.38382376587876e-05,
      "loss": 0.0007,
      "step": 4614
    },
    {
      "epoch": 6.67,
      "learning_rate": 7.380607814761215e-05,
      "loss": 0.0169,
      "step": 4615
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.377391863643674e-05,
      "loss": 0.0243,
      "step": 4616
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.37417591252613e-05,
      "loss": 0.0068,
      "step": 4617
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.370959961408586e-05,
      "loss": 0.0594,
      "step": 4618
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.367744010291045e-05,
      "loss": 0.0581,
      "step": 4619
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.3645280591735e-05,
      "loss": 0.0069,
      "step": 4620
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.361312108055957e-05,
      "loss": 0.0546,
      "step": 4621
    },
    {
      "epoch": 6.68,
      "learning_rate": 7.358096156938415e-05,
      "loss": 0.0106,
      "step": 4622
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.354880205820872e-05,
      "loss": 0.0549,
      "step": 4623
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.351664254703329e-05,
      "loss": 0.0047,
      "step": 4624
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.348448303585786e-05,
      "loss": 0.0148,
      "step": 4625
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.345232352468243e-05,
      "loss": 0.0362,
      "step": 4626
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.3420164013507e-05,
      "loss": 0.0633,
      "step": 4627
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.338800450233157e-05,
      "loss": 0.1009,
      "step": 4628
    },
    {
      "epoch": 6.69,
      "learning_rate": 7.335584499115614e-05,
      "loss": 0.0018,
      "step": 4629
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.332368547998071e-05,
      "loss": 0.0577,
      "step": 4630
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.329152596880528e-05,
      "loss": 0.0076,
      "step": 4631
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.325936645762985e-05,
      "loss": 0.0031,
      "step": 4632
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.322720694645442e-05,
      "loss": 0.0001,
      "step": 4633
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.319504743527899e-05,
      "loss": 0.003,
      "step": 4634
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.316288792410356e-05,
      "loss": 0.0002,
      "step": 4635
    },
    {
      "epoch": 6.7,
      "learning_rate": 7.313072841292813e-05,
      "loss": 0.0374,
      "step": 4636
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.309856890175269e-05,
      "loss": 0.0646,
      "step": 4637
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.306640939057726e-05,
      "loss": 0.0099,
      "step": 4638
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.303424987940184e-05,
      "loss": 0.0033,
      "step": 4639
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.30020903682264e-05,
      "loss": 0.0821,
      "step": 4640
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.296993085705097e-05,
      "loss": 0.0121,
      "step": 4641
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.293777134587556e-05,
      "loss": 0.0446,
      "step": 4642
    },
    {
      "epoch": 6.71,
      "learning_rate": 7.290561183470011e-05,
      "loss": 0.0641,
      "step": 4643
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.287345232352468e-05,
      "loss": 0.0153,
      "step": 4644
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.284129281234925e-05,
      "loss": 0.0239,
      "step": 4645
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.280913330117382e-05,
      "loss": 0.0003,
      "step": 4646
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.27769737899984e-05,
      "loss": 0.0094,
      "step": 4647
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.274481427882297e-05,
      "loss": 0.0095,
      "step": 4648
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.271265476764754e-05,
      "loss": 0.074,
      "step": 4649
    },
    {
      "epoch": 6.72,
      "learning_rate": 7.26804952564721e-05,
      "loss": 0.0049,
      "step": 4650
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.264833574529666e-05,
      "loss": 0.0024,
      "step": 4651
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.261617623412125e-05,
      "loss": 0.0071,
      "step": 4652
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.258401672294582e-05,
      "loss": 0.0026,
      "step": 4653
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.255185721177037e-05,
      "loss": 0.1013,
      "step": 4654
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.251969770059496e-05,
      "loss": 0.0128,
      "step": 4655
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.248753818941953e-05,
      "loss": 0.0542,
      "step": 4656
    },
    {
      "epoch": 6.73,
      "learning_rate": 7.245537867824409e-05,
      "loss": 0.1428,
      "step": 4657
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.242321916706867e-05,
      "loss": 0.0079,
      "step": 4658
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.239105965589324e-05,
      "loss": 0.0426,
      "step": 4659
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.23589001447178e-05,
      "loss": 0.0059,
      "step": 4660
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.232674063354238e-05,
      "loss": 0.0161,
      "step": 4661
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.229458112236694e-05,
      "loss": 0.0059,
      "step": 4662
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.226242161119151e-05,
      "loss": 0.0038,
      "step": 4663
    },
    {
      "epoch": 6.74,
      "learning_rate": 7.223026210001609e-05,
      "loss": 0.0119,
      "step": 4664
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.219810258884065e-05,
      "loss": 0.0133,
      "step": 4665
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.216594307766522e-05,
      "loss": 0.0786,
      "step": 4666
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.213378356648979e-05,
      "loss": 0.0001,
      "step": 4667
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.210162405531436e-05,
      "loss": 0.0483,
      "step": 4668
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.206946454413893e-05,
      "loss": 0.0496,
      "step": 4669
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.20373050329635e-05,
      "loss": 0.0044,
      "step": 4670
    },
    {
      "epoch": 6.75,
      "learning_rate": 7.200514552178807e-05,
      "loss": 0.0033,
      "step": 4671
    },
    {
      "epoch": 6.76,
      "learning_rate": 7.197298601061264e-05,
      "loss": 0.0078,
      "step": 4672
    },
    {
      "epoch": 6.76,
      "learning_rate": 7.194082649943721e-05,
      "loss": 0.075,
      "step": 4673
    },
    {
      "epoch": 6.76,
      "learning_rate": 7.190866698826178e-05,
      "loss": 0.0043,
      "step": 4674
    },
    {
      "epoch": 6.76,
      "learning_rate": 7.187650747708636e-05,
      "loss": 0.0036,
      "step": 4675
    },
    {
      "epoch": 6.76,
      "learning_rate": 7.184434796591091e-05,
      "loss": 0.0154,
      "step": 4676
    },
    {
      "epoch": 6.76,
      "learning_rate": 7.18121884547355e-05,
      "loss": 0.034,
      "step": 4677
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.178002894356007e-05,
      "loss": 0.0288,
      "step": 4678
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.174786943238462e-05,
      "loss": 0.0279,
      "step": 4679
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.171570992120921e-05,
      "loss": 0.0348,
      "step": 4680
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.168355041003378e-05,
      "loss": 0.0079,
      "step": 4681
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.165139089885834e-05,
      "loss": 0.0018,
      "step": 4682
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.16192313876829e-05,
      "loss": 0.0883,
      "step": 4683
    },
    {
      "epoch": 6.77,
      "learning_rate": 7.158707187650749e-05,
      "loss": 0.0184,
      "step": 4684
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.155491236533205e-05,
      "loss": 0.002,
      "step": 4685
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.152275285415662e-05,
      "loss": 0.0056,
      "step": 4686
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.149059334298119e-05,
      "loss": 0.0005,
      "step": 4687
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.145843383180576e-05,
      "loss": 0.0377,
      "step": 4688
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.142627432063033e-05,
      "loss": 0.0309,
      "step": 4689
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.13941148094549e-05,
      "loss": 0.0191,
      "step": 4690
    },
    {
      "epoch": 6.78,
      "learning_rate": 7.136195529827947e-05,
      "loss": 0.0563,
      "step": 4691
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.132979578710404e-05,
      "loss": 0.0153,
      "step": 4692
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.129763627592861e-05,
      "loss": 0.005,
      "step": 4693
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.126547676475318e-05,
      "loss": 0.0061,
      "step": 4694
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.123331725357775e-05,
      "loss": 0.0065,
      "step": 4695
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.120115774240231e-05,
      "loss": 0.0004,
      "step": 4696
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.116899823122689e-05,
      "loss": 0.0078,
      "step": 4697
    },
    {
      "epoch": 6.79,
      "learning_rate": 7.113683872005146e-05,
      "loss": 0.019,
      "step": 4698
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.110467920887602e-05,
      "loss": 0.0038,
      "step": 4699
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.10725196977006e-05,
      "loss": 0.0596,
      "step": 4700
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.104036018652516e-05,
      "loss": 0.0063,
      "step": 4701
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.100820067534973e-05,
      "loss": 0.0576,
      "step": 4702
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.097604116417432e-05,
      "loss": 0.0342,
      "step": 4703
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.094388165299887e-05,
      "loss": 0.0048,
      "step": 4704
    },
    {
      "epoch": 6.8,
      "learning_rate": 7.091172214182344e-05,
      "loss": 0.0444,
      "step": 4705
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.087956263064803e-05,
      "loss": 0.0374,
      "step": 4706
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.084740311947258e-05,
      "loss": 0.0515,
      "step": 4707
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.081524360829716e-05,
      "loss": 0.0452,
      "step": 4708
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.078308409712173e-05,
      "loss": 0.0325,
      "step": 4709
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.07509245859463e-05,
      "loss": 0.0079,
      "step": 4710
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.071876507477087e-05,
      "loss": 0.0735,
      "step": 4711
    },
    {
      "epoch": 6.81,
      "learning_rate": 7.068660556359544e-05,
      "loss": 0.125,
      "step": 4712
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.065444605242001e-05,
      "loss": 0.0323,
      "step": 4713
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.062228654124458e-05,
      "loss": 0.1091,
      "step": 4714
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.059012703006915e-05,
      "loss": 0.0024,
      "step": 4715
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.055796751889372e-05,
      "loss": 0.0011,
      "step": 4716
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.052580800771829e-05,
      "loss": 0.0271,
      "step": 4717
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.049364849654285e-05,
      "loss": 0.0011,
      "step": 4718
    },
    {
      "epoch": 6.82,
      "learning_rate": 7.046148898536743e-05,
      "loss": 0.0122,
      "step": 4719
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.0429329474192e-05,
      "loss": 0.0033,
      "step": 4720
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.039716996301656e-05,
      "loss": 0.0029,
      "step": 4721
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.036501045184114e-05,
      "loss": 0.1697,
      "step": 4722
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.033285094066571e-05,
      "loss": 0.0158,
      "step": 4723
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.030069142949027e-05,
      "loss": 0.0604,
      "step": 4724
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.026853191831485e-05,
      "loss": 0.1119,
      "step": 4725
    },
    {
      "epoch": 6.83,
      "learning_rate": 7.023637240713941e-05,
      "loss": 0.0291,
      "step": 4726
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.020421289596398e-05,
      "loss": 0.0082,
      "step": 4727
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.017205338478855e-05,
      "loss": 0.0634,
      "step": 4728
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.013989387361312e-05,
      "loss": 0.0797,
      "step": 4729
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.010773436243769e-05,
      "loss": 0.0036,
      "step": 4730
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.007557485126226e-05,
      "loss": 0.0095,
      "step": 4731
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.004341534008683e-05,
      "loss": 0.0071,
      "step": 4732
    },
    {
      "epoch": 6.84,
      "learning_rate": 7.00112558289114e-05,
      "loss": 0.0034,
      "step": 4733
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.997909631773597e-05,
      "loss": 0.017,
      "step": 4734
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.994693680656055e-05,
      "loss": 0.0061,
      "step": 4735
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.991477729538512e-05,
      "loss": 0.0029,
      "step": 4736
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.988261778420969e-05,
      "loss": 0.0264,
      "step": 4737
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.985045827303426e-05,
      "loss": 0.0623,
      "step": 4738
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.981829876185883e-05,
      "loss": 0.0218,
      "step": 4739
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.978613925068338e-05,
      "loss": 0.0098,
      "step": 4740
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.975397973950795e-05,
      "loss": 0.005,
      "step": 4741
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.972182022833254e-05,
      "loss": 0.0558,
      "step": 4742
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.96896607171571e-05,
      "loss": 0.0504,
      "step": 4743
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.965750120598167e-05,
      "loss": 0.0298,
      "step": 4744
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.962534169480625e-05,
      "loss": 0.0354,
      "step": 4745
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.959318218363081e-05,
      "loss": 0.032,
      "step": 4746
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.956102267245538e-05,
      "loss": 0.0058,
      "step": 4747
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.952886316127996e-05,
      "loss": 0.0229,
      "step": 4748
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.949670365010452e-05,
      "loss": 0.0113,
      "step": 4749
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.946454413892909e-05,
      "loss": 0.0503,
      "step": 4750
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.943238462775366e-05,
      "loss": 0.0222,
      "step": 4751
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.940022511657823e-05,
      "loss": 0.0112,
      "step": 4752
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.93680656054028e-05,
      "loss": 0.0093,
      "step": 4753
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.933590609422737e-05,
      "loss": 0.0102,
      "step": 4754
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.930374658305194e-05,
      "loss": 0.035,
      "step": 4755
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.927158707187651e-05,
      "loss": 0.0366,
      "step": 4756
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.923942756070107e-05,
      "loss": 0.017,
      "step": 4757
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.920726804952565e-05,
      "loss": 0.0933,
      "step": 4758
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.917510853835022e-05,
      "loss": 0.0312,
      "step": 4759
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.914294902717478e-05,
      "loss": 0.0005,
      "step": 4760
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.911078951599937e-05,
      "loss": 0.0287,
      "step": 4761
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.907863000482394e-05,
      "loss": 0.0367,
      "step": 4762
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.904647049364849e-05,
      "loss": 0.0767,
      "step": 4763
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.901431098247308e-05,
      "loss": 0.0061,
      "step": 4764
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.898215147129763e-05,
      "loss": 0.0138,
      "step": 4765
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.89499919601222e-05,
      "loss": 0.0972,
      "step": 4766
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.891783244894679e-05,
      "loss": 0.0329,
      "step": 4767
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.888567293777135e-05,
      "loss": 0.0266,
      "step": 4768
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.885351342659592e-05,
      "loss": 0.0265,
      "step": 4769
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.882135391542049e-05,
      "loss": 0.0977,
      "step": 4770
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.878919440424506e-05,
      "loss": 0.0043,
      "step": 4771
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.875703489306963e-05,
      "loss": 0.0268,
      "step": 4772
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.87248753818942e-05,
      "loss": 0.027,
      "step": 4773
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.869271587071877e-05,
      "loss": 0.0079,
      "step": 4774
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.866055635954334e-05,
      "loss": 0.1165,
      "step": 4775
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.862839684836791e-05,
      "loss": 0.0034,
      "step": 4776
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.859623733719248e-05,
      "loss": 0.0684,
      "step": 4777
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.856407782601705e-05,
      "loss": 0.0369,
      "step": 4778
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.853191831484162e-05,
      "loss": 0.1068,
      "step": 4779
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.849975880366619e-05,
      "loss": 0.0857,
      "step": 4780
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.846759929249076e-05,
      "loss": 0.0077,
      "step": 4781
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.843543978131532e-05,
      "loss": 0.0448,
      "step": 4782
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.84032802701399e-05,
      "loss": 0.0044,
      "step": 4783
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.837112075896447e-05,
      "loss": 0.1012,
      "step": 4784
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.833896124778903e-05,
      "loss": 0.0835,
      "step": 4785
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.83068017366136e-05,
      "loss": 0.0253,
      "step": 4786
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.827464222543818e-05,
      "loss": 0.003,
      "step": 4787
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.824248271426274e-05,
      "loss": 0.125,
      "step": 4788
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.821032320308731e-05,
      "loss": 0.0409,
      "step": 4789
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.817816369191188e-05,
      "loss": 0.0415,
      "step": 4790
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.814600418073645e-05,
      "loss": 0.0244,
      "step": 4791
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.811384466956102e-05,
      "loss": 0.0039,
      "step": 4792
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.80816851583856e-05,
      "loss": 0.0004,
      "step": 4793
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.804952564721016e-05,
      "loss": 0.0092,
      "step": 4794
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.801736613603474e-05,
      "loss": 0.0684,
      "step": 4795
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.79852066248593e-05,
      "loss": 0.0019,
      "step": 4796
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.795304711368388e-05,
      "loss": 0.005,
      "step": 4797
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.792088760250845e-05,
      "loss": 0.0001,
      "step": 4798
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.788872809133302e-05,
      "loss": 0.1151,
      "step": 4799
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.785656858015759e-05,
      "loss": 0.0639,
      "step": 4800
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.782440906898216e-05,
      "loss": 0.0111,
      "step": 4801
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.779224955780672e-05,
      "loss": 0.0795,
      "step": 4802
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.77600900466313e-05,
      "loss": 0.0624,
      "step": 4803
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.772793053545587e-05,
      "loss": 0.0641,
      "step": 4804
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.769577102428043e-05,
      "loss": 0.0359,
      "step": 4805
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.766361151310501e-05,
      "loss": 0.0095,
      "step": 4806
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.763145200192957e-05,
      "loss": 0.0332,
      "step": 4807
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.759929249075414e-05,
      "loss": 0.0401,
      "step": 4808
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.756713297957872e-05,
      "loss": 0.0066,
      "step": 4809
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.753497346840328e-05,
      "loss": 0.0016,
      "step": 4810
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.750281395722785e-05,
      "loss": 0.0132,
      "step": 4811
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.747065444605243e-05,
      "loss": 0.0564,
      "step": 4812
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.743849493487699e-05,
      "loss": 0.0042,
      "step": 4813
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.740633542370156e-05,
      "loss": 0.0006,
      "step": 4814
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.737417591252613e-05,
      "loss": 0.0014,
      "step": 4815
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.73420164013507e-05,
      "loss": 0.0212,
      "step": 4816
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.730985689017527e-05,
      "loss": 0.0362,
      "step": 4817
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.727769737899984e-05,
      "loss": 0.0641,
      "step": 4818
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.724553786782441e-05,
      "loss": 0.0686,
      "step": 4819
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.721337835664898e-05,
      "loss": 0.0009,
      "step": 4820
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.718121884547354e-05,
      "loss": 0.0597,
      "step": 4821
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.714905933429813e-05,
      "loss": 0.0004,
      "step": 4822
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.71168998231227e-05,
      "loss": 0.103,
      "step": 4823
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.708474031194725e-05,
      "loss": 0.0001,
      "step": 4824
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.705258080077184e-05,
      "loss": 0.0291,
      "step": 4825
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.702042128959641e-05,
      "loss": 0.0735,
      "step": 4826
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.698826177842096e-05,
      "loss": 0.0007,
      "step": 4827
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.695610226724555e-05,
      "loss": 0.0099,
      "step": 4828
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.692394275607012e-05,
      "loss": 0.0062,
      "step": 4829
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.689178324489468e-05,
      "loss": 0.0456,
      "step": 4830
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.685962373371925e-05,
      "loss": 0.0007,
      "step": 4831
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.682746422254382e-05,
      "loss": 0.0564,
      "step": 4832
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.679530471136839e-05,
      "loss": 0.0333,
      "step": 4833
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.676314520019296e-05,
      "loss": 0.0177,
      "step": 4834
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.673098568901753e-05,
      "loss": 0.0151,
      "step": 4835
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.66988261778421e-05,
      "loss": 0.0354,
      "step": 4836
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.666666666666667e-05,
      "loss": 0.0631,
      "step": 4837
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.663450715549124e-05,
      "loss": 0.0302,
      "step": 4838
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.660234764431581e-05,
      "loss": 0.0148,
      "step": 4839
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.657018813314038e-05,
      "loss": 0.125,
      "step": 4840
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.653802862196495e-05,
      "loss": 0.0004,
      "step": 4841
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.650586911078952e-05,
      "loss": 0.0049,
      "step": 4842
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.647370959961409e-05,
      "loss": 0.0255,
      "step": 4843
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.644155008843866e-05,
      "loss": 0.001,
      "step": 4844
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.640939057726323e-05,
      "loss": 0.0016,
      "step": 4845
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.637723106608779e-05,
      "loss": 0.0197,
      "step": 4846
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.634507155491236e-05,
      "loss": 0.0001,
      "step": 4847
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.631291204373695e-05,
      "loss": 0.0503,
      "step": 4848
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.62807525325615e-05,
      "loss": 0.023,
      "step": 4849
    },
    {
      "epoch": 7.01,
      "learning_rate": 6.624859302138607e-05,
      "loss": 0.0252,
      "step": 4850
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.621643351021066e-05,
      "loss": 0.0195,
      "step": 4851
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.618427399903521e-05,
      "loss": 0.0074,
      "step": 4852
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.615211448785978e-05,
      "loss": 0.006,
      "step": 4853
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.611995497668436e-05,
      "loss": 0.0055,
      "step": 4854
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.608779546550893e-05,
      "loss": 0.014,
      "step": 4855
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.60556359543335e-05,
      "loss": 0.0487,
      "step": 4856
    },
    {
      "epoch": 7.02,
      "learning_rate": 6.602347644315807e-05,
      "loss": 0.0562,
      "step": 4857
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.599131693198264e-05,
      "loss": 0.0113,
      "step": 4858
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.595915742080721e-05,
      "loss": 0.0052,
      "step": 4859
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.592699790963178e-05,
      "loss": 0.0156,
      "step": 4860
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.589483839845635e-05,
      "loss": 0.0703,
      "step": 4861
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.586267888728092e-05,
      "loss": 0.0393,
      "step": 4862
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.583051937610548e-05,
      "loss": 0.0021,
      "step": 4863
    },
    {
      "epoch": 7.03,
      "learning_rate": 6.579835986493006e-05,
      "loss": 0.0977,
      "step": 4864
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.576620035375463e-05,
      "loss": 0.0354,
      "step": 4865
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.573404084257919e-05,
      "loss": 0.0003,
      "step": 4866
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.570188133140377e-05,
      "loss": 0.001,
      "step": 4867
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.566972182022834e-05,
      "loss": 0.0194,
      "step": 4868
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.56375623090529e-05,
      "loss": 0.0349,
      "step": 4869
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.560540279787748e-05,
      "loss": 0.0123,
      "step": 4870
    },
    {
      "epoch": 7.04,
      "learning_rate": 6.557324328670204e-05,
      "loss": 0.0374,
      "step": 4871
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.554108377552661e-05,
      "loss": 0.0436,
      "step": 4872
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.55089242643512e-05,
      "loss": 0.0396,
      "step": 4873
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.547676475317575e-05,
      "loss": 0.0008,
      "step": 4874
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.544460524200032e-05,
      "loss": 0.0161,
      "step": 4875
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.541244573082489e-05,
      "loss": 0.0422,
      "step": 4876
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.538028621964946e-05,
      "loss": 0.0041,
      "step": 4877
    },
    {
      "epoch": 7.05,
      "learning_rate": 6.534812670847403e-05,
      "loss": 0.0036,
      "step": 4878
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.53159671972986e-05,
      "loss": 0.0136,
      "step": 4879
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.528380768612317e-05,
      "loss": 0.021,
      "step": 4880
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.525164817494775e-05,
      "loss": 0.023,
      "step": 4881
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.521948866377232e-05,
      "loss": 0.0005,
      "step": 4882
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.518732915259689e-05,
      "loss": 0.0211,
      "step": 4883
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.515516964142146e-05,
      "loss": 0.0305,
      "step": 4884
    },
    {
      "epoch": 7.06,
      "learning_rate": 6.512301013024601e-05,
      "loss": 0.0039,
      "step": 4885
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.50908506190706e-05,
      "loss": 0.005,
      "step": 4886
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.505869110789517e-05,
      "loss": 0.0004,
      "step": 4887
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.502653159671973e-05,
      "loss": 0.0068,
      "step": 4888
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.499437208554431e-05,
      "loss": 0.0873,
      "step": 4889
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.496221257436888e-05,
      "loss": 0.001,
      "step": 4890
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.493005306319344e-05,
      "loss": 0.0006,
      "step": 4891
    },
    {
      "epoch": 7.07,
      "learning_rate": 6.489789355201801e-05,
      "loss": 0.0076,
      "step": 4892
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.486573404084259e-05,
      "loss": 0.0376,
      "step": 4893
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.483357452966715e-05,
      "loss": 0.0482,
      "step": 4894
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.480141501849172e-05,
      "loss": 0.0325,
      "step": 4895
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.476925550731629e-05,
      "loss": 0.0486,
      "step": 4896
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.473709599614086e-05,
      "loss": 0.0172,
      "step": 4897
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.470493648496543e-05,
      "loss": 0.0347,
      "step": 4898
    },
    {
      "epoch": 7.08,
      "learning_rate": 6.467277697379e-05,
      "loss": 0.0104,
      "step": 4899
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.464061746261457e-05,
      "loss": 0.0011,
      "step": 4900
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.460845795143914e-05,
      "loss": 0.0069,
      "step": 4901
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.457629844026371e-05,
      "loss": 0.0036,
      "step": 4902
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.454413892908828e-05,
      "loss": 0.0139,
      "step": 4903
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.451197941791285e-05,
      "loss": 0.0329,
      "step": 4904
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.447981990673741e-05,
      "loss": 0.0693,
      "step": 4905
    },
    {
      "epoch": 7.09,
      "learning_rate": 6.4447660395562e-05,
      "loss": 0.0472,
      "step": 4906
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.441550088438657e-05,
      "loss": 0.0475,
      "step": 4907
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.438334137321112e-05,
      "loss": 0.0138,
      "step": 4908
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.43511818620357e-05,
      "loss": 0.0328,
      "step": 4909
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.431902235086026e-05,
      "loss": 0.0443,
      "step": 4910
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.428686283968483e-05,
      "loss": 0.0206,
      "step": 4911
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.425470332850942e-05,
      "loss": 0.0093,
      "step": 4912
    },
    {
      "epoch": 7.1,
      "learning_rate": 6.422254381733397e-05,
      "loss": 0.0256,
      "step": 4913
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.419038430615855e-05,
      "loss": 0.0013,
      "step": 4914
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.415822479498313e-05,
      "loss": 0.0504,
      "step": 4915
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.412606528380769e-05,
      "loss": 0.0515,
      "step": 4916
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.409390577263226e-05,
      "loss": 0.0023,
      "step": 4917
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.406174626145684e-05,
      "loss": 0.0062,
      "step": 4918
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.40295867502814e-05,
      "loss": 0.0272,
      "step": 4919
    },
    {
      "epoch": 7.11,
      "learning_rate": 6.399742723910597e-05,
      "loss": 0.053,
      "step": 4920
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.396526772793054e-05,
      "loss": 0.0023,
      "step": 4921
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.393310821675511e-05,
      "loss": 0.0177,
      "step": 4922
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.390094870557968e-05,
      "loss": 0.0174,
      "step": 4923
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.386878919440425e-05,
      "loss": 0.005,
      "step": 4924
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.383662968322882e-05,
      "loss": 0.0213,
      "step": 4925
    },
    {
      "epoch": 7.12,
      "learning_rate": 6.380447017205339e-05,
      "loss": 0.0562,
      "step": 4926
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.377231066087795e-05,
      "loss": 0.0031,
      "step": 4927
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.374015114970253e-05,
      "loss": 0.0363,
      "step": 4928
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.37079916385271e-05,
      "loss": 0.0084,
      "step": 4929
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.367583212735166e-05,
      "loss": 0.0046,
      "step": 4930
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.364367261617624e-05,
      "loss": 0.0043,
      "step": 4931
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.361151310500081e-05,
      "loss": 0.0211,
      "step": 4932
    },
    {
      "epoch": 7.13,
      "learning_rate": 6.357935359382537e-05,
      "loss": 0.0372,
      "step": 4933
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.354719408264994e-05,
      "loss": 0.0016,
      "step": 4934
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.351503457147451e-05,
      "loss": 0.0306,
      "step": 4935
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.348287506029908e-05,
      "loss": 0.0471,
      "step": 4936
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.345071554912365e-05,
      "loss": 0.0524,
      "step": 4937
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.341855603794822e-05,
      "loss": 0.0095,
      "step": 4938
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.33863965267728e-05,
      "loss": 0.0469,
      "step": 4939
    },
    {
      "epoch": 7.14,
      "learning_rate": 6.335423701559736e-05,
      "loss": 0.0757,
      "step": 4940
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.332207750442194e-05,
      "loss": 0.0003,
      "step": 4941
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.32899179932465e-05,
      "loss": 0.0019,
      "step": 4942
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.325775848207108e-05,
      "loss": 0.0112,
      "step": 4943
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.322559897089565e-05,
      "loss": 0.0696,
      "step": 4944
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.319343945972022e-05,
      "loss": 0.0232,
      "step": 4945
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.316127994854479e-05,
      "loss": 0.0001,
      "step": 4946
    },
    {
      "epoch": 7.15,
      "learning_rate": 6.312912043736936e-05,
      "loss": 0.0374,
      "step": 4947
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.309696092619393e-05,
      "loss": 0.0505,
      "step": 4948
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.30648014150185e-05,
      "loss": 0.05,
      "step": 4949
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.303264190384306e-05,
      "loss": 0.0077,
      "step": 4950
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.300048239266764e-05,
      "loss": 0.0001,
      "step": 4951
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.29683228814922e-05,
      "loss": 0.0598,
      "step": 4952
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.293616337031677e-05,
      "loss": 0.0384,
      "step": 4953
    },
    {
      "epoch": 7.16,
      "learning_rate": 6.290400385914135e-05,
      "loss": 0.0054,
      "step": 4954
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.287184434796591e-05,
      "loss": 0.0559,
      "step": 4955
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.283968483679048e-05,
      "loss": 0.0011,
      "step": 4956
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.280752532561506e-05,
      "loss": 0.0077,
      "step": 4957
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.277536581443962e-05,
      "loss": 0.0146,
      "step": 4958
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.274320630326419e-05,
      "loss": 0.0006,
      "step": 4959
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.271104679208876e-05,
      "loss": 0.002,
      "step": 4960
    },
    {
      "epoch": 7.17,
      "learning_rate": 6.267888728091333e-05,
      "loss": 0.002,
      "step": 4961
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.26467277697379e-05,
      "loss": 0.0001,
      "step": 4962
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.261456825856247e-05,
      "loss": 0.0158,
      "step": 4963
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.258240874738704e-05,
      "loss": 0.0335,
      "step": 4964
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.255024923621161e-05,
      "loss": 0.0591,
      "step": 4965
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.251808972503617e-05,
      "loss": 0.0396,
      "step": 4966
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.248593021386076e-05,
      "loss": 0.0,
      "step": 4967
    },
    {
      "epoch": 7.18,
      "learning_rate": 6.245377070268533e-05,
      "loss": 0.0027,
      "step": 4968
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.242161119150988e-05,
      "loss": 0.0495,
      "step": 4969
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.238945168033447e-05,
      "loss": 0.0225,
      "step": 4970
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.235729216915904e-05,
      "loss": 0.0231,
      "step": 4971
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.23251326579836e-05,
      "loss": 0.0335,
      "step": 4972
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.229297314680818e-05,
      "loss": 0.0008,
      "step": 4973
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.226081363563275e-05,
      "loss": 0.001,
      "step": 4974
    },
    {
      "epoch": 7.19,
      "learning_rate": 6.22286541244573e-05,
      "loss": 0.0187,
      "step": 4975
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.219649461328189e-05,
      "loss": 0.0694,
      "step": 4976
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.216433510210645e-05,
      "loss": 0.0698,
      "step": 4977
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.213217559093102e-05,
      "loss": 0.0129,
      "step": 4978
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.210001607975559e-05,
      "loss": 0.0004,
      "step": 4979
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.206785656858016e-05,
      "loss": 0.0008,
      "step": 4980
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.203569705740473e-05,
      "loss": 0.0153,
      "step": 4981
    },
    {
      "epoch": 7.2,
      "learning_rate": 6.20035375462293e-05,
      "loss": 0.0008,
      "step": 4982
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.197137803505387e-05,
      "loss": 0.0004,
      "step": 4983
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.193921852387844e-05,
      "loss": 0.0885,
      "step": 4984
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.190705901270301e-05,
      "loss": 0.0883,
      "step": 4985
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.187489950152758e-05,
      "loss": 0.0113,
      "step": 4986
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.184273999035215e-05,
      "loss": 0.0087,
      "step": 4987
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.181058047917672e-05,
      "loss": 0.0035,
      "step": 4988
    },
    {
      "epoch": 7.21,
      "learning_rate": 6.177842096800129e-05,
      "loss": 0.0002,
      "step": 4989
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.174626145682586e-05,
      "loss": 0.0427,
      "step": 4990
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.171410194565042e-05,
      "loss": 0.0047,
      "step": 4991
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.1681942434475e-05,
      "loss": 0.002,
      "step": 4992
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.164978292329957e-05,
      "loss": 0.0043,
      "step": 4993
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.161762341212413e-05,
      "loss": 0.0341,
      "step": 4994
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.15854639009487e-05,
      "loss": 0.043,
      "step": 4995
    },
    {
      "epoch": 7.22,
      "learning_rate": 6.155330438977329e-05,
      "loss": 0.0013,
      "step": 4996
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.152114487859784e-05,
      "loss": 0.0464,
      "step": 4997
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.148898536742241e-05,
      "loss": 0.0851,
      "step": 4998
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.145682585624698e-05,
      "loss": 0.0441,
      "step": 4999
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.142466634507155e-05,
      "loss": 0.0108,
      "step": 5000
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.139250683389613e-05,
      "loss": 0.0053,
      "step": 5001
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.13603473227207e-05,
      "loss": 0.0043,
      "step": 5002
    },
    {
      "epoch": 7.23,
      "learning_rate": 6.132818781154527e-05,
      "loss": 0.0012,
      "step": 5003
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.129602830036984e-05,
      "loss": 0.0023,
      "step": 5004
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.126386878919441e-05,
      "loss": 0.0002,
      "step": 5005
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.123170927801898e-05,
      "loss": 0.0221,
      "step": 5006
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.119954976684355e-05,
      "loss": 0.0629,
      "step": 5007
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.116739025566812e-05,
      "loss": 0.0218,
      "step": 5008
    },
    {
      "epoch": 7.24,
      "learning_rate": 6.113523074449269e-05,
      "loss": 0.0085,
      "step": 5009
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.110307123331726e-05,
      "loss": 0.0664,
      "step": 5010
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.107091172214182e-05,
      "loss": 0.0899,
      "step": 5011
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.10387522109664e-05,
      "loss": 0.0346,
      "step": 5012
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.1006592699790965e-05,
      "loss": 0.0105,
      "step": 5013
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.097443318861553e-05,
      "loss": 0.0002,
      "step": 5014
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.094227367744011e-05,
      "loss": 0.0045,
      "step": 5015
    },
    {
      "epoch": 7.25,
      "learning_rate": 6.0910114166264676e-05,
      "loss": 0.0307,
      "step": 5016
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.087795465508924e-05,
      "loss": 0.0901,
      "step": 5017
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.084579514391382e-05,
      "loss": 0.0015,
      "step": 5018
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.081363563273839e-05,
      "loss": 0.0001,
      "step": 5019
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.078147612156295e-05,
      "loss": 0.056,
      "step": 5020
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.074931661038753e-05,
      "loss": 0.0523,
      "step": 5021
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.07171570992121e-05,
      "loss": 0.0005,
      "step": 5022
    },
    {
      "epoch": 7.26,
      "learning_rate": 6.068499758803666e-05,
      "loss": 0.0001,
      "step": 5023
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.065283807686123e-05,
      "loss": 0.0155,
      "step": 5024
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.0620678565685804e-05,
      "loss": 0.0067,
      "step": 5025
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.0588519054510375e-05,
      "loss": 0.0214,
      "step": 5026
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.055635954333494e-05,
      "loss": 0.049,
      "step": 5027
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.0524200032159516e-05,
      "loss": 0.0041,
      "step": 5028
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.0492040520984086e-05,
      "loss": 0.0106,
      "step": 5029
    },
    {
      "epoch": 7.27,
      "learning_rate": 6.045988100980865e-05,
      "loss": 0.0501,
      "step": 5030
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.042772149863323e-05,
      "loss": 0.0502,
      "step": 5031
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.039556198745779e-05,
      "loss": 0.1036,
      "step": 5032
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.036340247628236e-05,
      "loss": 0.0496,
      "step": 5033
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.033124296510694e-05,
      "loss": 0.04,
      "step": 5034
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.02990834539315e-05,
      "loss": 0.0054,
      "step": 5035
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.026692394275607e-05,
      "loss": 0.0393,
      "step": 5036
    },
    {
      "epoch": 7.28,
      "learning_rate": 6.023476443158065e-05,
      "loss": 0.0001,
      "step": 5037
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.0202604920405214e-05,
      "loss": 0.0708,
      "step": 5038
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.017044540922978e-05,
      "loss": 0.0583,
      "step": 5039
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.013828589805435e-05,
      "loss": 0.0008,
      "step": 5040
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.0106126386878926e-05,
      "loss": 0.0571,
      "step": 5041
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.007396687570349e-05,
      "loss": 0.0024,
      "step": 5042
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.004180736452806e-05,
      "loss": 0.0018,
      "step": 5043
    },
    {
      "epoch": 7.29,
      "learning_rate": 6.000964785335264e-05,
      "loss": 0.0032,
      "step": 5044
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.99774883421772e-05,
      "loss": 0.0336,
      "step": 5045
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.9945328831001765e-05,
      "loss": 0.0312,
      "step": 5046
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.991316931982635e-05,
      "loss": 0.0001,
      "step": 5047
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.988100980865091e-05,
      "loss": 0.0901,
      "step": 5048
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.9848850297475476e-05,
      "loss": 0.0234,
      "step": 5049
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.9816690786300053e-05,
      "loss": 0.0376,
      "step": 5050
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.9784531275124624e-05,
      "loss": 0.0004,
      "step": 5051
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.975237176394919e-05,
      "loss": 0.0042,
      "step": 5052
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.9720212252773765e-05,
      "loss": 0.013,
      "step": 5053
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.9688052741598335e-05,
      "loss": 0.013,
      "step": 5054
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.96558932304229e-05,
      "loss": 0.0037,
      "step": 5055
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.962373371924746e-05,
      "loss": 0.061,
      "step": 5056
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.959157420807204e-05,
      "loss": 0.0009,
      "step": 5057
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.955941469689661e-05,
      "loss": 0.025,
      "step": 5058
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.9527255185721174e-05,
      "loss": 0.0071,
      "step": 5059
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.949509567454575e-05,
      "loss": 0.0077,
      "step": 5060
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.946293616337032e-05,
      "loss": 0.049,
      "step": 5061
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.9430776652194886e-05,
      "loss": 0.0,
      "step": 5062
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.939861714101946e-05,
      "loss": 0.0679,
      "step": 5063
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.936645762984403e-05,
      "loss": 0.041,
      "step": 5064
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.93342981186686e-05,
      "loss": 0.0698,
      "step": 5065
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.9302138607493175e-05,
      "loss": 0.0048,
      "step": 5066
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.926997909631774e-05,
      "loss": 0.0075,
      "step": 5067
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.923781958514231e-05,
      "loss": 0.0639,
      "step": 5068
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.920566007396687e-05,
      "loss": 0.0012,
      "step": 5069
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.917350056279145e-05,
      "loss": 0.0568,
      "step": 5070
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.9141341051616014e-05,
      "loss": 0.0244,
      "step": 5071
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.9109181540440584e-05,
      "loss": 0.0137,
      "step": 5072
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.907702202926516e-05,
      "loss": 0.0024,
      "step": 5073
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.9044862518089725e-05,
      "loss": 0.0108,
      "step": 5074
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.9012703006914296e-05,
      "loss": 0.0051,
      "step": 5075
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.898054349573887e-05,
      "loss": 0.0466,
      "step": 5076
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.894838398456344e-05,
      "loss": 0.0105,
      "step": 5077
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.891622447338801e-05,
      "loss": 0.0004,
      "step": 5078
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.8884064962212585e-05,
      "loss": 0.0261,
      "step": 5079
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.885190545103715e-05,
      "loss": 0.0029,
      "step": 5080
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.881974593986171e-05,
      "loss": 0.014,
      "step": 5081
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.878758642868629e-05,
      "loss": 0.0248,
      "step": 5082
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.875542691751086e-05,
      "loss": 0.0009,
      "step": 5083
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.8723267406335424e-05,
      "loss": 0.0862,
      "step": 5084
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.8691107895159994e-05,
      "loss": 0.0762,
      "step": 5085
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.865894838398457e-05,
      "loss": 0.0049,
      "step": 5086
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.8626788872809135e-05,
      "loss": 0.0085,
      "step": 5087
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.85946293616337e-05,
      "loss": 0.0668,
      "step": 5088
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.8562469850458276e-05,
      "loss": 0.0,
      "step": 5089
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.853031033928285e-05,
      "loss": 0.0582,
      "step": 5090
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.849815082810741e-05,
      "loss": 0.0007,
      "step": 5091
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.846599131693199e-05,
      "loss": 0.005,
      "step": 5092
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.843383180575656e-05,
      "loss": 0.0002,
      "step": 5093
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.840167229458112e-05,
      "loss": 0.0543,
      "step": 5094
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.83695127834057e-05,
      "loss": 0.0405,
      "step": 5095
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.833735327223026e-05,
      "loss": 0.0019,
      "step": 5096
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.8305193761054834e-05,
      "loss": 0.006,
      "step": 5097
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.82730342498794e-05,
      "loss": 0.0143,
      "step": 5098
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.8240874738703975e-05,
      "loss": 0.0117,
      "step": 5099
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.8208715227528545e-05,
      "loss": 0.0486,
      "step": 5100
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.817655571635311e-05,
      "loss": 0.0049,
      "step": 5101
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.8144396205177686e-05,
      "loss": 0.001,
      "step": 5102
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.811223669400226e-05,
      "loss": 0.0012,
      "step": 5103
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.808007718282682e-05,
      "loss": 0.0586,
      "step": 5104
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.80479176716514e-05,
      "loss": 0.0435,
      "step": 5105
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.801575816047596e-05,
      "loss": 0.1699,
      "step": 5106
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.798359864930053e-05,
      "loss": 0.0004,
      "step": 5107
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.795143913812511e-05,
      "loss": 0.0168,
      "step": 5108
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.791927962694967e-05,
      "loss": 0.0012,
      "step": 5109
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.7887120115774244e-05,
      "loss": 0.0009,
      "step": 5110
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.785496060459882e-05,
      "loss": 0.0701,
      "step": 5111
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.7822801093423385e-05,
      "loss": 0.0113,
      "step": 5112
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.779064158224795e-05,
      "loss": 0.032,
      "step": 5113
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.775848207107252e-05,
      "loss": 0.032,
      "step": 5114
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.7726322559897096e-05,
      "loss": 0.0001,
      "step": 5115
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.769416304872166e-05,
      "loss": 0.0162,
      "step": 5116
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.766200353754623e-05,
      "loss": 0.0221,
      "step": 5117
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.762984402637081e-05,
      "loss": 0.0033,
      "step": 5118
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.759768451519537e-05,
      "loss": 0.0078,
      "step": 5119
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.7565525004019935e-05,
      "loss": 0.0109,
      "step": 5120
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.753336549284451e-05,
      "loss": 0.0004,
      "step": 5121
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.750120598166908e-05,
      "loss": 0.0224,
      "step": 5122
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.746904647049365e-05,
      "loss": 0.0066,
      "step": 5123
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.7436886959318224e-05,
      "loss": 0.0533,
      "step": 5124
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.7404727448142794e-05,
      "loss": 0.1218,
      "step": 5125
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.737256793696736e-05,
      "loss": 0.1104,
      "step": 5126
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.7340408425791935e-05,
      "loss": 0.0076,
      "step": 5127
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.73082489146165e-05,
      "loss": 0.0071,
      "step": 5128
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.727608940344107e-05,
      "loss": 0.0003,
      "step": 5129
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.7243929892265633e-05,
      "loss": 0.0155,
      "step": 5130
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.721177038109021e-05,
      "loss": 0.0826,
      "step": 5131
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.717961086991478e-05,
      "loss": 0.0013,
      "step": 5132
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.7147451358739345e-05,
      "loss": 0.0357,
      "step": 5133
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.711529184756392e-05,
      "loss": 0.0001,
      "step": 5134
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.708313233638849e-05,
      "loss": 0.0447,
      "step": 5135
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.7050972825213057e-05,
      "loss": 0.0012,
      "step": 5136
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.7018813314037634e-05,
      "loss": 0.0011,
      "step": 5137
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.69866538028622e-05,
      "loss": 0.0139,
      "step": 5138
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.695449429168677e-05,
      "loss": 0.0068,
      "step": 5139
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.6922334780511345e-05,
      "loss": 0.0014,
      "step": 5140
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.689017526933591e-05,
      "loss": 0.0165,
      "step": 5141
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.685801575816048e-05,
      "loss": 0.0002,
      "step": 5142
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.682585624698504e-05,
      "loss": 0.0328,
      "step": 5143
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.679369673580962e-05,
      "loss": 0.0048,
      "step": 5144
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.6761537224634184e-05,
      "loss": 0.0011,
      "step": 5145
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.6729377713458755e-05,
      "loss": 0.0396,
      "step": 5146
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.669721820228333e-05,
      "loss": 0.0012,
      "step": 5147
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.6665058691107896e-05,
      "loss": 0.0203,
      "step": 5148
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.6632899179932466e-05,
      "loss": 0.0062,
      "step": 5149
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.6600739668757044e-05,
      "loss": 0.0099,
      "step": 5150
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.656858015758161e-05,
      "loss": 0.0014,
      "step": 5151
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.653642064640617e-05,
      "loss": 0.0612,
      "step": 5152
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.650426113523075e-05,
      "loss": 0.0357,
      "step": 5153
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.647210162405532e-05,
      "loss": 0.0015,
      "step": 5154
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.643994211287988e-05,
      "loss": 0.0097,
      "step": 5155
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.640778260170446e-05,
      "loss": 0.0402,
      "step": 5156
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.637562309052903e-05,
      "loss": 0.0,
      "step": 5157
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.6343463579353594e-05,
      "loss": 0.0958,
      "step": 5158
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.631130406817816e-05,
      "loss": 0.0046,
      "step": 5159
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.627914455700274e-05,
      "loss": 0.0472,
      "step": 5160
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.6246985045827306e-05,
      "loss": 0.0044,
      "step": 5161
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.621482553465187e-05,
      "loss": 0.1125,
      "step": 5162
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.618266602347645e-05,
      "loss": 0.0007,
      "step": 5163
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.615050651230102e-05,
      "loss": 0.0024,
      "step": 5164
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.611834700112558e-05,
      "loss": 0.0829,
      "step": 5165
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.608618748995016e-05,
      "loss": 0.0075,
      "step": 5166
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.605402797877473e-05,
      "loss": 0.0199,
      "step": 5167
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.602186846759929e-05,
      "loss": 0.0339,
      "step": 5168
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.598970895642387e-05,
      "loss": 0.0879,
      "step": 5169
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.5957549445248434e-05,
      "loss": 0.0083,
      "step": 5170
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.5925389934073004e-05,
      "loss": 0.0031,
      "step": 5171
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.589323042289758e-05,
      "loss": 0.0044,
      "step": 5172
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.5861070911722145e-05,
      "loss": 0.0393,
      "step": 5173
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.5828911400546716e-05,
      "loss": 0.002,
      "step": 5174
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.579675188937128e-05,
      "loss": 0.066,
      "step": 5175
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.576459237819586e-05,
      "loss": 0.0193,
      "step": 5176
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.573243286702042e-05,
      "loss": 0.0179,
      "step": 5177
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.570027335584499e-05,
      "loss": 0.0037,
      "step": 5178
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.566811384466957e-05,
      "loss": 0.0704,
      "step": 5179
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.563595433349413e-05,
      "loss": 0.0281,
      "step": 5180
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.56037948223187e-05,
      "loss": 0.0633,
      "step": 5181
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.557163531114328e-05,
      "loss": 0.0077,
      "step": 5182
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.5539475799967843e-05,
      "loss": 0.0031,
      "step": 5183
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.550731628879241e-05,
      "loss": 0.0434,
      "step": 5184
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.5475156777616984e-05,
      "loss": 0.0052,
      "step": 5185
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.5442997266441555e-05,
      "loss": 0.0402,
      "step": 5186
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.541083775526612e-05,
      "loss": 0.0532,
      "step": 5187
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.537867824409069e-05,
      "loss": 0.0118,
      "step": 5188
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.5346518732915267e-05,
      "loss": 0.012,
      "step": 5189
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.531435922173983e-05,
      "loss": 0.0181,
      "step": 5190
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.5282199710564394e-05,
      "loss": 0.0867,
      "step": 5191
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.525004019938898e-05,
      "loss": 0.0718,
      "step": 5192
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.521788068821354e-05,
      "loss": 0.0131,
      "step": 5193
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.5185721177038106e-05,
      "loss": 0.0003,
      "step": 5194
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.515356166586268e-05,
      "loss": 0.0155,
      "step": 5195
    },
    {
      "epoch": 7.51,
      "learning_rate": 5.512140215468725e-05,
      "loss": 0.0153,
      "step": 5196
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.508924264351182e-05,
      "loss": 0.0368,
      "step": 5197
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.5057083132336394e-05,
      "loss": 0.0226,
      "step": 5198
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.5024923621160965e-05,
      "loss": 0.0003,
      "step": 5199
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.499276410998553e-05,
      "loss": 0.0425,
      "step": 5200
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.4960604598810106e-05,
      "loss": 0.0065,
      "step": 5201
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.492844508763467e-05,
      "loss": 0.008,
      "step": 5202
    },
    {
      "epoch": 7.52,
      "learning_rate": 5.489628557645924e-05,
      "loss": 0.1279,
      "step": 5203
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.4864126065283804e-05,
      "loss": 0.0321,
      "step": 5204
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.483196655410838e-05,
      "loss": 0.0027,
      "step": 5205
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.479980704293295e-05,
      "loss": 0.0363,
      "step": 5206
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.4767647531757515e-05,
      "loss": 0.0059,
      "step": 5207
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.473548802058209e-05,
      "loss": 0.0576,
      "step": 5208
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.4703328509406656e-05,
      "loss": 0.0035,
      "step": 5209
    },
    {
      "epoch": 7.53,
      "learning_rate": 5.467116899823123e-05,
      "loss": 0.1033,
      "step": 5210
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.4639009487055804e-05,
      "loss": 0.0077,
      "step": 5211
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.460684997588037e-05,
      "loss": 0.0003,
      "step": 5212
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.457469046470494e-05,
      "loss": 0.0153,
      "step": 5213
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.4542530953529516e-05,
      "loss": 0.0074,
      "step": 5214
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.451037144235408e-05,
      "loss": 0.0172,
      "step": 5215
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.447821193117864e-05,
      "loss": 0.0105,
      "step": 5216
    },
    {
      "epoch": 7.54,
      "learning_rate": 5.444605242000323e-05,
      "loss": 0.027,
      "step": 5217
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.441389290882779e-05,
      "loss": 0.0027,
      "step": 5218
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.4381733397652355e-05,
      "loss": 0.0434,
      "step": 5219
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.4349573886476925e-05,
      "loss": 0.0977,
      "step": 5220
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.43174143753015e-05,
      "loss": 0.0046,
      "step": 5221
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.4285254864126066e-05,
      "loss": 0.0012,
      "step": 5222
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.425309535295064e-05,
      "loss": 0.0041,
      "step": 5223
    },
    {
      "epoch": 7.55,
      "learning_rate": 5.4220935841775214e-05,
      "loss": 0.0,
      "step": 5224
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.418877633059978e-05,
      "loss": 0.0019,
      "step": 5225
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.415661681942434e-05,
      "loss": 0.0187,
      "step": 5226
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.412445730824892e-05,
      "loss": 0.1328,
      "step": 5227
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.409229779707349e-05,
      "loss": 0.0,
      "step": 5228
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.406013828589805e-05,
      "loss": 0.043,
      "step": 5229
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.402797877472263e-05,
      "loss": 0.0564,
      "step": 5230
    },
    {
      "epoch": 7.56,
      "learning_rate": 5.39958192635472e-05,
      "loss": 0.1058,
      "step": 5231
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.3963659752371765e-05,
      "loss": 0.0017,
      "step": 5232
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.393150024119633e-05,
      "loss": 0.0001,
      "step": 5233
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.3899340730020906e-05,
      "loss": 0.0067,
      "step": 5234
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.3867181218845476e-05,
      "loss": 0.014,
      "step": 5235
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.383502170767004e-05,
      "loss": 0.0757,
      "step": 5236
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.380286219649462e-05,
      "loss": 0.0313,
      "step": 5237
    },
    {
      "epoch": 7.57,
      "learning_rate": 5.377070268531919e-05,
      "loss": 0.0104,
      "step": 5238
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.373854317414375e-05,
      "loss": 0.0,
      "step": 5239
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.370638366296833e-05,
      "loss": 0.0052,
      "step": 5240
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.367422415179289e-05,
      "loss": 0.0089,
      "step": 5241
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.364206464061746e-05,
      "loss": 0.0534,
      "step": 5242
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.360990512944204e-05,
      "loss": 0.0,
      "step": 5243
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.3577745618266604e-05,
      "loss": 0.0048,
      "step": 5244
    },
    {
      "epoch": 7.58,
      "learning_rate": 5.3545586107091175e-05,
      "loss": 0.0044,
      "step": 5245
    },
    {
      "epoch": 7.59,
      "learning_rate": 5.351342659591575e-05,
      "loss": 0.0056,
      "step": 5246
    },
    {
      "epoch": 7.59,
      "learning_rate": 5.3481267084740316e-05,
      "loss": 0.0007,
      "step": 5247
    },
    {
      "epoch": 7.59,
      "learning_rate": 5.3449107573564886e-05,
      "loss": 0.0003,
      "step": 5248
    },
    {
      "epoch": 7.59,
      "learning_rate": 5.341694806238945e-05,
      "loss": 0.001,
      "step": 5249
    },
    {
      "epoch": 7.59,
      "learning_rate": 5.338478855121403e-05,
      "loss": 0.0003,
      "step": 5250
    },
    {
      "epoch": 7.59,
      "learning_rate": 5.335262904003859e-05,
      "loss": 0.0384,
      "step": 5251
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.332046952886316e-05,
      "loss": 0.0,
      "step": 5252
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.328831001768774e-05,
      "loss": 0.0757,
      "step": 5253
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.32561505065123e-05,
      "loss": 0.0318,
      "step": 5254
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.322399099533687e-05,
      "loss": 0.0266,
      "step": 5255
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.319183148416145e-05,
      "loss": 0.004,
      "step": 5256
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.3159671972986014e-05,
      "loss": 0.0003,
      "step": 5257
    },
    {
      "epoch": 7.6,
      "learning_rate": 5.312751246181058e-05,
      "loss": 0.0822,
      "step": 5258
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.3095352950635155e-05,
      "loss": 0.0003,
      "step": 5259
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.3063193439459725e-05,
      "loss": 0.0051,
      "step": 5260
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.303103392828429e-05,
      "loss": 0.069,
      "step": 5261
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.2998874417108867e-05,
      "loss": 0.0134,
      "step": 5262
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.296671490593344e-05,
      "loss": 0.0442,
      "step": 5263
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.2934555394758e-05,
      "loss": 0.0024,
      "step": 5264
    },
    {
      "epoch": 7.61,
      "learning_rate": 5.2902395883582564e-05,
      "loss": 0.0346,
      "step": 5265
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.287023637240714e-05,
      "loss": 0.0178,
      "step": 5266
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.283807686123171e-05,
      "loss": 0.0051,
      "step": 5267
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.2805917350056276e-05,
      "loss": 0.0535,
      "step": 5268
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.277375783888085e-05,
      "loss": 0.0826,
      "step": 5269
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.2741598327705424e-05,
      "loss": 0.0526,
      "step": 5270
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.270943881652999e-05,
      "loss": 0.053,
      "step": 5271
    },
    {
      "epoch": 7.62,
      "learning_rate": 5.2677279305354565e-05,
      "loss": 0.0035,
      "step": 5272
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.264511979417913e-05,
      "loss": 0.0,
      "step": 5273
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.26129602830037e-05,
      "loss": 0.0477,
      "step": 5274
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.2580800771828276e-05,
      "loss": 0.0034,
      "step": 5275
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.254864126065284e-05,
      "loss": 0.002,
      "step": 5276
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.251648174947741e-05,
      "loss": 0.0372,
      "step": 5277
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.2484322238301974e-05,
      "loss": 0.0759,
      "step": 5278
    },
    {
      "epoch": 7.63,
      "learning_rate": 5.245216272712655e-05,
      "loss": 0.0011,
      "step": 5279
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.242000321595112e-05,
      "loss": 0.0188,
      "step": 5280
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.2387843704775686e-05,
      "loss": 0.0084,
      "step": 5281
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.235568419360026e-05,
      "loss": 0.0286,
      "step": 5282
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.232352468242483e-05,
      "loss": 0.0751,
      "step": 5283
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.22913651712494e-05,
      "loss": 0.0032,
      "step": 5284
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.2259205660073975e-05,
      "loss": 0.0054,
      "step": 5285
    },
    {
      "epoch": 7.64,
      "learning_rate": 5.222704614889854e-05,
      "loss": 0.0035,
      "step": 5286
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.219488663772311e-05,
      "loss": 0.0012,
      "step": 5287
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.2162727126547686e-05,
      "loss": 0.0,
      "step": 5288
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.213056761537225e-05,
      "loss": 0.0004,
      "step": 5289
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.2098408104196814e-05,
      "loss": 0.0105,
      "step": 5290
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.206624859302139e-05,
      "loss": 0.0,
      "step": 5291
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.203408908184596e-05,
      "loss": 0.0235,
      "step": 5292
    },
    {
      "epoch": 7.65,
      "learning_rate": 5.2001929570670525e-05,
      "loss": 0.0009,
      "step": 5293
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.1969770059495096e-05,
      "loss": 0.0396,
      "step": 5294
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.193761054831967e-05,
      "loss": 0.0767,
      "step": 5295
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.190545103714424e-05,
      "loss": 0.0176,
      "step": 5296
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.18732915259688e-05,
      "loss": 0.0063,
      "step": 5297
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.184113201479338e-05,
      "loss": 0.0978,
      "step": 5298
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.180897250361795e-05,
      "loss": 0.0192,
      "step": 5299
    },
    {
      "epoch": 7.66,
      "learning_rate": 5.177681299244251e-05,
      "loss": 0.0195,
      "step": 5300
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.174465348126709e-05,
      "loss": 0.0085,
      "step": 5301
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.171249397009166e-05,
      "loss": 0.1357,
      "step": 5302
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.1680334458916224e-05,
      "loss": 0.0158,
      "step": 5303
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.16481749477408e-05,
      "loss": 0.0059,
      "step": 5304
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.161601543656537e-05,
      "loss": 0.0232,
      "step": 5305
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.1583855925389935e-05,
      "loss": 0.0051,
      "step": 5306
    },
    {
      "epoch": 7.67,
      "learning_rate": 5.15516964142145e-05,
      "loss": 0.0154,
      "step": 5307
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.1519536903039076e-05,
      "loss": 0.004,
      "step": 5308
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.148737739186365e-05,
      "loss": 0.0716,
      "step": 5309
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.145521788068821e-05,
      "loss": 0.0001,
      "step": 5310
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.142305836951279e-05,
      "loss": 0.0073,
      "step": 5311
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.139089885833736e-05,
      "loss": 0.0055,
      "step": 5312
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.135873934716192e-05,
      "loss": 0.0278,
      "step": 5313
    },
    {
      "epoch": 7.68,
      "learning_rate": 5.13265798359865e-05,
      "loss": 0.0248,
      "step": 5314
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.129442032481106e-05,
      "loss": 0.0003,
      "step": 5315
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.1262260813635634e-05,
      "loss": 0.0028,
      "step": 5316
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.123010130246021e-05,
      "loss": 0.0271,
      "step": 5317
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.1197941791284775e-05,
      "loss": 0.0006,
      "step": 5318
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.1165782280109345e-05,
      "loss": 0.0001,
      "step": 5319
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.113362276893392e-05,
      "loss": 0.0,
      "step": 5320
    },
    {
      "epoch": 7.69,
      "learning_rate": 5.1101463257758486e-05,
      "loss": 0.0602,
      "step": 5321
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.106930374658305e-05,
      "loss": 0.0391,
      "step": 5322
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.103714423540762e-05,
      "loss": 0.0013,
      "step": 5323
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.10049847242322e-05,
      "loss": 0.0075,
      "step": 5324
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.097282521305676e-05,
      "loss": 0.0041,
      "step": 5325
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.094066570188133e-05,
      "loss": 0.0152,
      "step": 5326
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.090850619070591e-05,
      "loss": 0.0001,
      "step": 5327
    },
    {
      "epoch": 7.7,
      "learning_rate": 5.087634667953047e-05,
      "loss": 0.0412,
      "step": 5328
    },
    {
      "epoch": 7.71,
      "learning_rate": 5.084418716835504e-05,
      "loss": 0.0066,
      "step": 5329
    },
    {
      "epoch": 7.71,
      "learning_rate": 5.0812027657179614e-05,
      "loss": 0.0072,
      "step": 5330
    },
    {
      "epoch": 7.71,
      "learning_rate": 5.0779868146004184e-05,
      "loss": 0.0294,
      "step": 5331
    },
    {
      "epoch": 7.71,
      "learning_rate": 5.074770863482875e-05,
      "loss": 0.055,
      "step": 5332
    },
    {
      "epoch": 7.71,
      "learning_rate": 5.0715549123653325e-05,
      "loss": 0.0095,
      "step": 5333
    },
    {
      "epoch": 7.71,
      "learning_rate": 5.0683389612477896e-05,
      "loss": 0.0005,
      "step": 5334
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.065123010130246e-05,
      "loss": 0.0283,
      "step": 5335
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.061907059012704e-05,
      "loss": 0.0164,
      "step": 5336
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.058691107895161e-05,
      "loss": 0.005,
      "step": 5337
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.055475156777617e-05,
      "loss": 0.0034,
      "step": 5338
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.0522592056600735e-05,
      "loss": 0.0023,
      "step": 5339
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.049043254542531e-05,
      "loss": 0.0002,
      "step": 5340
    },
    {
      "epoch": 7.72,
      "learning_rate": 5.045827303424988e-05,
      "loss": 0.0005,
      "step": 5341
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.0426113523074446e-05,
      "loss": 0.0012,
      "step": 5342
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.0393954011899024e-05,
      "loss": 0.0001,
      "step": 5343
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.0361794500723594e-05,
      "loss": 0.0015,
      "step": 5344
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.032963498954816e-05,
      "loss": 0.0031,
      "step": 5345
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.0297475478372735e-05,
      "loss": 0.0292,
      "step": 5346
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.02653159671973e-05,
      "loss": 0.0084,
      "step": 5347
    },
    {
      "epoch": 7.73,
      "learning_rate": 5.023315645602187e-05,
      "loss": 0.0226,
      "step": 5348
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.020099694484645e-05,
      "loss": 0.0117,
      "step": 5349
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.016883743367101e-05,
      "loss": 0.0019,
      "step": 5350
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.013667792249558e-05,
      "loss": 0.0294,
      "step": 5351
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.0104518411320145e-05,
      "loss": 0.0,
      "step": 5352
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.007235890014472e-05,
      "loss": 0.0134,
      "step": 5353
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.0040199388969286e-05,
      "loss": 0.004,
      "step": 5354
    },
    {
      "epoch": 7.74,
      "learning_rate": 5.0008039877793856e-05,
      "loss": 0.0109,
      "step": 5355
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.997588036661843e-05,
      "loss": 0.0542,
      "step": 5356
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.9943720855443e-05,
      "loss": 0.0196,
      "step": 5357
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.991156134426757e-05,
      "loss": 0.0066,
      "step": 5358
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.987940183309214e-05,
      "loss": 0.0039,
      "step": 5359
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.984724232191671e-05,
      "loss": 0.0378,
      "step": 5360
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.981508281074128e-05,
      "loss": 0.0,
      "step": 5361
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.978292329956585e-05,
      "loss": 0.0004,
      "step": 5362
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.975076378839042e-05,
      "loss": 0.0052,
      "step": 5363
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.9718604277214984e-05,
      "loss": 0.0013,
      "step": 5364
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.968644476603956e-05,
      "loss": 0.0218,
      "step": 5365
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.965428525486413e-05,
      "loss": 0.0,
      "step": 5366
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.9622125743688696e-05,
      "loss": 0.0206,
      "step": 5367
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.9589966232513266e-05,
      "loss": 0.0072,
      "step": 5368
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.9557806721337844e-05,
      "loss": 0.0131,
      "step": 5369
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.952564721016241e-05,
      "loss": 0.0023,
      "step": 5370
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.949348769898698e-05,
      "loss": 0.0029,
      "step": 5371
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.946132818781155e-05,
      "loss": 0.0079,
      "step": 5372
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.942916867663612e-05,
      "loss": 0.0399,
      "step": 5373
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.939700916546069e-05,
      "loss": 0.0068,
      "step": 5374
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.936484965428525e-05,
      "loss": 0.03,
      "step": 5375
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.933269014310983e-05,
      "loss": 0.0205,
      "step": 5376
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.93005306319344e-05,
      "loss": 0.0269,
      "step": 5377
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.9268371120758965e-05,
      "loss": 0.0069,
      "step": 5378
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.9236211609583535e-05,
      "loss": 0.0127,
      "step": 5379
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.9204052098408106e-05,
      "loss": 0.0808,
      "step": 5380
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.9171892587232676e-05,
      "loss": 0.0026,
      "step": 5381
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.913973307605725e-05,
      "loss": 0.0233,
      "step": 5382
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.910757356488182e-05,
      "loss": 0.0457,
      "step": 5383
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.907541405370639e-05,
      "loss": 0.0001,
      "step": 5384
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.904325454253096e-05,
      "loss": 0.0691,
      "step": 5385
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.901109503135552e-05,
      "loss": 0.0002,
      "step": 5386
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.89789355201801e-05,
      "loss": 0.0204,
      "step": 5387
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.894677600900466e-05,
      "loss": 0.0049,
      "step": 5388
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.8914616497829233e-05,
      "loss": 0.0238,
      "step": 5389
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.8882456986653804e-05,
      "loss": 0.0016,
      "step": 5390
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.8850297475478374e-05,
      "loss": 0.003,
      "step": 5391
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.8818137964302945e-05,
      "loss": 0.0058,
      "step": 5392
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.8785978453127516e-05,
      "loss": 0.0002,
      "step": 5393
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.8753818941952086e-05,
      "loss": 0.0121,
      "step": 5394
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.8721659430776657e-05,
      "loss": 0.0004,
      "step": 5395
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.868949991960122e-05,
      "loss": 0.0032,
      "step": 5396
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.86573404084258e-05,
      "loss": 0.0459,
      "step": 5397
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.862518089725037e-05,
      "loss": 0.0405,
      "step": 5398
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.859302138607493e-05,
      "loss": 0.0084,
      "step": 5399
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.85608618748995e-05,
      "loss": 0.0023,
      "step": 5400
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.852870236372407e-05,
      "loss": 0.0613,
      "step": 5401
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.849654285254864e-05,
      "loss": 0.0159,
      "step": 5402
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.8464383341373214e-05,
      "loss": 0.0188,
      "step": 5403
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.8432223830197784e-05,
      "loss": 0.0011,
      "step": 5404
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.8400064319022355e-05,
      "loss": 0.0033,
      "step": 5405
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.8367904807846925e-05,
      "loss": 0.0779,
      "step": 5406
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.833574529667149e-05,
      "loss": 0.0096,
      "step": 5407
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.8303585785496066e-05,
      "loss": 0.0278,
      "step": 5408
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.827142627432063e-05,
      "loss": 0.0471,
      "step": 5409
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.82392667631452e-05,
      "loss": 0.0064,
      "step": 5410
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.820710725196977e-05,
      "loss": 0.001,
      "step": 5411
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.817494774079434e-05,
      "loss": 0.1589,
      "step": 5412
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.814278822961891e-05,
      "loss": 0.068,
      "step": 5413
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.811062871844348e-05,
      "loss": 0.0023,
      "step": 5414
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.807846920726805e-05,
      "loss": 0.0397,
      "step": 5415
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.8046309696092624e-05,
      "loss": 0.0596,
      "step": 5416
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.801415018491719e-05,
      "loss": 0.0009,
      "step": 5417
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.798199067374176e-05,
      "loss": 0.0002,
      "step": 5418
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.7949831162566335e-05,
      "loss": 0.0327,
      "step": 5419
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.79176716513909e-05,
      "loss": 0.0874,
      "step": 5420
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.788551214021547e-05,
      "loss": 0.0045,
      "step": 5421
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.785335262904005e-05,
      "loss": 0.004,
      "step": 5422
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.782119311786461e-05,
      "loss": 0.0553,
      "step": 5423
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.778903360668918e-05,
      "loss": 0.0469,
      "step": 5424
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.775687409551375e-05,
      "loss": 0.0003,
      "step": 5425
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.772471458433832e-05,
      "loss": 0.0319,
      "step": 5426
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.769255507316289e-05,
      "loss": 0.0974,
      "step": 5427
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.7660395561987456e-05,
      "loss": 0.0018,
      "step": 5428
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.7628236050812034e-05,
      "loss": 0.0022,
      "step": 5429
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.75960765396366e-05,
      "loss": 0.0007,
      "step": 5430
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.756391702846117e-05,
      "loss": 0.0031,
      "step": 5431
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.753175751728574e-05,
      "loss": 0.0137,
      "step": 5432
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.749959800611031e-05,
      "loss": 0.2048,
      "step": 5433
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.746743849493488e-05,
      "loss": 0.0135,
      "step": 5434
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.743527898375945e-05,
      "loss": 0.0204,
      "step": 5435
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.740311947258402e-05,
      "loss": 0.0001,
      "step": 5436
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.737095996140859e-05,
      "loss": 0.0033,
      "step": 5437
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.7338800450233155e-05,
      "loss": 0.0335,
      "step": 5438
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.7306640939057725e-05,
      "loss": 0.0041,
      "step": 5439
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.72744814278823e-05,
      "loss": 0.0552,
      "step": 5440
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.7242321916706866e-05,
      "loss": 0.0167,
      "step": 5441
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.721016240553144e-05,
      "loss": 0.055,
      "step": 5442
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.717800289435601e-05,
      "loss": 0.0112,
      "step": 5443
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.714584338318058e-05,
      "loss": 0.0021,
      "step": 5444
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.711368387200515e-05,
      "loss": 0.0315,
      "step": 5445
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.708152436082971e-05,
      "loss": 0.058,
      "step": 5446
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.704936484965429e-05,
      "loss": 0.0173,
      "step": 5447
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.701720533847886e-05,
      "loss": 0.0612,
      "step": 5448
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.6985045827303424e-05,
      "loss": 0.0427,
      "step": 5449
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.6952886316128e-05,
      "loss": 0.06,
      "step": 5450
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.692072680495257e-05,
      "loss": 0.0146,
      "step": 5451
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.6888567293777135e-05,
      "loss": 0.0493,
      "step": 5452
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.6856407782601706e-05,
      "loss": 0.0233,
      "step": 5453
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.6824248271426276e-05,
      "loss": 0.0005,
      "step": 5454
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.679208876025085e-05,
      "loss": 0.0173,
      "step": 5455
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.675992924907542e-05,
      "loss": 0.0796,
      "step": 5456
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.672776973789999e-05,
      "loss": 0.0044,
      "step": 5457
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.669561022672456e-05,
      "loss": 0.0019,
      "step": 5458
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.666345071554913e-05,
      "loss": 0.0292,
      "step": 5459
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.663129120437369e-05,
      "loss": 0.0066,
      "step": 5460
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.659913169319827e-05,
      "loss": 0.0588,
      "step": 5461
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.6566972182022833e-05,
      "loss": 0.0029,
      "step": 5462
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.6534812670847404e-05,
      "loss": 0.066,
      "step": 5463
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.6502653159671974e-05,
      "loss": 0.0012,
      "step": 5464
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.6470493648496545e-05,
      "loss": 0.052,
      "step": 5465
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.6438334137321115e-05,
      "loss": 0.0073,
      "step": 5466
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.6406174626145686e-05,
      "loss": 0.0015,
      "step": 5467
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.6374015114970256e-05,
      "loss": 0.0001,
      "step": 5468
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.634185560379483e-05,
      "loss": 0.054,
      "step": 5469
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.630969609261939e-05,
      "loss": 0.0006,
      "step": 5470
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.627753658144396e-05,
      "loss": 0.0075,
      "step": 5471
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.624537707026854e-05,
      "loss": 0.0701,
      "step": 5472
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.62132175590931e-05,
      "loss": 0.0331,
      "step": 5473
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.618105804791767e-05,
      "loss": 0.064,
      "step": 5474
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.614889853674224e-05,
      "loss": 0.0027,
      "step": 5475
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.6116739025566814e-05,
      "loss": 0.0266,
      "step": 5476
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.6084579514391384e-05,
      "loss": 0.059,
      "step": 5477
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.605242000321595e-05,
      "loss": 0.0002,
      "step": 5478
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.6020260492040525e-05,
      "loss": 0.0409,
      "step": 5479
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.5988100980865096e-05,
      "loss": 0.0609,
      "step": 5480
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.595594146968966e-05,
      "loss": 0.0045,
      "step": 5481
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.592378195851424e-05,
      "loss": 0.0333,
      "step": 5482
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.58916224473388e-05,
      "loss": 0.0164,
      "step": 5483
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.585946293616337e-05,
      "loss": 0.0001,
      "step": 5484
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.582730342498794e-05,
      "loss": 0.0235,
      "step": 5485
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.579514391381251e-05,
      "loss": 0.0313,
      "step": 5486
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.576298440263708e-05,
      "loss": 0.0022,
      "step": 5487
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.573082489146165e-05,
      "loss": 0.0065,
      "step": 5488
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.5698665380286224e-05,
      "loss": 0.0107,
      "step": 5489
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.5666505869110794e-05,
      "loss": 0.0002,
      "step": 5490
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.563434635793536e-05,
      "loss": 0.0548,
      "step": 5491
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.560218684675993e-05,
      "loss": 0.0069,
      "step": 5492
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.5570027335584506e-05,
      "loss": 0.0122,
      "step": 5493
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.553786782440907e-05,
      "loss": 0.0064,
      "step": 5494
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.550570831323364e-05,
      "loss": 0.1196,
      "step": 5495
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.547354880205821e-05,
      "loss": 0.0,
      "step": 5496
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.544138929088278e-05,
      "loss": 0.0294,
      "step": 5497
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.540922977970735e-05,
      "loss": 0.0138,
      "step": 5498
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.5377070268531915e-05,
      "loss": 0.0018,
      "step": 5499
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.534491075735649e-05,
      "loss": 0.0454,
      "step": 5500
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.531275124618106e-05,
      "loss": 0.0003,
      "step": 5501
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.528059173500563e-05,
      "loss": 0.1141,
      "step": 5502
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.52484322238302e-05,
      "loss": 0.0114,
      "step": 5503
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.5216272712654775e-05,
      "loss": 0.012,
      "step": 5504
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.518411320147934e-05,
      "loss": 0.0494,
      "step": 5505
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.515195369030391e-05,
      "loss": 0.0344,
      "step": 5506
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.511979417912848e-05,
      "loss": 0.0719,
      "step": 5507
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.508763466795305e-05,
      "loss": 0.0512,
      "step": 5508
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.505547515677762e-05,
      "loss": 0.01,
      "step": 5509
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.502331564560219e-05,
      "loss": 0.0066,
      "step": 5510
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.499115613442676e-05,
      "loss": 0.0048,
      "step": 5511
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.4958996623251325e-05,
      "loss": 0.0066,
      "step": 5512
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.4926837112075896e-05,
      "loss": 0.0896,
      "step": 5513
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.489467760090047e-05,
      "loss": 0.0517,
      "step": 5514
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.486251808972504e-05,
      "loss": 0.0013,
      "step": 5515
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.483035857854961e-05,
      "loss": 0.0421,
      "step": 5516
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.479819906737418e-05,
      "loss": 0.0118,
      "step": 5517
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.476603955619875e-05,
      "loss": 0.074,
      "step": 5518
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.473388004502332e-05,
      "loss": 0.0369,
      "step": 5519
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.470172053384788e-05,
      "loss": 0.0052,
      "step": 5520
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.466956102267246e-05,
      "loss": 0.0456,
      "step": 5521
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.463740151149703e-05,
      "loss": 0.009,
      "step": 5522
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.4605242000321594e-05,
      "loss": 0.0002,
      "step": 5523
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.4573082489146165e-05,
      "loss": 0.0618,
      "step": 5524
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.454092297797074e-05,
      "loss": 0.0438,
      "step": 5525
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.4508763466795306e-05,
      "loss": 0.0016,
      "step": 5526
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.4476603955619876e-05,
      "loss": 0.017,
      "step": 5527
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.4444444444444447e-05,
      "loss": 0.0654,
      "step": 5528
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.441228493326902e-05,
      "loss": 0.0001,
      "step": 5529
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.438012542209359e-05,
      "loss": 0.0176,
      "step": 5530
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.434796591091815e-05,
      "loss": 0.0081,
      "step": 5531
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.431580639974273e-05,
      "loss": 0.0391,
      "step": 5532
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.42836468885673e-05,
      "loss": 0.0115,
      "step": 5533
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.425148737739186e-05,
      "loss": 0.0005,
      "step": 5534
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.421932786621643e-05,
      "loss": 0.0103,
      "step": 5535
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.4187168355041004e-05,
      "loss": 0.0006,
      "step": 5536
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.4155008843865574e-05,
      "loss": 0.0812,
      "step": 5537
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.4122849332690145e-05,
      "loss": 0.0002,
      "step": 5538
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.4090689821514715e-05,
      "loss": 0.0003,
      "step": 5539
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.4058530310339286e-05,
      "loss": 0.011,
      "step": 5540
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.4026370799163856e-05,
      "loss": 0.0282,
      "step": 5541
    },
    {
      "epoch": 8.01,
      "learning_rate": 4.399421128798843e-05,
      "loss": 0.051,
      "step": 5542
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.3962051776813e-05,
      "loss": 0.0207,
      "step": 5543
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.392989226563756e-05,
      "loss": 0.0497,
      "step": 5544
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.389773275446213e-05,
      "loss": 0.0004,
      "step": 5545
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.386557324328671e-05,
      "loss": 0.0024,
      "step": 5546
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.383341373211127e-05,
      "loss": 0.0064,
      "step": 5547
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.380125422093584e-05,
      "loss": 0.0068,
      "step": 5548
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.3769094709760414e-05,
      "loss": 0.0258,
      "step": 5549
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.3736935198584984e-05,
      "loss": 0.0067,
      "step": 5550
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.3704775687409555e-05,
      "loss": 0.0002,
      "step": 5551
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.367261617623412e-05,
      "loss": 0.0018,
      "step": 5552
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.3640456665058696e-05,
      "loss": 0.0149,
      "step": 5553
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.3608297153883266e-05,
      "loss": 0.0079,
      "step": 5554
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.357613764270783e-05,
      "loss": 0.0328,
      "step": 5555
    },
    {
      "epoch": 8.03,
      "learning_rate": 4.35439781315324e-05,
      "loss": 0.0522,
      "step": 5556
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.351181862035697e-05,
      "loss": 0.0459,
      "step": 5557
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.347965910918154e-05,
      "loss": 0.0157,
      "step": 5558
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.344749959800611e-05,
      "loss": 0.0312,
      "step": 5559
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.341534008683068e-05,
      "loss": 0.004,
      "step": 5560
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.338318057565525e-05,
      "loss": 0.0734,
      "step": 5561
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.3351021064479824e-05,
      "loss": 0.0062,
      "step": 5562
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.331886155330439e-05,
      "loss": 0.0,
      "step": 5563
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.3286702042128965e-05,
      "loss": 0.014,
      "step": 5564
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.325454253095353e-05,
      "loss": 0.0001,
      "step": 5565
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.32223830197781e-05,
      "loss": 0.0238,
      "step": 5566
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.3190223508602676e-05,
      "loss": 0.0352,
      "step": 5567
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.315806399742724e-05,
      "loss": 0.0432,
      "step": 5568
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.312590448625181e-05,
      "loss": 0.0707,
      "step": 5569
    },
    {
      "epoch": 8.05,
      "learning_rate": 4.309374497507638e-05,
      "loss": 0.0004,
      "step": 5570
    },
    {
      "epoch": 8.06,
      "learning_rate": 4.306158546390095e-05,
      "loss": 0.0037,
      "step": 5571
    },
    {
      "epoch": 8.06,
      "learning_rate": 4.302942595272552e-05,
      "loss": 0.0017,
      "step": 5572
    },
    {
      "epoch": 8.06,
      "learning_rate": 4.2997266441550086e-05,
      "loss": 0.0637,
      "step": 5573
    },
    {
      "epoch": 8.06,
      "learning_rate": 4.296510693037466e-05,
      "loss": 0.0316,
      "step": 5574
    },
    {
      "epoch": 8.06,
      "learning_rate": 4.2932947419199234e-05,
      "loss": 0.0172,
      "step": 5575
    },
    {
      "epoch": 8.06,
      "learning_rate": 4.29007879080238e-05,
      "loss": 0.0029,
      "step": 5576
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.286862839684837e-05,
      "loss": 0.0023,
      "step": 5577
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.2836468885672945e-05,
      "loss": 0.0405,
      "step": 5578
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.280430937449751e-05,
      "loss": 0.0066,
      "step": 5579
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.277214986332208e-05,
      "loss": 0.0384,
      "step": 5580
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.273999035214665e-05,
      "loss": 0.0035,
      "step": 5581
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.270783084097122e-05,
      "loss": 0.0048,
      "step": 5582
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.267567132979579e-05,
      "loss": 0.0894,
      "step": 5583
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.2643511818620355e-05,
      "loss": 0.003,
      "step": 5584
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.261135230744493e-05,
      "loss": 0.0081,
      "step": 5585
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.25791927962695e-05,
      "loss": 0.0474,
      "step": 5586
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.2547033285094066e-05,
      "loss": 0.0414,
      "step": 5587
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.251487377391864e-05,
      "loss": 0.0176,
      "step": 5588
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.248271426274321e-05,
      "loss": 0.08,
      "step": 5589
    },
    {
      "epoch": 8.08,
      "learning_rate": 4.245055475156778e-05,
      "loss": 0.022,
      "step": 5590
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.241839524039235e-05,
      "loss": 0.0469,
      "step": 5591
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.238623572921692e-05,
      "loss": 0.0714,
      "step": 5592
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.235407621804149e-05,
      "loss": 0.0525,
      "step": 5593
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.232191670686605e-05,
      "loss": 0.0057,
      "step": 5594
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.228975719569063e-05,
      "loss": 0.0371,
      "step": 5595
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.22575976845152e-05,
      "loss": 0.0371,
      "step": 5596
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.2225438173339764e-05,
      "loss": 0.0004,
      "step": 5597
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.2193278662164335e-05,
      "loss": 0.034,
      "step": 5598
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.216111915098891e-05,
      "loss": 0.0374,
      "step": 5599
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.2128959639813476e-05,
      "loss": 0.0133,
      "step": 5600
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.2096800128638047e-05,
      "loss": 0.063,
      "step": 5601
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.206464061746262e-05,
      "loss": 0.0322,
      "step": 5602
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.203248110628719e-05,
      "loss": 0.0041,
      "step": 5603
    },
    {
      "epoch": 8.1,
      "learning_rate": 4.200032159511176e-05,
      "loss": 0.0286,
      "step": 5604
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.196816208393632e-05,
      "loss": 0.007,
      "step": 5605
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.19360025727609e-05,
      "loss": 0.0645,
      "step": 5606
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.190384306158547e-05,
      "loss": 0.0153,
      "step": 5607
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.187168355041003e-05,
      "loss": 0.0148,
      "step": 5608
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.1839524039234604e-05,
      "loss": 0.0001,
      "step": 5609
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.1807364528059174e-05,
      "loss": 0.0004,
      "step": 5610
    },
    {
      "epoch": 8.11,
      "learning_rate": 4.1775205016883745e-05,
      "loss": 0.0007,
      "step": 5611
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.1743045505708315e-05,
      "loss": 0.0806,
      "step": 5612
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.1710885994532886e-05,
      "loss": 0.0006,
      "step": 5613
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.1678726483357456e-05,
      "loss": 0.0138,
      "step": 5614
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.164656697218203e-05,
      "loss": 0.0256,
      "step": 5615
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.161440746100659e-05,
      "loss": 0.0613,
      "step": 5616
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.158224794983117e-05,
      "loss": 0.0319,
      "step": 5617
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.155008843865573e-05,
      "loss": 0.0002,
      "step": 5618
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.15179289274803e-05,
      "loss": 0.0121,
      "step": 5619
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.148576941630487e-05,
      "loss": 0.0001,
      "step": 5620
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.145360990512944e-05,
      "loss": 0.0056,
      "step": 5621
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.1421450393954014e-05,
      "loss": 0.0397,
      "step": 5622
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.1389290882778584e-05,
      "loss": 0.0772,
      "step": 5623
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.1357131371603155e-05,
      "loss": 0.0189,
      "step": 5624
    },
    {
      "epoch": 8.13,
      "learning_rate": 4.1324971860427725e-05,
      "loss": 0.0058,
      "step": 5625
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.129281234925229e-05,
      "loss": 0.0372,
      "step": 5626
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.1260652838076866e-05,
      "loss": 0.0001,
      "step": 5627
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.122849332690144e-05,
      "loss": 0.0234,
      "step": 5628
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.1196333815726e-05,
      "loss": 0.0036,
      "step": 5629
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.116417430455057e-05,
      "loss": 0.0247,
      "step": 5630
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.113201479337515e-05,
      "loss": 0.018,
      "step": 5631
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.109985528219971e-05,
      "loss": 0.0232,
      "step": 5632
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.106769577102428e-05,
      "loss": 0.0192,
      "step": 5633
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.103553625984885e-05,
      "loss": 0.02,
      "step": 5634
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.1003376748673424e-05,
      "loss": 0.0406,
      "step": 5635
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.0971217237497994e-05,
      "loss": 0.0232,
      "step": 5636
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.093905772632256e-05,
      "loss": 0.0126,
      "step": 5637
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.0906898215147135e-05,
      "loss": 0.0301,
      "step": 5638
    },
    {
      "epoch": 8.15,
      "learning_rate": 4.08747387039717e-05,
      "loss": 0.0006,
      "step": 5639
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.084257919279627e-05,
      "loss": 0.0022,
      "step": 5640
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.081041968162084e-05,
      "loss": 0.0002,
      "step": 5641
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.077826017044541e-05,
      "loss": 0.0273,
      "step": 5642
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.074610065926998e-05,
      "loss": 0.0005,
      "step": 5643
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.071394114809455e-05,
      "loss": 0.0164,
      "step": 5644
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.068178163691912e-05,
      "loss": 0.0001,
      "step": 5645
    },
    {
      "epoch": 8.16,
      "learning_rate": 4.064962212574369e-05,
      "loss": 0.0509,
      "step": 5646
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.0617462614568256e-05,
      "loss": 0.0002,
      "step": 5647
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.058530310339283e-05,
      "loss": 0.0072,
      "step": 5648
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.0553143592217404e-05,
      "loss": 0.0012,
      "step": 5649
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.052098408104197e-05,
      "loss": 0.0126,
      "step": 5650
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.048882456986654e-05,
      "loss": 0.0208,
      "step": 5651
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.0456665058691116e-05,
      "loss": 0.0223,
      "step": 5652
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.042450554751568e-05,
      "loss": 0.0107,
      "step": 5653
    },
    {
      "epoch": 8.18,
      "learning_rate": 4.039234603634025e-05,
      "loss": 0.0004,
      "step": 5654
    },
    {
      "epoch": 8.18,
      "learning_rate": 4.036018652516482e-05,
      "loss": 0.0413,
      "step": 5655
    },
    {
      "epoch": 8.18,
      "learning_rate": 4.032802701398939e-05,
      "loss": 0.0282,
      "step": 5656
    },
    {
      "epoch": 8.18,
      "learning_rate": 4.029586750281396e-05,
      "loss": 0.0491,
      "step": 5657
    },
    {
      "epoch": 8.18,
      "learning_rate": 4.0263707991638525e-05,
      "loss": 0.0052,
      "step": 5658
    },
    {
      "epoch": 8.18,
      "learning_rate": 4.02315484804631e-05,
      "loss": 0.0307,
      "step": 5659
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.019938896928767e-05,
      "loss": 0.0,
      "step": 5660
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.0167229458112237e-05,
      "loss": 0.0128,
      "step": 5661
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.013506994693681e-05,
      "loss": 0.0743,
      "step": 5662
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.010291043576138e-05,
      "loss": 0.0005,
      "step": 5663
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.007075092458595e-05,
      "loss": 0.0002,
      "step": 5664
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.003859141341052e-05,
      "loss": 0.0808,
      "step": 5665
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.000643190223509e-05,
      "loss": 0.0454,
      "step": 5666
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.997427239105966e-05,
      "loss": 0.0024,
      "step": 5667
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.994211287988423e-05,
      "loss": 0.0009,
      "step": 5668
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.9909953368708794e-05,
      "loss": 0.0137,
      "step": 5669
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.987779385753337e-05,
      "loss": 0.0018,
      "step": 5670
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.9845634346357935e-05,
      "loss": 0.0928,
      "step": 5671
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.9813474835182505e-05,
      "loss": 0.013,
      "step": 5672
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.9781315324007076e-05,
      "loss": 0.0031,
      "step": 5673
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.9749155812831646e-05,
      "loss": 0.0069,
      "step": 5674
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.971699630165622e-05,
      "loss": 0.0027,
      "step": 5675
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.968483679048078e-05,
      "loss": 0.0152,
      "step": 5676
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.965267727930536e-05,
      "loss": 0.0001,
      "step": 5677
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.962051776812993e-05,
      "loss": 0.0064,
      "step": 5678
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.958835825695449e-05,
      "loss": 0.0007,
      "step": 5679
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.955619874577906e-05,
      "loss": 0.0248,
      "step": 5680
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.952403923460364e-05,
      "loss": 0.0003,
      "step": 5681
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.9491879723428204e-05,
      "loss": 0.0088,
      "step": 5682
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.9459720212252774e-05,
      "loss": 0.0643,
      "step": 5683
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.9427560701077345e-05,
      "loss": 0.0007,
      "step": 5684
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.9395401189901915e-05,
      "loss": 0.0572,
      "step": 5685
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.9363241678726486e-05,
      "loss": 0.0267,
      "step": 5686
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.9331082167551056e-05,
      "loss": 0.0001,
      "step": 5687
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.929892265637563e-05,
      "loss": 0.0145,
      "step": 5688
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.92667631452002e-05,
      "loss": 0.0072,
      "step": 5689
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.923460363402476e-05,
      "loss": 0.026,
      "step": 5690
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.920244412284934e-05,
      "loss": 0.0158,
      "step": 5691
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.91702846116739e-05,
      "loss": 0.0002,
      "step": 5692
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.913812510049847e-05,
      "loss": 0.0013,
      "step": 5693
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.910596558932304e-05,
      "loss": 0.0025,
      "step": 5694
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.9073806078147614e-05,
      "loss": 0.002,
      "step": 5695
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.9041646566972184e-05,
      "loss": 0.0342,
      "step": 5696
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.9009487055796755e-05,
      "loss": 0.0541,
      "step": 5697
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.8977327544621325e-05,
      "loss": 0.0187,
      "step": 5698
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.8945168033445896e-05,
      "loss": 0.0001,
      "step": 5699
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.891300852227046e-05,
      "loss": 0.0011,
      "step": 5700
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.888084901109503e-05,
      "loss": 0.0401,
      "step": 5701
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.884868949991961e-05,
      "loss": 0.0095,
      "step": 5702
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.881652998874417e-05,
      "loss": 0.0198,
      "step": 5703
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.878437047756874e-05,
      "loss": 0.0427,
      "step": 5704
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.875221096639331e-05,
      "loss": 0.0037,
      "step": 5705
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.872005145521788e-05,
      "loss": 0.0125,
      "step": 5706
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.868789194404245e-05,
      "loss": 0.0111,
      "step": 5707
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.865573243286702e-05,
      "loss": 0.0361,
      "step": 5708
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.8623572921691594e-05,
      "loss": 0.0041,
      "step": 5709
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.8591413410516165e-05,
      "loss": 0.1292,
      "step": 5710
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.855925389934073e-05,
      "loss": 0.0086,
      "step": 5711
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.8527094388165306e-05,
      "loss": 0.0001,
      "step": 5712
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.8494934876989876e-05,
      "loss": 0.0618,
      "step": 5713
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.846277536581444e-05,
      "loss": 0.0002,
      "step": 5714
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.843061585463901e-05,
      "loss": 0.0884,
      "step": 5715
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.839845634346358e-05,
      "loss": 0.0002,
      "step": 5716
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.836629683228815e-05,
      "loss": 0.0113,
      "step": 5717
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.833413732111272e-05,
      "loss": 0.0441,
      "step": 5718
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.830197780993729e-05,
      "loss": 0.0002,
      "step": 5719
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.826981829876186e-05,
      "loss": 0.0181,
      "step": 5720
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.823765878758643e-05,
      "loss": 0.0002,
      "step": 5721
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.8205499276411e-05,
      "loss": 0.0001,
      "step": 5722
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.8173339765235574e-05,
      "loss": 0.0127,
      "step": 5723
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.814118025406014e-05,
      "loss": 0.0111,
      "step": 5724
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.810902074288471e-05,
      "loss": 0.0028,
      "step": 5725
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.807686123170928e-05,
      "loss": 0.0315,
      "step": 5726
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.804470172053385e-05,
      "loss": 0.0016,
      "step": 5727
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.801254220935842e-05,
      "loss": 0.0241,
      "step": 5728
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.7980382698182984e-05,
      "loss": 0.0429,
      "step": 5729
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.794822318700756e-05,
      "loss": 0.0003,
      "step": 5730
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.791606367583213e-05,
      "loss": 0.0006,
      "step": 5731
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.7883904164656696e-05,
      "loss": 0.0105,
      "step": 5732
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.7851744653481266e-05,
      "loss": 0.0,
      "step": 5733
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.781958514230584e-05,
      "loss": 0.0252,
      "step": 5734
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.778742563113041e-05,
      "loss": 0.0525,
      "step": 5735
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.775526611995498e-05,
      "loss": 0.0049,
      "step": 5736
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.772310660877955e-05,
      "loss": 0.0427,
      "step": 5737
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.769094709760412e-05,
      "loss": 0.0121,
      "step": 5738
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.765878758642869e-05,
      "loss": 0.0257,
      "step": 5739
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.762662807525326e-05,
      "loss": 0.0011,
      "step": 5740
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.759446856407783e-05,
      "loss": 0.0001,
      "step": 5741
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.75623090529024e-05,
      "loss": 0.0001,
      "step": 5742
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.7530149541726964e-05,
      "loss": 0.0002,
      "step": 5743
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.749799003055154e-05,
      "loss": 0.0008,
      "step": 5744
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.7465830519376105e-05,
      "loss": 0.0,
      "step": 5745
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.7433671008200676e-05,
      "loss": 0.0004,
      "step": 5746
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.7401511497025246e-05,
      "loss": 0.0658,
      "step": 5747
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.736935198584982e-05,
      "loss": 0.0,
      "step": 5748
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.733719247467439e-05,
      "loss": 0.0152,
      "step": 5749
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.730503296349896e-05,
      "loss": 0.0023,
      "step": 5750
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.727287345232353e-05,
      "loss": 0.0117,
      "step": 5751
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.72407139411481e-05,
      "loss": 0.0226,
      "step": 5752
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.720855442997266e-05,
      "loss": 0.0003,
      "step": 5753
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.717639491879723e-05,
      "loss": 0.0017,
      "step": 5754
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.714423540762181e-05,
      "loss": 0.0293,
      "step": 5755
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.7112075896446374e-05,
      "loss": 0.0529,
      "step": 5756
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.7079916385270945e-05,
      "loss": 0.0414,
      "step": 5757
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.7047756874095515e-05,
      "loss": 0.0005,
      "step": 5758
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.7015597362920086e-05,
      "loss": 0.0695,
      "step": 5759
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.6983437851744656e-05,
      "loss": 0.066,
      "step": 5760
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.695127834056922e-05,
      "loss": 0.0004,
      "step": 5761
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.69191188293938e-05,
      "loss": 0.0547,
      "step": 5762
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.688695931821837e-05,
      "loss": 0.0156,
      "step": 5763
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.685479980704293e-05,
      "loss": 0.0219,
      "step": 5764
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.68226402958675e-05,
      "loss": 0.0008,
      "step": 5765
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.679048078469207e-05,
      "loss": 0.098,
      "step": 5766
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.675832127351664e-05,
      "loss": 0.0001,
      "step": 5767
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.6726161762341214e-05,
      "loss": 0.0001,
      "step": 5768
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.6694002251165784e-05,
      "loss": 0.0092,
      "step": 5769
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.6661842739990355e-05,
      "loss": 0.0528,
      "step": 5770
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.6629683228814925e-05,
      "loss": 0.036,
      "step": 5771
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.6597523717639496e-05,
      "loss": 0.0002,
      "step": 5772
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.6565364206464066e-05,
      "loss": 0.0,
      "step": 5773
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.653320469528863e-05,
      "loss": 0.001,
      "step": 5774
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.65010451841132e-05,
      "loss": 0.0126,
      "step": 5775
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.646888567293778e-05,
      "loss": 0.1199,
      "step": 5776
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.643672616176234e-05,
      "loss": 0.0736,
      "step": 5777
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.640456665058691e-05,
      "loss": 0.0383,
      "step": 5778
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.637240713941148e-05,
      "loss": 0.0035,
      "step": 5779
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.634024762823605e-05,
      "loss": 0.0003,
      "step": 5780
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.6308088117060624e-05,
      "loss": 0.0103,
      "step": 5781
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.627592860588519e-05,
      "loss": 0.0003,
      "step": 5782
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.6243769094709765e-05,
      "loss": 0.0136,
      "step": 5783
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.6211609583534335e-05,
      "loss": 0.0535,
      "step": 5784
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.61794500723589e-05,
      "loss": 0.0009,
      "step": 5785
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.614729056118347e-05,
      "loss": 0.0008,
      "step": 5786
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.6115131050008047e-05,
      "loss": 0.0659,
      "step": 5787
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.608297153883261e-05,
      "loss": 0.0511,
      "step": 5788
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.605081202765718e-05,
      "loss": 0.0004,
      "step": 5789
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.601865251648175e-05,
      "loss": 0.0205,
      "step": 5790
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.598649300530632e-05,
      "loss": 0.0112,
      "step": 5791
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.595433349413089e-05,
      "loss": 0.0014,
      "step": 5792
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.5922173982955456e-05,
      "loss": 0.0012,
      "step": 5793
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.5890014471780033e-05,
      "loss": 0.0227,
      "step": 5794
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.5857854960604604e-05,
      "loss": 0.0171,
      "step": 5795
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.582569544942917e-05,
      "loss": 0.0005,
      "step": 5796
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.5793535938253745e-05,
      "loss": 0.0076,
      "step": 5797
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.576137642707831e-05,
      "loss": 0.0026,
      "step": 5798
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.572921691590288e-05,
      "loss": 0.0025,
      "step": 5799
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.569705740472745e-05,
      "loss": 0.0214,
      "step": 5800
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.566489789355202e-05,
      "loss": 0.018,
      "step": 5801
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.563273838237659e-05,
      "loss": 0.0192,
      "step": 5802
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.5600578871201154e-05,
      "loss": 0.0005,
      "step": 5803
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.556841936002573e-05,
      "loss": 0.0027,
      "step": 5804
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.55362598488503e-05,
      "loss": 0.0284,
      "step": 5805
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.5504100337674866e-05,
      "loss": 0.0004,
      "step": 5806
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.5471940826499437e-05,
      "loss": 0.0157,
      "step": 5807
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.5439781315324014e-05,
      "loss": 0.0014,
      "step": 5808
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.540762180414858e-05,
      "loss": 0.0109,
      "step": 5809
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.537546229297315e-05,
      "loss": 0.0053,
      "step": 5810
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.534330278179772e-05,
      "loss": 0.035,
      "step": 5811
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.531114327062229e-05,
      "loss": 0.0004,
      "step": 5812
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.527898375944686e-05,
      "loss": 0.0003,
      "step": 5813
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.524682424827142e-05,
      "loss": 0.0381,
      "step": 5814
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.5214664737096e-05,
      "loss": 0.0002,
      "step": 5815
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.518250522592057e-05,
      "loss": 0.0184,
      "step": 5816
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.5150345714745135e-05,
      "loss": 0.0313,
      "step": 5817
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.5118186203569705e-05,
      "loss": 0.0002,
      "step": 5818
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.5086026692394276e-05,
      "loss": 0.0006,
      "step": 5819
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.5053867181218846e-05,
      "loss": 0.0007,
      "step": 5820
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.502170767004342e-05,
      "loss": 0.0777,
      "step": 5821
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.498954815886799e-05,
      "loss": 0.0158,
      "step": 5822
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.495738864769256e-05,
      "loss": 0.013,
      "step": 5823
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.492522913651713e-05,
      "loss": 0.0012,
      "step": 5824
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.489306962534169e-05,
      "loss": 0.0404,
      "step": 5825
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.486091011416627e-05,
      "loss": 0.0006,
      "step": 5826
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.482875060299083e-05,
      "loss": 0.0129,
      "step": 5827
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.4796591091815404e-05,
      "loss": 0.0006,
      "step": 5828
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.476443158063998e-05,
      "loss": 0.02,
      "step": 5829
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.4732272069464545e-05,
      "loss": 0.0005,
      "step": 5830
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.4700112558289115e-05,
      "loss": 0.0429,
      "step": 5831
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.4667953047113686e-05,
      "loss": 0.0402,
      "step": 5832
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.4635793535938256e-05,
      "loss": 0.0186,
      "step": 5833
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.460363402476283e-05,
      "loss": 0.0041,
      "step": 5834
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.457147451358739e-05,
      "loss": 0.0001,
      "step": 5835
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.453931500241197e-05,
      "loss": 0.0001,
      "step": 5836
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.450715549123654e-05,
      "loss": 0.012,
      "step": 5837
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.44749959800611e-05,
      "loss": 0.0115,
      "step": 5838
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.444283646888567e-05,
      "loss": 0.006,
      "step": 5839
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.441067695771024e-05,
      "loss": 0.0104,
      "step": 5840
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.4378517446534814e-05,
      "loss": 0.0001,
      "step": 5841
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.4346357935359384e-05,
      "loss": 0.0005,
      "step": 5842
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.4314198424183955e-05,
      "loss": 0.0503,
      "step": 5843
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.4282038913008525e-05,
      "loss": 0.0002,
      "step": 5844
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.4249879401833096e-05,
      "loss": 0.0244,
      "step": 5845
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.421771989065766e-05,
      "loss": 0.0338,
      "step": 5846
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.418556037948224e-05,
      "loss": 0.0379,
      "step": 5847
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.41534008683068e-05,
      "loss": 0.0011,
      "step": 5848
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.412124135713137e-05,
      "loss": 0.0004,
      "step": 5849
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.408908184595594e-05,
      "loss": 0.0025,
      "step": 5850
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.405692233478051e-05,
      "loss": 0.006,
      "step": 5851
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.402476282360508e-05,
      "loss": 0.0001,
      "step": 5852
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.399260331242965e-05,
      "loss": 0.0006,
      "step": 5853
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.3960443801254223e-05,
      "loss": 0.0406,
      "step": 5854
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.3928284290078794e-05,
      "loss": 0.0072,
      "step": 5855
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.389612477890336e-05,
      "loss": 0.0007,
      "step": 5856
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.3863965267727935e-05,
      "loss": 0.0017,
      "step": 5857
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.3831805756552506e-05,
      "loss": 0.0988,
      "step": 5858
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.379964624537707e-05,
      "loss": 0.0189,
      "step": 5859
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.376748673420164e-05,
      "loss": 0.0258,
      "step": 5860
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.373532722302622e-05,
      "loss": 0.0029,
      "step": 5861
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.370316771185078e-05,
      "loss": 0.0469,
      "step": 5862
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.367100820067535e-05,
      "loss": 0.0732,
      "step": 5863
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.363884868949992e-05,
      "loss": 0.0092,
      "step": 5864
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.360668917832449e-05,
      "loss": 0.0762,
      "step": 5865
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.357452966714906e-05,
      "loss": 0.0,
      "step": 5866
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.3542370155973627e-05,
      "loss": 0.0016,
      "step": 5867
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.3510210644798204e-05,
      "loss": 0.0074,
      "step": 5868
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.3478051133622774e-05,
      "loss": 0.0019,
      "step": 5869
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.344589162244734e-05,
      "loss": 0.0214,
      "step": 5870
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.341373211127191e-05,
      "loss": 0.0039,
      "step": 5871
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.338157260009648e-05,
      "loss": 0.0023,
      "step": 5872
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.334941308892105e-05,
      "loss": 0.0489,
      "step": 5873
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.331725357774562e-05,
      "loss": 0.0002,
      "step": 5874
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.328509406657019e-05,
      "loss": 0.0056,
      "step": 5875
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.325293455539476e-05,
      "loss": 0.003,
      "step": 5876
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.322077504421933e-05,
      "loss": 0.0002,
      "step": 5877
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.3188615533043895e-05,
      "loss": 0.0013,
      "step": 5878
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.315645602186847e-05,
      "loss": 0.0245,
      "step": 5879
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.3124296510693036e-05,
      "loss": 0.0001,
      "step": 5880
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.309213699951761e-05,
      "loss": 0.0435,
      "step": 5881
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.305997748834218e-05,
      "loss": 0.0221,
      "step": 5882
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.302781797716675e-05,
      "loss": 0.0308,
      "step": 5883
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.299565846599132e-05,
      "loss": 0.0608,
      "step": 5884
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.296349895481589e-05,
      "loss": 0.0042,
      "step": 5885
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.293133944364046e-05,
      "loss": 0.0009,
      "step": 5886
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.289917993246503e-05,
      "loss": 0.0309,
      "step": 5887
    },
    {
      "epoch": 8.51,
      "learning_rate": 3.2867020421289594e-05,
      "loss": 0.0012,
      "step": 5888
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.283486091011417e-05,
      "loss": 0.0364,
      "step": 5889
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.280270139893874e-05,
      "loss": 0.014,
      "step": 5890
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.2770541887763305e-05,
      "loss": 0.0365,
      "step": 5891
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.2738382376587876e-05,
      "loss": 0.0008,
      "step": 5892
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.2706222865412446e-05,
      "loss": 0.001,
      "step": 5893
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.267406335423702e-05,
      "loss": 0.0021,
      "step": 5894
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.264190384306159e-05,
      "loss": 0.0474,
      "step": 5895
    },
    {
      "epoch": 8.53,
      "learning_rate": 3.260974433188616e-05,
      "loss": 0.0502,
      "step": 5896
    },
    {
      "epoch": 8.53,
      "learning_rate": 3.257758482071073e-05,
      "loss": 0.0005,
      "step": 5897
    },
    {
      "epoch": 8.53,
      "learning_rate": 3.25454253095353e-05,
      "loss": 0.0013,
      "step": 5898
    },
    {
      "epoch": 8.53,
      "learning_rate": 3.251326579835986e-05,
      "loss": 0.025,
      "step": 5899
    },
    {
      "epoch": 8.53,
      "learning_rate": 3.248110628718444e-05,
      "loss": 0.0156,
      "step": 5900
    },
    {
      "epoch": 8.53,
      "learning_rate": 3.2448946776009004e-05,
      "loss": 0.0,
      "step": 5901
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.2416787264833574e-05,
      "loss": 0.0142,
      "step": 5902
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.2384627753658145e-05,
      "loss": 0.0123,
      "step": 5903
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.2352468242482715e-05,
      "loss": 0.0023,
      "step": 5904
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.2320308731307286e-05,
      "loss": 0.0028,
      "step": 5905
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.2288149220131856e-05,
      "loss": 0.0446,
      "step": 5906
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.225598970895643e-05,
      "loss": 0.0011,
      "step": 5907
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.2223830197781e-05,
      "loss": 0.0797,
      "step": 5908
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.219167068660556e-05,
      "loss": 0.041,
      "step": 5909
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.215951117543013e-05,
      "loss": 0.0101,
      "step": 5910
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.212735166425471e-05,
      "loss": 0.0572,
      "step": 5911
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.209519215307927e-05,
      "loss": 0.0084,
      "step": 5912
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.206303264190384e-05,
      "loss": 0.0415,
      "step": 5913
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.203087313072842e-05,
      "loss": 0.0002,
      "step": 5914
    },
    {
      "epoch": 8.55,
      "learning_rate": 3.1998713619552984e-05,
      "loss": 0.0014,
      "step": 5915
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.1966554108377555e-05,
      "loss": 0.0473,
      "step": 5916
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.1934394597202125e-05,
      "loss": 0.001,
      "step": 5917
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.1902235086026696e-05,
      "loss": 0.007,
      "step": 5918
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.1870075574851266e-05,
      "loss": 0.0648,
      "step": 5919
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.183791606367583e-05,
      "loss": 0.0002,
      "step": 5920
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.180575655250041e-05,
      "loss": 0.0,
      "step": 5921
    },
    {
      "epoch": 8.56,
      "learning_rate": 3.177359704132497e-05,
      "loss": 0.0137,
      "step": 5922
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.174143753014954e-05,
      "loss": 0.0374,
      "step": 5923
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.170927801897411e-05,
      "loss": 0.0369,
      "step": 5924
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.167711850779868e-05,
      "loss": 0.0012,
      "step": 5925
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.164495899662325e-05,
      "loss": 0.0001,
      "step": 5926
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.1612799485447823e-05,
      "loss": 0.0,
      "step": 5927
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.1580639974272394e-05,
      "loss": 0.0073,
      "step": 5928
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.1548480463096964e-05,
      "loss": 0.0381,
      "step": 5929
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.151632095192153e-05,
      "loss": 0.0,
      "step": 5930
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.14841614407461e-05,
      "loss": 0.0364,
      "step": 5931
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.1452001929570676e-05,
      "loss": 0.0,
      "step": 5932
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.141984241839524e-05,
      "loss": 0.0298,
      "step": 5933
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.138768290721981e-05,
      "loss": 0.005,
      "step": 5934
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.135552339604438e-05,
      "loss": 0.0699,
      "step": 5935
    },
    {
      "epoch": 8.58,
      "learning_rate": 3.132336388486895e-05,
      "loss": 0.0803,
      "step": 5936
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.129120437369352e-05,
      "loss": 0.0003,
      "step": 5937
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.1259044862518086e-05,
      "loss": 0.0002,
      "step": 5938
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.122688535134266e-05,
      "loss": 0.0591,
      "step": 5939
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.119472584016723e-05,
      "loss": 0.0064,
      "step": 5940
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.11625663289918e-05,
      "loss": 0.0705,
      "step": 5941
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.1130406817816374e-05,
      "loss": 0.0036,
      "step": 5942
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.1098247306640945e-05,
      "loss": 0.0003,
      "step": 5943
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.106608779546551e-05,
      "loss": 0.0001,
      "step": 5944
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.103392828429008e-05,
      "loss": 0.0353,
      "step": 5945
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.100176877311465e-05,
      "loss": 0.0,
      "step": 5946
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.096960926193922e-05,
      "loss": 0.0008,
      "step": 5947
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.093744975076379e-05,
      "loss": 0.0369,
      "step": 5948
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.090529023958836e-05,
      "loss": 0.0003,
      "step": 5949
    },
    {
      "epoch": 8.6,
      "learning_rate": 3.087313072841293e-05,
      "loss": 0.0468,
      "step": 5950
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.08409712172375e-05,
      "loss": 0.0063,
      "step": 5951
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.0808811706062066e-05,
      "loss": 0.0005,
      "step": 5952
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.077665219488664e-05,
      "loss": 0.0087,
      "step": 5953
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.074449268371121e-05,
      "loss": 0.0462,
      "step": 5954
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.071233317253578e-05,
      "loss": 0.0,
      "step": 5955
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.068017366136035e-05,
      "loss": 0.0053,
      "step": 5956
    },
    {
      "epoch": 8.61,
      "learning_rate": 3.064801415018492e-05,
      "loss": 0.0044,
      "step": 5957
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.061585463900949e-05,
      "loss": 0.0137,
      "step": 5958
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.058369512783406e-05,
      "loss": 0.0001,
      "step": 5959
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.055153561665863e-05,
      "loss": 0.0,
      "step": 5960
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.05193761054832e-05,
      "loss": 0.0119,
      "step": 5961
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.0487216594307764e-05,
      "loss": 0.0003,
      "step": 5962
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.0455057083132338e-05,
      "loss": 0.0695,
      "step": 5963
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.042289757195691e-05,
      "loss": 0.0168,
      "step": 5964
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.0390738060781476e-05,
      "loss": 0.0315,
      "step": 5965
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.035857854960605e-05,
      "loss": 0.0109,
      "step": 5966
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.0326419038430613e-05,
      "loss": 0.0039,
      "step": 5967
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.0294259527255187e-05,
      "loss": 0.0036,
      "step": 5968
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.0262100016079758e-05,
      "loss": 0.0064,
      "step": 5969
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.0229940504904325e-05,
      "loss": 0.0118,
      "step": 5970
    },
    {
      "epoch": 8.63,
      "learning_rate": 3.0197780993728895e-05,
      "loss": 0.0031,
      "step": 5971
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.016562148255347e-05,
      "loss": 0.0194,
      "step": 5972
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.0133461971378037e-05,
      "loss": 0.0245,
      "step": 5973
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.0101302460202607e-05,
      "loss": 0.0224,
      "step": 5974
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.0069142949027174e-05,
      "loss": 0.0379,
      "step": 5975
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.0036983437851745e-05,
      "loss": 0.0001,
      "step": 5976
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.000482392667632e-05,
      "loss": 0.0824,
      "step": 5977
    },
    {
      "epoch": 8.64,
      "learning_rate": 2.9972664415500882e-05,
      "loss": 0.0198,
      "step": 5978
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.9940504904325456e-05,
      "loss": 0.0002,
      "step": 5979
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.9908345393150027e-05,
      "loss": 0.0593,
      "step": 5980
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.9876185881974594e-05,
      "loss": 0.0021,
      "step": 5981
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.9844026370799168e-05,
      "loss": 0.0396,
      "step": 5982
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.981186685962373e-05,
      "loss": 0.0016,
      "step": 5983
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.9779707348448305e-05,
      "loss": 0.0299,
      "step": 5984
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9747547837272876e-05,
      "loss": 0.03,
      "step": 5985
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9715388326097443e-05,
      "loss": 0.0041,
      "step": 5986
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9683228814922014e-05,
      "loss": 0.0,
      "step": 5987
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9651069303746587e-05,
      "loss": 0.0004,
      "step": 5988
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9618909792571155e-05,
      "loss": 0.0234,
      "step": 5989
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9586750281395725e-05,
      "loss": 0.0319,
      "step": 5990
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.9554590770220292e-05,
      "loss": 0.0259,
      "step": 5991
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9522431259044863e-05,
      "loss": 0.0487,
      "step": 5992
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9490271747869437e-05,
      "loss": 0.0689,
      "step": 5993
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9458112236694004e-05,
      "loss": 0.0013,
      "step": 5994
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9425952725518574e-05,
      "loss": 0.0007,
      "step": 5995
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9393793214343145e-05,
      "loss": 0.0064,
      "step": 5996
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9361633703167712e-05,
      "loss": 0.0007,
      "step": 5997
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.9329474191992286e-05,
      "loss": 0.045,
      "step": 5998
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.929731468081685e-05,
      "loss": 0.0006,
      "step": 5999
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.9265155169641423e-05,
      "loss": 0.0165,
      "step": 6000
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.9232995658465994e-05,
      "loss": 0.0524,
      "step": 6001
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.920083614729056e-05,
      "loss": 0.0043,
      "step": 6002
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.916867663611513e-05,
      "loss": 0.0101,
      "step": 6003
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.91365171249397e-05,
      "loss": 0.032,
      "step": 6004
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.9104357613764273e-05,
      "loss": 0.003,
      "step": 6005
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.9072198102588843e-05,
      "loss": 0.0073,
      "step": 6006
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.904003859141341e-05,
      "loss": 0.0313,
      "step": 6007
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.900787908023798e-05,
      "loss": 0.0005,
      "step": 6008
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.8975719569062555e-05,
      "loss": 0.0194,
      "step": 6009
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.8943560057887122e-05,
      "loss": 0.0,
      "step": 6010
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.8911400546711692e-05,
      "loss": 0.0532,
      "step": 6011
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.887924103553626e-05,
      "loss": 0.0001,
      "step": 6012
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.884708152436083e-05,
      "loss": 0.0003,
      "step": 6013
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.8814922013185404e-05,
      "loss": 0.0482,
      "step": 6014
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.8782762502009968e-05,
      "loss": 0.0002,
      "step": 6015
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.875060299083454e-05,
      "loss": 0.002,
      "step": 6016
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.8718443479659112e-05,
      "loss": 0.0243,
      "step": 6017
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.868628396848368e-05,
      "loss": 0.0002,
      "step": 6018
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.865412445730825e-05,
      "loss": 0.0045,
      "step": 6019
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.8621964946132817e-05,
      "loss": 0.0003,
      "step": 6020
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.858980543495739e-05,
      "loss": 0.0073,
      "step": 6021
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.855764592378196e-05,
      "loss": 0.002,
      "step": 6022
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.8525486412606528e-05,
      "loss": 0.005,
      "step": 6023
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.84933269014311e-05,
      "loss": 0.0154,
      "step": 6024
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.8461167390255673e-05,
      "loss": 0.0388,
      "step": 6025
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.842900787908024e-05,
      "loss": 0.0448,
      "step": 6026
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.839684836790481e-05,
      "loss": 0.0356,
      "step": 6027
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.8364688856729377e-05,
      "loss": 0.0113,
      "step": 6028
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.8332529345553948e-05,
      "loss": 0.0461,
      "step": 6029
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.8300369834378522e-05,
      "loss": 0.0332,
      "step": 6030
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.8268210323203086e-05,
      "loss": 0.0171,
      "step": 6031
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.823605081202766e-05,
      "loss": 0.0049,
      "step": 6032
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.820389130085223e-05,
      "loss": 0.0,
      "step": 6033
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.8171731789676797e-05,
      "loss": 0.0186,
      "step": 6034
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.813957227850137e-05,
      "loss": 0.0231,
      "step": 6035
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.8107412767325935e-05,
      "loss": 0.0508,
      "step": 6036
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.807525325615051e-05,
      "loss": 0.0541,
      "step": 6037
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.804309374497508e-05,
      "loss": 0.0475,
      "step": 6038
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.8010934233799646e-05,
      "loss": 0.0043,
      "step": 6039
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.7978774722624217e-05,
      "loss": 0.0308,
      "step": 6040
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.794661521144879e-05,
      "loss": 0.0039,
      "step": 6041
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.7914455700273358e-05,
      "loss": 0.0476,
      "step": 6042
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.788229618909793e-05,
      "loss": 0.0001,
      "step": 6043
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.7850136677922495e-05,
      "loss": 0.0,
      "step": 6044
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.7817977166747066e-05,
      "loss": 0.0,
      "step": 6045
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.778581765557164e-05,
      "loss": 0.0317,
      "step": 6046
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.7753658144396204e-05,
      "loss": 0.0364,
      "step": 6047
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.7721498633220778e-05,
      "loss": 0.0383,
      "step": 6048
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.7689339122045345e-05,
      "loss": 0.0739,
      "step": 6049
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.7657179610869915e-05,
      "loss": 0.0,
      "step": 6050
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.762502009969449e-05,
      "loss": 0.0188,
      "step": 6051
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.7592860588519053e-05,
      "loss": 0.0397,
      "step": 6052
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.7560701077343627e-05,
      "loss": 0.0013,
      "step": 6053
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.7528541566168197e-05,
      "loss": 0.0,
      "step": 6054
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.7496382054992764e-05,
      "loss": 0.0273,
      "step": 6055
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.7464222543817335e-05,
      "loss": 0.0471,
      "step": 6056
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.7432063032641902e-05,
      "loss": 0.0026,
      "step": 6057
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.7399903521466476e-05,
      "loss": 0.0021,
      "step": 6058
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.7367744010291046e-05,
      "loss": 0.0619,
      "step": 6059
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.7335584499115613e-05,
      "loss": 0.0103,
      "step": 6060
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.7303424987940184e-05,
      "loss": 0.0536,
      "step": 6061
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.7271265476764758e-05,
      "loss": 0.0396,
      "step": 6062
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.723910596558932e-05,
      "loss": 0.0206,
      "step": 6063
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.7206946454413896e-05,
      "loss": 0.0009,
      "step": 6064
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.7174786943238463e-05,
      "loss": 0.0366,
      "step": 6065
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.7142627432063033e-05,
      "loss": 0.0006,
      "step": 6066
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.7110467920887607e-05,
      "loss": 0.0079,
      "step": 6067
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.707830840971217e-05,
      "loss": 0.0814,
      "step": 6068
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.7046148898536745e-05,
      "loss": 0.0268,
      "step": 6069
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.7013989387361315e-05,
      "loss": 0.0001,
      "step": 6070
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.6981829876185882e-05,
      "loss": 0.0,
      "step": 6071
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.6949670365010453e-05,
      "loss": 0.006,
      "step": 6072
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.691751085383502e-05,
      "loss": 0.0125,
      "step": 6073
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.6885351342659594e-05,
      "loss": 0.005,
      "step": 6074
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.6853191831484164e-05,
      "loss": 0.0019,
      "step": 6075
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.682103232030873e-05,
      "loss": 0.0074,
      "step": 6076
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.6788872809133302e-05,
      "loss": 0.0001,
      "step": 6077
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.6756713297957876e-05,
      "loss": 0.0045,
      "step": 6078
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.6724553786782443e-05,
      "loss": 0.0384,
      "step": 6079
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.6692394275607014e-05,
      "loss": 0.0014,
      "step": 6080
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.666023476443158e-05,
      "loss": 0.0405,
      "step": 6081
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.662807525325615e-05,
      "loss": 0.0085,
      "step": 6082
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.6595915742080725e-05,
      "loss": 0.0972,
      "step": 6083
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.656375623090529e-05,
      "loss": 0.0235,
      "step": 6084
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.6531596719729863e-05,
      "loss": 0.0027,
      "step": 6085
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.6499437208554433e-05,
      "loss": 0.0708,
      "step": 6086
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.6467277697379e-05,
      "loss": 0.0147,
      "step": 6087
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.643511818620357e-05,
      "loss": 0.0005,
      "step": 6088
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.6402958675028138e-05,
      "loss": 0.0003,
      "step": 6089
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.6370799163852712e-05,
      "loss": 0.0262,
      "step": 6090
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.6338639652677282e-05,
      "loss": 0.0097,
      "step": 6091
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.630648014150185e-05,
      "loss": 0.0,
      "step": 6092
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.627432063032642e-05,
      "loss": 0.0515,
      "step": 6093
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.6242161119150987e-05,
      "loss": 0.0061,
      "step": 6094
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.621000160797556e-05,
      "loss": 0.0133,
      "step": 6095
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.617784209680013e-05,
      "loss": 0.0349,
      "step": 6096
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.61456825856247e-05,
      "loss": 0.0095,
      "step": 6097
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.611352307444927e-05,
      "loss": 0.0008,
      "step": 6098
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.6081363563273843e-05,
      "loss": 0.0211,
      "step": 6099
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.6049204052098407e-05,
      "loss": 0.0036,
      "step": 6100
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.601704454092298e-05,
      "loss": 0.046,
      "step": 6101
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.5984885029747548e-05,
      "loss": 0.0044,
      "step": 6102
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.595272551857212e-05,
      "loss": 0.0332,
      "step": 6103
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.592056600739669e-05,
      "loss": 0.002,
      "step": 6104
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.5888406496221256e-05,
      "loss": 0.0203,
      "step": 6105
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.585624698504583e-05,
      "loss": 0.0002,
      "step": 6106
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.58240874738704e-05,
      "loss": 0.0223,
      "step": 6107
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.5791927962694968e-05,
      "loss": 0.0294,
      "step": 6108
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.5759768451519538e-05,
      "loss": 0.0471,
      "step": 6109
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.5727608940344105e-05,
      "loss": 0.0005,
      "step": 6110
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.569544942916868e-05,
      "loss": 0.0004,
      "step": 6111
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.566328991799325e-05,
      "loss": 0.0083,
      "step": 6112
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.5631130406817817e-05,
      "loss": 0.0606,
      "step": 6113
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.5598970895642387e-05,
      "loss": 0.0006,
      "step": 6114
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.556681138446696e-05,
      "loss": 0.0203,
      "step": 6115
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.5534651873291525e-05,
      "loss": 0.0133,
      "step": 6116
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.55024923621161e-05,
      "loss": 0.0002,
      "step": 6117
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.5470332850940666e-05,
      "loss": 0.0248,
      "step": 6118
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.5438173339765236e-05,
      "loss": 0.0797,
      "step": 6119
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.5406013828589807e-05,
      "loss": 0.0,
      "step": 6120
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.5373854317414374e-05,
      "loss": 0.0295,
      "step": 6121
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.5341694806238948e-05,
      "loss": 0.0503,
      "step": 6122
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.530953529506352e-05,
      "loss": 0.0686,
      "step": 6123
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5277375783888086e-05,
      "loss": 0.0083,
      "step": 6124
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5245216272712656e-05,
      "loss": 0.0001,
      "step": 6125
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5213056761537223e-05,
      "loss": 0.0092,
      "step": 6126
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5180897250361797e-05,
      "loss": 0.0203,
      "step": 6127
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5148737739186368e-05,
      "loss": 0.0049,
      "step": 6128
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5116578228010935e-05,
      "loss": 0.0016,
      "step": 6129
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.5084418716835505e-05,
      "loss": 0.0191,
      "step": 6130
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.5052259205660072e-05,
      "loss": 0.0104,
      "step": 6131
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.5020099694484643e-05,
      "loss": 0.0004,
      "step": 6132
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.4987940183309213e-05,
      "loss": 0.0043,
      "step": 6133
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.4955780672133784e-05,
      "loss": 0.059,
      "step": 6134
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.4923621160958354e-05,
      "loss": 0.0455,
      "step": 6135
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.4891461649782925e-05,
      "loss": 0.0032,
      "step": 6136
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.4859302138607492e-05,
      "loss": 0.0343,
      "step": 6137
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.4827142627432066e-05,
      "loss": 0.035,
      "step": 6138
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.4794983116256633e-05,
      "loss": 0.0,
      "step": 6139
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.4762823605081204e-05,
      "loss": 0.0366,
      "step": 6140
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.4730664093905774e-05,
      "loss": 0.048,
      "step": 6141
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.4698504582730345e-05,
      "loss": 0.0424,
      "step": 6142
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.4666345071554915e-05,
      "loss": 0.0177,
      "step": 6143
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.4634185560379482e-05,
      "loss": 0.0,
      "step": 6144
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.4602026049204053e-05,
      "loss": 0.0111,
      "step": 6145
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.4569866538028623e-05,
      "loss": 0.0462,
      "step": 6146
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.4537707026853194e-05,
      "loss": 0.0008,
      "step": 6147
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.450554751567776e-05,
      "loss": 0.0593,
      "step": 6148
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.447338800450233e-05,
      "loss": 0.0002,
      "step": 6149
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.4441228493326902e-05,
      "loss": 0.0077,
      "step": 6150
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.4409068982151473e-05,
      "loss": 0.025,
      "step": 6151
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.4376909470976043e-05,
      "loss": 0.0131,
      "step": 6152
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.434474995980061e-05,
      "loss": 0.0,
      "step": 6153
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.4312590448625184e-05,
      "loss": 0.0302,
      "step": 6154
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.428043093744975e-05,
      "loss": 0.0001,
      "step": 6155
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.424827142627432e-05,
      "loss": 0.0019,
      "step": 6156
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.4216111915098892e-05,
      "loss": 0.0,
      "step": 6157
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.4183952403923463e-05,
      "loss": 0.0176,
      "step": 6158
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.4151792892748033e-05,
      "loss": 0.0008,
      "step": 6159
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.41196333815726e-05,
      "loss": 0.0579,
      "step": 6160
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.408747387039717e-05,
      "loss": 0.0162,
      "step": 6161
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.405531435922174e-05,
      "loss": 0.0325,
      "step": 6162
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.4023154848046312e-05,
      "loss": 0.0003,
      "step": 6163
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.399099533687088e-05,
      "loss": 0.0002,
      "step": 6164
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.395883582569545e-05,
      "loss": 0.012,
      "step": 6165
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.3926676314520023e-05,
      "loss": 0.0016,
      "step": 6166
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.389451680334459e-05,
      "loss": 0.0011,
      "step": 6167
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.386235729216916e-05,
      "loss": 0.0006,
      "step": 6168
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.3830197780993728e-05,
      "loss": 0.0427,
      "step": 6169
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.37980382698183e-05,
      "loss": 0.0006,
      "step": 6170
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.376587875864287e-05,
      "loss": 0.0023,
      "step": 6171
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.373371924746744e-05,
      "loss": 0.0006,
      "step": 6172
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.370155973629201e-05,
      "loss": 0.0001,
      "step": 6173
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.3669400225116577e-05,
      "loss": 0.0,
      "step": 6174
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.363724071394115e-05,
      "loss": 0.0009,
      "step": 6175
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.360508120276572e-05,
      "loss": 0.0186,
      "step": 6176
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.357292169159029e-05,
      "loss": 0.0713,
      "step": 6177
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.3540762180414856e-05,
      "loss": 0.0107,
      "step": 6178
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.350860266923943e-05,
      "loss": 0.0904,
      "step": 6179
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.3476443158064e-05,
      "loss": 0.0014,
      "step": 6180
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.3444283646888568e-05,
      "loss": 0.0012,
      "step": 6181
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.3412124135713138e-05,
      "loss": 0.0259,
      "step": 6182
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.337996462453771e-05,
      "loss": 0.0084,
      "step": 6183
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.334780511336228e-05,
      "loss": 0.0026,
      "step": 6184
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.3315645602186846e-05,
      "loss": 0.0466,
      "step": 6185
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.3283486091011417e-05,
      "loss": 0.0049,
      "step": 6186
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.3251326579835987e-05,
      "loss": 0.0002,
      "step": 6187
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.3219167068660558e-05,
      "loss": 0.0392,
      "step": 6188
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.3187007557485128e-05,
      "loss": 0.0,
      "step": 6189
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.3154848046309695e-05,
      "loss": 0.0011,
      "step": 6190
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.312268853513427e-05,
      "loss": 0.0272,
      "step": 6191
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.3090529023958836e-05,
      "loss": 0.0405,
      "step": 6192
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.3058369512783407e-05,
      "loss": 0.0387,
      "step": 6193
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.3026210001607974e-05,
      "loss": 0.0535,
      "step": 6194
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.2994050490432548e-05,
      "loss": 0.0076,
      "step": 6195
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.296189097925712e-05,
      "loss": 0.0001,
      "step": 6196
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.2929731468081686e-05,
      "loss": 0.0045,
      "step": 6197
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.2897571956906256e-05,
      "loss": 0.0141,
      "step": 6198
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.2865412445730827e-05,
      "loss": 0.0037,
      "step": 6199
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2833252934555397e-05,
      "loss": 0.0171,
      "step": 6200
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2801093423379964e-05,
      "loss": 0.0,
      "step": 6201
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2768933912204535e-05,
      "loss": 0.1466,
      "step": 6202
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2736774401029105e-05,
      "loss": 0.0148,
      "step": 6203
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2704614889853676e-05,
      "loss": 0.0095,
      "step": 6204
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2672455378678246e-05,
      "loss": 0.0999,
      "step": 6205
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.2640295867502813e-05,
      "loss": 0.0362,
      "step": 6206
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2608136356327387e-05,
      "loss": 0.0199,
      "step": 6207
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2575976845151954e-05,
      "loss": 0.0002,
      "step": 6208
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2543817333976525e-05,
      "loss": 0.0003,
      "step": 6209
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2511657822801095e-05,
      "loss": 0.0072,
      "step": 6210
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2479498311625663e-05,
      "loss": 0.0147,
      "step": 6211
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2447338800450236e-05,
      "loss": 0.0006,
      "step": 6212
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.2415179289274804e-05,
      "loss": 0.0206,
      "step": 6213
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.2383019778099374e-05,
      "loss": 0.0253,
      "step": 6214
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.235086026692394e-05,
      "loss": 0.0226,
      "step": 6215
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.2318700755748515e-05,
      "loss": 0.0713,
      "step": 6216
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.2286541244573082e-05,
      "loss": 0.0011,
      "step": 6217
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.2254381733397653e-05,
      "loss": 0.001,
      "step": 6218
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.2222222222222223e-05,
      "loss": 0.0002,
      "step": 6219
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.2190062711046794e-05,
      "loss": 0.0046,
      "step": 6220
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.2157903199871364e-05,
      "loss": 0.0684,
      "step": 6221
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.212574368869593e-05,
      "loss": 0.0601,
      "step": 6222
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.2093584177520502e-05,
      "loss": 0.0049,
      "step": 6223
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.2061424666345072e-05,
      "loss": 0.04,
      "step": 6224
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.2029265155169643e-05,
      "loss": 0.001,
      "step": 6225
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.1997105643994213e-05,
      "loss": 0.0211,
      "step": 6226
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.196494613281878e-05,
      "loss": 0.0142,
      "step": 6227
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.1932786621643355e-05,
      "loss": 0.0021,
      "step": 6228
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.190062711046792e-05,
      "loss": 0.0311,
      "step": 6229
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.1868467599292492e-05,
      "loss": 0.0389,
      "step": 6230
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.183630808811706e-05,
      "loss": 0.0071,
      "step": 6231
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.1804148576941633e-05,
      "loss": 0.0059,
      "step": 6232
    },
    {
      "epoch": 9.01,
      "learning_rate": 2.17719890657662e-05,
      "loss": 0.0093,
      "step": 6233
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.173982955459077e-05,
      "loss": 0.0006,
      "step": 6234
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.170767004341534e-05,
      "loss": 0.0,
      "step": 6235
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.1675510532239912e-05,
      "loss": 0.0008,
      "step": 6236
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.1643351021064482e-05,
      "loss": 0.0366,
      "step": 6237
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.161119150988905e-05,
      "loss": 0.0,
      "step": 6238
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.157903199871362e-05,
      "loss": 0.0,
      "step": 6239
    },
    {
      "epoch": 9.02,
      "learning_rate": 2.154687248753819e-05,
      "loss": 0.0013,
      "step": 6240
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.151471297636276e-05,
      "loss": 0.0034,
      "step": 6241
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.148255346518733e-05,
      "loss": 0.0004,
      "step": 6242
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.14503939540119e-05,
      "loss": 0.04,
      "step": 6243
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.1418234442836473e-05,
      "loss": 0.0197,
      "step": 6244
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.138607493166104e-05,
      "loss": 0.038,
      "step": 6245
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.135391542048561e-05,
      "loss": 0.0001,
      "step": 6246
    },
    {
      "epoch": 9.03,
      "learning_rate": 2.1321755909310177e-05,
      "loss": 0.0072,
      "step": 6247
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.128959639813475e-05,
      "loss": 0.0021,
      "step": 6248
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.125743688695932e-05,
      "loss": 0.001,
      "step": 6249
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.122527737578389e-05,
      "loss": 0.0032,
      "step": 6250
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.119311786460846e-05,
      "loss": 0.1011,
      "step": 6251
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.1160958353433026e-05,
      "loss": 0.015,
      "step": 6252
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.11287988422576e-05,
      "loss": 0.0004,
      "step": 6253
    },
    {
      "epoch": 9.04,
      "learning_rate": 2.1096639331082167e-05,
      "loss": 0.0099,
      "step": 6254
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.1064479819906738e-05,
      "loss": 0.0297,
      "step": 6255
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.103232030873131e-05,
      "loss": 0.0217,
      "step": 6256
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.100016079755588e-05,
      "loss": 0.0002,
      "step": 6257
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.096800128638045e-05,
      "loss": 0.0108,
      "step": 6258
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.0935841775205017e-05,
      "loss": 0.0365,
      "step": 6259
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.0903682264029587e-05,
      "loss": 0.063,
      "step": 6260
    },
    {
      "epoch": 9.05,
      "learning_rate": 2.0871522752854158e-05,
      "loss": 0.0002,
      "step": 6261
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0839363241678728e-05,
      "loss": 0.0001,
      "step": 6262
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0807203730503295e-05,
      "loss": 0.0066,
      "step": 6263
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0775044219327866e-05,
      "loss": 0.0134,
      "step": 6264
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0742884708152436e-05,
      "loss": 0.0304,
      "step": 6265
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0710725196977007e-05,
      "loss": 0.0143,
      "step": 6266
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0678565685801577e-05,
      "loss": 0.0379,
      "step": 6267
    },
    {
      "epoch": 9.06,
      "learning_rate": 2.0646406174626145e-05,
      "loss": 0.0732,
      "step": 6268
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.061424666345072e-05,
      "loss": 0.0361,
      "step": 6269
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.0582087152275286e-05,
      "loss": 0.0332,
      "step": 6270
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.0549927641099856e-05,
      "loss": 0.0,
      "step": 6271
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.0517768129924427e-05,
      "loss": 0.0033,
      "step": 6272
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.0485608618748997e-05,
      "loss": 0.0464,
      "step": 6273
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.0453449107573568e-05,
      "loss": 0.0,
      "step": 6274
    },
    {
      "epoch": 9.07,
      "learning_rate": 2.0421289596398135e-05,
      "loss": 0.0354,
      "step": 6275
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0389130085222705e-05,
      "loss": 0.0251,
      "step": 6276
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0356970574047276e-05,
      "loss": 0.0047,
      "step": 6277
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0324811062871846e-05,
      "loss": 0.0004,
      "step": 6278
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0292651551696413e-05,
      "loss": 0.0369,
      "step": 6279
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0260492040520984e-05,
      "loss": 0.0096,
      "step": 6280
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0228332529345558e-05,
      "loss": 0.0004,
      "step": 6281
    },
    {
      "epoch": 9.08,
      "learning_rate": 2.0196173018170125e-05,
      "loss": 0.0408,
      "step": 6282
    },
    {
      "epoch": 9.09,
      "learning_rate": 2.0164013506994695e-05,
      "loss": 0.0025,
      "step": 6283
    },
    {
      "epoch": 9.09,
      "learning_rate": 2.0131853995819263e-05,
      "loss": 0.0012,
      "step": 6284
    },
    {
      "epoch": 9.09,
      "learning_rate": 2.0099694484643836e-05,
      "loss": 0.0003,
      "step": 6285
    },
    {
      "epoch": 9.09,
      "learning_rate": 2.0067534973468404e-05,
      "loss": 0.0188,
      "step": 6286
    },
    {
      "epoch": 9.09,
      "learning_rate": 2.0035375462292974e-05,
      "loss": 0.0226,
      "step": 6287
    },
    {
      "epoch": 9.09,
      "learning_rate": 2.0003215951117545e-05,
      "loss": 0.0103,
      "step": 6288
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.9971056439942115e-05,
      "loss": 0.0289,
      "step": 6289
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.9938896928766686e-05,
      "loss": 0.049,
      "step": 6290
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.9906737417591253e-05,
      "loss": 0.0421,
      "step": 6291
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.9874577906415823e-05,
      "loss": 0.008,
      "step": 6292
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.984241839524039e-05,
      "loss": 0.0002,
      "step": 6293
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.9810258884064964e-05,
      "loss": 0.0337,
      "step": 6294
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.977809937288953e-05,
      "loss": 0.0,
      "step": 6295
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.9745939861714102e-05,
      "loss": 0.0028,
      "step": 6296
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.9713780350538672e-05,
      "loss": 0.0144,
      "step": 6297
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.9681620839363243e-05,
      "loss": 0.0041,
      "step": 6298
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.9649461328187813e-05,
      "loss": 0.0141,
      "step": 6299
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.961730181701238e-05,
      "loss": 0.0002,
      "step": 6300
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.958514230583695e-05,
      "loss": 0.0239,
      "step": 6301
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.955298279466152e-05,
      "loss": 0.0011,
      "step": 6302
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.9520823283486092e-05,
      "loss": 0.0019,
      "step": 6303
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.9488663772310663e-05,
      "loss": 0.0001,
      "step": 6304
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.945650426113523e-05,
      "loss": 0.0205,
      "step": 6305
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.9424344749959804e-05,
      "loss": 0.0069,
      "step": 6306
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.939218523878437e-05,
      "loss": 0.0003,
      "step": 6307
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.936002572760894e-05,
      "loss": 0.0442,
      "step": 6308
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.932786621643351e-05,
      "loss": 0.0479,
      "step": 6309
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.9295706705258082e-05,
      "loss": 0.0707,
      "step": 6310
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.9263547194082653e-05,
      "loss": 0.0001,
      "step": 6311
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.923138768290722e-05,
      "loss": 0.0111,
      "step": 6312
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.919922817173179e-05,
      "loss": 0.0424,
      "step": 6313
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.916706866055636e-05,
      "loss": 0.0005,
      "step": 6314
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.913490914938093e-05,
      "loss": 0.0017,
      "step": 6315
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.91027496382055e-05,
      "loss": 0.0004,
      "step": 6316
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.907059012703007e-05,
      "loss": 0.0004,
      "step": 6317
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.903843061585464e-05,
      "loss": 0.0042,
      "step": 6318
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.900627110467921e-05,
      "loss": 0.0071,
      "step": 6319
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.897411159350378e-05,
      "loss": 0.0312,
      "step": 6320
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.8941952082328348e-05,
      "loss": 0.0006,
      "step": 6321
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.890979257115292e-05,
      "loss": 0.0004,
      "step": 6322
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.887763305997749e-05,
      "loss": 0.0623,
      "step": 6323
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.884547354880206e-05,
      "loss": 0.004,
      "step": 6324
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.881331403762663e-05,
      "loss": 0.0001,
      "step": 6325
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.87811545264512e-05,
      "loss": 0.0,
      "step": 6326
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.874899501527577e-05,
      "loss": 0.0069,
      "step": 6327
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.8716835504100338e-05,
      "loss": 0.0002,
      "step": 6328
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.868467599292491e-05,
      "loss": 0.0049,
      "step": 6329
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.865251648174948e-05,
      "loss": 0.0008,
      "step": 6330
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.862035697057405e-05,
      "loss": 0.088,
      "step": 6331
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8588197459398617e-05,
      "loss": 0.0225,
      "step": 6332
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8556037948223187e-05,
      "loss": 0.0278,
      "step": 6333
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8523878437047758e-05,
      "loss": 0.0383,
      "step": 6334
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8491718925872328e-05,
      "loss": 0.0054,
      "step": 6335
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.84595594146969e-05,
      "loss": 0.0025,
      "step": 6336
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8427399903521466e-05,
      "loss": 0.0007,
      "step": 6337
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8395240392346036e-05,
      "loss": 0.0006,
      "step": 6338
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8363080881170607e-05,
      "loss": 0.0029,
      "step": 6339
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8330921369995177e-05,
      "loss": 0.0369,
      "step": 6340
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8298761858819748e-05,
      "loss": 0.0687,
      "step": 6341
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8266602347644315e-05,
      "loss": 0.0,
      "step": 6342
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.823444283646889e-05,
      "loss": 0.0,
      "step": 6343
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8202283325293456e-05,
      "loss": 0.0001,
      "step": 6344
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.8170123814118027e-05,
      "loss": 0.0269,
      "step": 6345
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.8137964302942594e-05,
      "loss": 0.0508,
      "step": 6346
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.8105804791767168e-05,
      "loss": 0.0011,
      "step": 6347
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.8073645280591735e-05,
      "loss": 0.0532,
      "step": 6348
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.8041485769416305e-05,
      "loss": 0.0041,
      "step": 6349
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.8009326258240876e-05,
      "loss": 0.0001,
      "step": 6350
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.7977166747065446e-05,
      "loss": 0.0274,
      "step": 6351
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7945007235890017e-05,
      "loss": 0.0286,
      "step": 6352
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7912847724714584e-05,
      "loss": 0.0006,
      "step": 6353
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7880688213539154e-05,
      "loss": 0.026,
      "step": 6354
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7848528702363725e-05,
      "loss": 0.039,
      "step": 6355
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7816369191188295e-05,
      "loss": 0.0132,
      "step": 6356
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7784209680012866e-05,
      "loss": 0.0623,
      "step": 6357
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.7752050168837433e-05,
      "loss": 0.0005,
      "step": 6358
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.7719890657662007e-05,
      "loss": 0.0001,
      "step": 6359
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.7687731146486574e-05,
      "loss": 0.0012,
      "step": 6360
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.7655571635311145e-05,
      "loss": 0.0293,
      "step": 6361
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.762341212413571e-05,
      "loss": 0.0271,
      "step": 6362
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.7591252612960286e-05,
      "loss": 0.0016,
      "step": 6363
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.7559093101784853e-05,
      "loss": 0.0091,
      "step": 6364
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.7526933590609423e-05,
      "loss": 0.0497,
      "step": 6365
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7494774079433994e-05,
      "loss": 0.0017,
      "step": 6366
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7462614568258564e-05,
      "loss": 0.0001,
      "step": 6367
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7430455057083135e-05,
      "loss": 0.0001,
      "step": 6368
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7398295545907702e-05,
      "loss": 0.0253,
      "step": 6369
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7366136034732272e-05,
      "loss": 0.0006,
      "step": 6370
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7333976523556843e-05,
      "loss": 0.002,
      "step": 6371
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.7301817012381413e-05,
      "loss": 0.0084,
      "step": 6372
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.7269657501205984e-05,
      "loss": 0.0074,
      "step": 6373
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.723749799003055e-05,
      "loss": 0.0018,
      "step": 6374
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.720533847885512e-05,
      "loss": 0.0601,
      "step": 6375
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.7173178967679692e-05,
      "loss": 0.0244,
      "step": 6376
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.7141019456504263e-05,
      "loss": 0.0034,
      "step": 6377
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.710885994532883e-05,
      "loss": 0.0112,
      "step": 6378
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.70767004341534e-05,
      "loss": 0.0002,
      "step": 6379
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.704454092297797e-05,
      "loss": 0.0,
      "step": 6380
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.701238141180254e-05,
      "loss": 0.016,
      "step": 6381
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.6980221900627112e-05,
      "loss": 0.0002,
      "step": 6382
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.694806238945168e-05,
      "loss": 0.0243,
      "step": 6383
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.6915902878276253e-05,
      "loss": 0.0055,
      "step": 6384
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.688374336710082e-05,
      "loss": 0.001,
      "step": 6385
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.685158385592539e-05,
      "loss": 0.0095,
      "step": 6386
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.681942434474996e-05,
      "loss": 0.0496,
      "step": 6387
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.678726483357453e-05,
      "loss": 0.0075,
      "step": 6388
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.6755105322399102e-05,
      "loss": 0.0001,
      "step": 6389
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.672294581122367e-05,
      "loss": 0.002,
      "step": 6390
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.669078630004824e-05,
      "loss": 0.0002,
      "step": 6391
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.665862678887281e-05,
      "loss": 0.0134,
      "step": 6392
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.662646727769738e-05,
      "loss": 0.0166,
      "step": 6393
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.6594307766521948e-05,
      "loss": 0.0391,
      "step": 6394
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.6562148255346518e-05,
      "loss": 0.0162,
      "step": 6395
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.652998874417109e-05,
      "loss": 0.0004,
      "step": 6396
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.649782923299566e-05,
      "loss": 0.0146,
      "step": 6397
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.646566972182023e-05,
      "loss": 0.0427,
      "step": 6398
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.6433510210644797e-05,
      "loss": 0.0294,
      "step": 6399
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.640135069946937e-05,
      "loss": 0.0005,
      "step": 6400
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.6369191188293938e-05,
      "loss": 0.0326,
      "step": 6401
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.633703167711851e-05,
      "loss": 0.0212,
      "step": 6402
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.630487216594308e-05,
      "loss": 0.0011,
      "step": 6403
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.627271265476765e-05,
      "loss": 0.0067,
      "step": 6404
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.624055314359222e-05,
      "loss": 0.0604,
      "step": 6405
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.6208393632416787e-05,
      "loss": 0.0681,
      "step": 6406
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.6176234121241358e-05,
      "loss": 0.0433,
      "step": 6407
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.6144074610065928e-05,
      "loss": 0.0235,
      "step": 6408
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.61119150988905e-05,
      "loss": 0.0047,
      "step": 6409
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.6079755587715066e-05,
      "loss": 0.0021,
      "step": 6410
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.6047596076539636e-05,
      "loss": 0.0003,
      "step": 6411
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.601543656536421e-05,
      "loss": 0.0266,
      "step": 6412
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.5983277054188777e-05,
      "loss": 0.0001,
      "step": 6413
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5951117543013348e-05,
      "loss": 0.0001,
      "step": 6414
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5918958031837915e-05,
      "loss": 0.0472,
      "step": 6415
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5886798520662485e-05,
      "loss": 0.001,
      "step": 6416
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5854639009487056e-05,
      "loss": 0.0,
      "step": 6417
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5822479498311626e-05,
      "loss": 0.0037,
      "step": 6418
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5790319987136197e-05,
      "loss": 0.0536,
      "step": 6419
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.5758160475960764e-05,
      "loss": 0.0344,
      "step": 6420
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5726000964785338e-05,
      "loss": 0.0004,
      "step": 6421
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5693841453609905e-05,
      "loss": 0.0086,
      "step": 6422
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5661681942434476e-05,
      "loss": 0.0554,
      "step": 6423
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5629522431259043e-05,
      "loss": 0.0036,
      "step": 6424
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5597362920083617e-05,
      "loss": 0.0506,
      "step": 6425
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5565203408908187e-05,
      "loss": 0.0003,
      "step": 6426
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.5533043897732754e-05,
      "loss": 0.019,
      "step": 6427
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5500884386557325e-05,
      "loss": 0.0023,
      "step": 6428
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5468724875381895e-05,
      "loss": 0.0002,
      "step": 6429
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5436565364206466e-05,
      "loss": 0.0018,
      "step": 6430
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5404405853031033e-05,
      "loss": 0.0001,
      "step": 6431
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5372246341855603e-05,
      "loss": 0.0027,
      "step": 6432
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5340086830680174e-05,
      "loss": 0.0002,
      "step": 6433
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.5307927319504745e-05,
      "loss": 0.0006,
      "step": 6434
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5275767808329315e-05,
      "loss": 0.0323,
      "step": 6435
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5243608297153882e-05,
      "loss": 0.0408,
      "step": 6436
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5211448785978454e-05,
      "loss": 0.0443,
      "step": 6437
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5179289274803025e-05,
      "loss": 0.0623,
      "step": 6438
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5147129763627594e-05,
      "loss": 0.0081,
      "step": 6439
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5114970252452162e-05,
      "loss": 0.0,
      "step": 6440
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.5082810741276735e-05,
      "loss": 0.0225,
      "step": 6441
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.5050651230101304e-05,
      "loss": 0.0004,
      "step": 6442
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.5018491718925872e-05,
      "loss": 0.0036,
      "step": 6443
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.4986332207750441e-05,
      "loss": 0.0128,
      "step": 6444
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.4954172696575013e-05,
      "loss": 0.0004,
      "step": 6445
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.4922013185399584e-05,
      "loss": 0.0001,
      "step": 6446
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.4889853674224153e-05,
      "loss": 0.0058,
      "step": 6447
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.4857694163048722e-05,
      "loss": 0.0774,
      "step": 6448
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4825534651873294e-05,
      "loss": 0.008,
      "step": 6449
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4793375140697863e-05,
      "loss": 0.0533,
      "step": 6450
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4761215629522431e-05,
      "loss": 0.0109,
      "step": 6451
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4729056118347002e-05,
      "loss": 0.0141,
      "step": 6452
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4696896607171572e-05,
      "loss": 0.0372,
      "step": 6453
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4664737095996143e-05,
      "loss": 0.0006,
      "step": 6454
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.4632577584820712e-05,
      "loss": 0.0178,
      "step": 6455
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.460041807364528e-05,
      "loss": 0.0039,
      "step": 6456
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.456825856246985e-05,
      "loss": 0.0001,
      "step": 6457
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.4536099051294422e-05,
      "loss": 0.0347,
      "step": 6458
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.450393954011899e-05,
      "loss": 0.0311,
      "step": 6459
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.4471780028943561e-05,
      "loss": 0.0199,
      "step": 6460
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.443962051776813e-05,
      "loss": 0.0333,
      "step": 6461
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.4407461006592702e-05,
      "loss": 0.0001,
      "step": 6462
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.437530149541727e-05,
      "loss": 0.0,
      "step": 6463
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.434314198424184e-05,
      "loss": 0.0004,
      "step": 6464
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.4310982473066408e-05,
      "loss": 0.0202,
      "step": 6465
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.427882296189098e-05,
      "loss": 0.0496,
      "step": 6466
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.424666345071555e-05,
      "loss": 0.0225,
      "step": 6467
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.421450393954012e-05,
      "loss": 0.0498,
      "step": 6468
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.4182344428364689e-05,
      "loss": 0.0826,
      "step": 6469
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.4150184917189261e-05,
      "loss": 0.0011,
      "step": 6470
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.411802540601383e-05,
      "loss": 0.0015,
      "step": 6471
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.4085865894838399e-05,
      "loss": 0.0017,
      "step": 6472
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.4053706383662967e-05,
      "loss": 0.0035,
      "step": 6473
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.402154687248754e-05,
      "loss": 0.0659,
      "step": 6474
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.3989387361312108e-05,
      "loss": 0.0004,
      "step": 6475
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.3957227850136679e-05,
      "loss": 0.014,
      "step": 6476
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.3925068338961248e-05,
      "loss": 0.0139,
      "step": 6477
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.389290882778582e-05,
      "loss": 0.0004,
      "step": 6478
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.3860749316610389e-05,
      "loss": 0.0004,
      "step": 6479
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.3828589805434958e-05,
      "loss": 0.02,
      "step": 6480
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.3796430294259526e-05,
      "loss": 0.0004,
      "step": 6481
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.3764270783084099e-05,
      "loss": 0.0261,
      "step": 6482
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3732111271908667e-05,
      "loss": 0.0478,
      "step": 6483
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3699951760733238e-05,
      "loss": 0.0055,
      "step": 6484
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3667792249557807e-05,
      "loss": 0.0223,
      "step": 6485
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3635632738382379e-05,
      "loss": 0.0697,
      "step": 6486
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3603473227206948e-05,
      "loss": 0.0002,
      "step": 6487
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3571313716031517e-05,
      "loss": 0.0008,
      "step": 6488
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.3539154204856085e-05,
      "loss": 0.0042,
      "step": 6489
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3506994693680658e-05,
      "loss": 0.0698,
      "step": 6490
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3474835182505226e-05,
      "loss": 0.0029,
      "step": 6491
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3442675671329797e-05,
      "loss": 0.0534,
      "step": 6492
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3410516160154366e-05,
      "loss": 0.026,
      "step": 6493
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3378356648978938e-05,
      "loss": 0.0053,
      "step": 6494
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3346197137803507e-05,
      "loss": 0.0144,
      "step": 6495
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.3314037626628076e-05,
      "loss": 0.0034,
      "step": 6496
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3281878115452644e-05,
      "loss": 0.0256,
      "step": 6497
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3249718604277217e-05,
      "loss": 0.0024,
      "step": 6498
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3217559093101785e-05,
      "loss": 0.0002,
      "step": 6499
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3185399581926356e-05,
      "loss": 0.0325,
      "step": 6500
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3153240070750925e-05,
      "loss": 0.0003,
      "step": 6501
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3121080559575494e-05,
      "loss": 0.0292,
      "step": 6502
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.3088921048400066e-05,
      "loss": 0.0337,
      "step": 6503
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.3056761537224635e-05,
      "loss": 0.0112,
      "step": 6504
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.3024602026049203e-05,
      "loss": 0.0014,
      "step": 6505
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.2992442514873774e-05,
      "loss": 0.002,
      "step": 6506
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.2960283003698344e-05,
      "loss": 0.0059,
      "step": 6507
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.2928123492522915e-05,
      "loss": 0.0,
      "step": 6508
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.2895963981347484e-05,
      "loss": 0.0559,
      "step": 6509
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.2863804470172053e-05,
      "loss": 0.0027,
      "step": 6510
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2831644958996625e-05,
      "loss": 0.0064,
      "step": 6511
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2799485447821194e-05,
      "loss": 0.0268,
      "step": 6512
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2767325936645762e-05,
      "loss": 0.0092,
      "step": 6513
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2735166425470333e-05,
      "loss": 0.0013,
      "step": 6514
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2703006914294903e-05,
      "loss": 0.0063,
      "step": 6515
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2670847403119474e-05,
      "loss": 0.0866,
      "step": 6516
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.2638687891944043e-05,
      "loss": 0.0015,
      "step": 6517
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2606528380768612e-05,
      "loss": 0.0572,
      "step": 6518
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2574368869593184e-05,
      "loss": 0.0385,
      "step": 6519
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2542209358417753e-05,
      "loss": 0.0006,
      "step": 6520
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2510049847242321e-05,
      "loss": 0.0,
      "step": 6521
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2477890336066892e-05,
      "loss": 0.0007,
      "step": 6522
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2445730824891462e-05,
      "loss": 0.0003,
      "step": 6523
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.2413571313716033e-05,
      "loss": 0.0,
      "step": 6524
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.2381411802540602e-05,
      "loss": 0.0413,
      "step": 6525
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.2349252291365172e-05,
      "loss": 0.0089,
      "step": 6526
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.2317092780189741e-05,
      "loss": 0.0074,
      "step": 6527
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.2284933269014312e-05,
      "loss": 0.0547,
      "step": 6528
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.225277375783888e-05,
      "loss": 0.0015,
      "step": 6529
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.2220614246663451e-05,
      "loss": 0.0134,
      "step": 6530
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.2188454735488022e-05,
      "loss": 0.0252,
      "step": 6531
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.2156295224312592e-05,
      "loss": 0.0007,
      "step": 6532
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.212413571313716e-05,
      "loss": 0.0383,
      "step": 6533
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.2091976201961731e-05,
      "loss": 0.0001,
      "step": 6534
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.20598166907863e-05,
      "loss": 0.0432,
      "step": 6535
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.202765717961087e-05,
      "loss": 0.0379,
      "step": 6536
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.199549766843544e-05,
      "loss": 0.0184,
      "step": 6537
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.1963338157260012e-05,
      "loss": 0.0031,
      "step": 6538
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.193117864608458e-05,
      "loss": 0.0289,
      "step": 6539
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.189901913490915e-05,
      "loss": 0.0155,
      "step": 6540
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.186685962373372e-05,
      "loss": 0.0186,
      "step": 6541
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.1834700112558289e-05,
      "loss": 0.0002,
      "step": 6542
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.180254060138286e-05,
      "loss": 0.0003,
      "step": 6543
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.1770381090207428e-05,
      "loss": 0.0138,
      "step": 6544
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.1738221579032e-05,
      "loss": 0.0176,
      "step": 6545
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.1706062067856569e-05,
      "loss": 0.0444,
      "step": 6546
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.167390255668114e-05,
      "loss": 0.0214,
      "step": 6547
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.1641743045505708e-05,
      "loss": 0.0096,
      "step": 6548
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.1609583534330279e-05,
      "loss": 0.0001,
      "step": 6549
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.1577424023154848e-05,
      "loss": 0.0001,
      "step": 6550
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.1545264511979418e-05,
      "loss": 0.0105,
      "step": 6551
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.1513105000803987e-05,
      "loss": 0.0076,
      "step": 6552
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.148094548962856e-05,
      "loss": 0.0337,
      "step": 6553
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.1448785978453128e-05,
      "loss": 0.0111,
      "step": 6554
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.1416626467277699e-05,
      "loss": 0.0106,
      "step": 6555
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.1384466956102267e-05,
      "loss": 0.0586,
      "step": 6556
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.1352307444926838e-05,
      "loss": 0.057,
      "step": 6557
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.1320147933751407e-05,
      "loss": 0.0015,
      "step": 6558
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1287988422575977e-05,
      "loss": 0.001,
      "step": 6559
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1255828911400548e-05,
      "loss": 0.0298,
      "step": 6560
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1223669400225118e-05,
      "loss": 0.044,
      "step": 6561
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1191509889049687e-05,
      "loss": 0.0107,
      "step": 6562
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1159350377874258e-05,
      "loss": 0.0659,
      "step": 6563
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1127190866698826e-05,
      "loss": 0.0154,
      "step": 6564
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.1095031355523397e-05,
      "loss": 0.0,
      "step": 6565
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.1062871844347966e-05,
      "loss": 0.0285,
      "step": 6566
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.1030712333172536e-05,
      "loss": 0.0006,
      "step": 6567
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0998552821997107e-05,
      "loss": 0.0009,
      "step": 6568
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0966393310821677e-05,
      "loss": 0.0253,
      "step": 6569
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0934233799646246e-05,
      "loss": 0.0005,
      "step": 6570
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0902074288470817e-05,
      "loss": 0.0002,
      "step": 6571
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0869914777295385e-05,
      "loss": 0.0039,
      "step": 6572
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0837755266119956e-05,
      "loss": 0.0007,
      "step": 6573
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0805595754944525e-05,
      "loss": 0.0,
      "step": 6574
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0773436243769095e-05,
      "loss": 0.0025,
      "step": 6575
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0741276732593666e-05,
      "loss": 0.0537,
      "step": 6576
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0709117221418236e-05,
      "loss": 0.011,
      "step": 6577
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0676957710242805e-05,
      "loss": 0.0391,
      "step": 6578
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.0644798199067376e-05,
      "loss": 0.0254,
      "step": 6579
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0612638687891944e-05,
      "loss": 0.0013,
      "step": 6580
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0580479176716513e-05,
      "loss": 0.0069,
      "step": 6581
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0548319665541084e-05,
      "loss": 0.0024,
      "step": 6582
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0516160154365654e-05,
      "loss": 0.0595,
      "step": 6583
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0484000643190225e-05,
      "loss": 0.0118,
      "step": 6584
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0451841132014794e-05,
      "loss": 0.0112,
      "step": 6585
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.0419681620839364e-05,
      "loss": 0.0005,
      "step": 6586
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0387522109663933e-05,
      "loss": 0.0054,
      "step": 6587
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0355362598488503e-05,
      "loss": 0.0003,
      "step": 6588
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0323203087313072e-05,
      "loss": 0.0358,
      "step": 6589
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0291043576137643e-05,
      "loss": 0.0039,
      "step": 6590
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0258884064962213e-05,
      "loss": 0.0092,
      "step": 6591
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0226724553786784e-05,
      "loss": 0.0003,
      "step": 6592
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.0194565042611353e-05,
      "loss": 0.0051,
      "step": 6593
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.0162405531435923e-05,
      "loss": 0.0,
      "step": 6594
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.0130246020260492e-05,
      "loss": 0.0002,
      "step": 6595
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.0098086509085062e-05,
      "loss": 0.0006,
      "step": 6596
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.0065926997909631e-05,
      "loss": 0.0281,
      "step": 6597
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.0033767486734202e-05,
      "loss": 0.0117,
      "step": 6598
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.0001607975558772e-05,
      "loss": 0.01,
      "step": 6599
    },
    {
      "epoch": 9.54,
      "learning_rate": 9.969448464383343e-06,
      "loss": 0.0001,
      "step": 6600
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.937288953207912e-06,
      "loss": 0.0197,
      "step": 6601
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.905129442032482e-06,
      "loss": 0.0146,
      "step": 6602
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.872969930857051e-06,
      "loss": 0.0561,
      "step": 6603
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.840810419681621e-06,
      "loss": 0.0004,
      "step": 6604
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.80865090850619e-06,
      "loss": 0.0546,
      "step": 6605
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.77649139733076e-06,
      "loss": 0.0427,
      "step": 6606
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.744331886155331e-06,
      "loss": 0.062,
      "step": 6607
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.712172374979902e-06,
      "loss": 0.0144,
      "step": 6608
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.68001286380447e-06,
      "loss": 0.0042,
      "step": 6609
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.647853352629041e-06,
      "loss": 0.0005,
      "step": 6610
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.61569384145361e-06,
      "loss": 0.0002,
      "step": 6611
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.58353433027818e-06,
      "loss": 0.0019,
      "step": 6612
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.55137481910275e-06,
      "loss": 0.0001,
      "step": 6613
    },
    {
      "epoch": 9.56,
      "learning_rate": 9.51921530792732e-06,
      "loss": 0.0264,
      "step": 6614
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.48705579675189e-06,
      "loss": 0.0547,
      "step": 6615
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.45489628557646e-06,
      "loss": 0.0019,
      "step": 6616
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.42273677440103e-06,
      "loss": 0.0001,
      "step": 6617
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.3905772632256e-06,
      "loss": 0.0062,
      "step": 6618
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.358417752050169e-06,
      "loss": 0.0363,
      "step": 6619
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.32625824087474e-06,
      "loss": 0.0189,
      "step": 6620
    },
    {
      "epoch": 9.57,
      "learning_rate": 9.294098729699308e-06,
      "loss": 0.0288,
      "step": 6621
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.261939218523879e-06,
      "loss": 0.0669,
      "step": 6622
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.22977970734845e-06,
      "loss": 0.0011,
      "step": 6623
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.197620196173018e-06,
      "loss": 0.0002,
      "step": 6624
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.165460684997589e-06,
      "loss": 0.0001,
      "step": 6625
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.133301173822157e-06,
      "loss": 0.0001,
      "step": 6626
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.101141662646728e-06,
      "loss": 0.0,
      "step": 6627
    },
    {
      "epoch": 9.58,
      "learning_rate": 9.068982151471297e-06,
      "loss": 0.0501,
      "step": 6628
    },
    {
      "epoch": 9.59,
      "learning_rate": 9.036822640295867e-06,
      "loss": 0.0516,
      "step": 6629
    },
    {
      "epoch": 9.59,
      "learning_rate": 9.004663129120438e-06,
      "loss": 0.0141,
      "step": 6630
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.972503617945008e-06,
      "loss": 0.0042,
      "step": 6631
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.940344106769577e-06,
      "loss": 0.0181,
      "step": 6632
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.908184595594148e-06,
      "loss": 0.0764,
      "step": 6633
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.876025084418717e-06,
      "loss": 0.0026,
      "step": 6634
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.843865573243287e-06,
      "loss": 0.0015,
      "step": 6635
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.811706062067856e-06,
      "loss": 0.0368,
      "step": 6636
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.779546550892426e-06,
      "loss": 0.0005,
      "step": 6637
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.747387039716997e-06,
      "loss": 0.0052,
      "step": 6638
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.715227528541567e-06,
      "loss": 0.0008,
      "step": 6639
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.683068017366136e-06,
      "loss": 0.0,
      "step": 6640
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.650908506190707e-06,
      "loss": 0.0001,
      "step": 6641
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.618748995015276e-06,
      "loss": 0.0001,
      "step": 6642
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.586589483839846e-06,
      "loss": 0.0059,
      "step": 6643
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.554429972664415e-06,
      "loss": 0.0004,
      "step": 6644
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.522270461488985e-06,
      "loss": 0.0227,
      "step": 6645
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.490110950313556e-06,
      "loss": 0.017,
      "step": 6646
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.457951439138126e-06,
      "loss": 0.0214,
      "step": 6647
    },
    {
      "epoch": 9.61,
      "learning_rate": 8.425791927962695e-06,
      "loss": 0.0078,
      "step": 6648
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.393632416787266e-06,
      "loss": 0.0016,
      "step": 6649
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.361472905611835e-06,
      "loss": 0.021,
      "step": 6650
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.329313394436405e-06,
      "loss": 0.0013,
      "step": 6651
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.297153883260974e-06,
      "loss": 0.0005,
      "step": 6652
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.264994372085544e-06,
      "loss": 0.0358,
      "step": 6653
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.232834860910115e-06,
      "loss": 0.0012,
      "step": 6654
    },
    {
      "epoch": 9.62,
      "learning_rate": 8.200675349734685e-06,
      "loss": 0.0018,
      "step": 6655
    },
    {
      "epoch": 9.63,
      "learning_rate": 8.168515838559254e-06,
      "loss": 0.0003,
      "step": 6656
    },
    {
      "epoch": 9.63,
      "learning_rate": 8.136356327383825e-06,
      "loss": 0.0224,
      "step": 6657
    },
    {
      "epoch": 9.63,
      "learning_rate": 8.104196816208394e-06,
      "loss": 0.0408,
      "step": 6658
    },
    {
      "epoch": 9.63,
      "learning_rate": 8.072037305032964e-06,
      "loss": 0.0,
      "step": 6659
    },
    {
      "epoch": 9.63,
      "learning_rate": 8.039877793857533e-06,
      "loss": 0.0337,
      "step": 6660
    },
    {
      "epoch": 9.63,
      "learning_rate": 8.007718282682105e-06,
      "loss": 0.016,
      "step": 6661
    },
    {
      "epoch": 9.63,
      "learning_rate": 7.975558771506674e-06,
      "loss": 0.0,
      "step": 6662
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.943399260331243e-06,
      "loss": 0.0191,
      "step": 6663
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.911239749155813e-06,
      "loss": 0.0275,
      "step": 6664
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.879080237980382e-06,
      "loss": 0.0091,
      "step": 6665
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.846920726804953e-06,
      "loss": 0.0272,
      "step": 6666
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.814761215629521e-06,
      "loss": 0.0158,
      "step": 6667
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.782601704454094e-06,
      "loss": 0.0003,
      "step": 6668
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.750442193278662e-06,
      "loss": 0.0359,
      "step": 6669
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.718282682103233e-06,
      "loss": 0.0036,
      "step": 6670
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.686123170927802e-06,
      "loss": 0.0051,
      "step": 6671
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.653963659752372e-06,
      "loss": 0.0002,
      "step": 6672
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.621804148576941e-06,
      "loss": 0.014,
      "step": 6673
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.5896446374015124e-06,
      "loss": 0.0813,
      "step": 6674
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.557485126226081e-06,
      "loss": 0.0131,
      "step": 6675
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.525325615050652e-06,
      "loss": 0.0001,
      "step": 6676
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.493166103875221e-06,
      "loss": 0.0471,
      "step": 6677
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.461006592699792e-06,
      "loss": 0.0095,
      "step": 6678
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.428847081524361e-06,
      "loss": 0.0218,
      "step": 6679
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.396687570348931e-06,
      "loss": 0.0019,
      "step": 6680
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.364528059173501e-06,
      "loss": 0.0135,
      "step": 6681
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.3323685479980714e-06,
      "loss": 0.0,
      "step": 6682
    },
    {
      "epoch": 9.66,
      "learning_rate": 7.30020903682264e-06,
      "loss": 0.0407,
      "step": 6683
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.268049525647211e-06,
      "loss": 0.0022,
      "step": 6684
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.2358900144717804e-06,
      "loss": 0.0757,
      "step": 6685
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.203730503296351e-06,
      "loss": 0.0091,
      "step": 6686
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.17157099212092e-06,
      "loss": 0.0196,
      "step": 6687
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.13941148094549e-06,
      "loss": 0.0002,
      "step": 6688
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.10725196977006e-06,
      "loss": 0.0009,
      "step": 6689
    },
    {
      "epoch": 9.67,
      "learning_rate": 7.0750924585946305e-06,
      "loss": 0.0002,
      "step": 6690
    },
    {
      "epoch": 9.68,
      "learning_rate": 7.042932947419199e-06,
      "loss": 0.0803,
      "step": 6691
    },
    {
      "epoch": 9.68,
      "learning_rate": 7.01077343624377e-06,
      "loss": 0.0421,
      "step": 6692
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.9786139250683395e-06,
      "loss": 0.0004,
      "step": 6693
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.94645441389291e-06,
      "loss": 0.0003,
      "step": 6694
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.914294902717479e-06,
      "loss": 0.0005,
      "step": 6695
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.882135391542049e-06,
      "loss": 0.0201,
      "step": 6696
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.849975880366619e-06,
      "loss": 0.0085,
      "step": 6697
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.8178163691911895e-06,
      "loss": 0.0003,
      "step": 6698
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.785656858015758e-06,
      "loss": 0.0192,
      "step": 6699
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.753497346840329e-06,
      "loss": 0.0614,
      "step": 6700
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.7213378356648985e-06,
      "loss": 0.0,
      "step": 6701
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.689178324489469e-06,
      "loss": 0.0018,
      "step": 6702
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.657018813314038e-06,
      "loss": 0.0478,
      "step": 6703
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.624859302138608e-06,
      "loss": 0.0475,
      "step": 6704
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.592699790963178e-06,
      "loss": 0.0501,
      "step": 6705
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.560540279787747e-06,
      "loss": 0.0026,
      "step": 6706
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.528380768612317e-06,
      "loss": 0.0649,
      "step": 6707
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.496221257436887e-06,
      "loss": 0.0598,
      "step": 6708
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.4640617462614575e-06,
      "loss": 0.0062,
      "step": 6709
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.431902235086026e-06,
      "loss": 0.0574,
      "step": 6710
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.399742723910597e-06,
      "loss": 0.0507,
      "step": 6711
    },
    {
      "epoch": 9.71,
      "learning_rate": 6.3675832127351665e-06,
      "loss": 0.0012,
      "step": 6712
    },
    {
      "epoch": 9.71,
      "learning_rate": 6.335423701559737e-06,
      "loss": 0.0173,
      "step": 6713
    },
    {
      "epoch": 9.71,
      "learning_rate": 6.303264190384306e-06,
      "loss": 0.0365,
      "step": 6714
    },
    {
      "epoch": 9.71,
      "learning_rate": 6.271104679208876e-06,
      "loss": 0.0002,
      "step": 6715
    },
    {
      "epoch": 9.71,
      "learning_rate": 6.238945168033446e-06,
      "loss": 0.0011,
      "step": 6716
    },
    {
      "epoch": 9.71,
      "learning_rate": 6.2067856568580165e-06,
      "loss": 0.0002,
      "step": 6717
    },
    {
      "epoch": 9.72,
      "learning_rate": 6.174626145682586e-06,
      "loss": 0.0,
      "step": 6718
    },
    {
      "epoch": 9.72,
      "learning_rate": 6.142466634507156e-06,
      "loss": 0.0003,
      "step": 6719
    },
    {
      "epoch": 9.72,
      "learning_rate": 6.1103071233317255e-06,
      "loss": 0.0123,
      "step": 6720
    },
    {
      "epoch": 9.72,
      "learning_rate": 6.078147612156296e-06,
      "loss": 0.0017,
      "step": 6721
    },
    {
      "epoch": 9.72,
      "learning_rate": 6.045988100980866e-06,
      "loss": 0.027,
      "step": 6722
    },
    {
      "epoch": 9.72,
      "learning_rate": 6.013828589805435e-06,
      "loss": 0.0053,
      "step": 6723
    },
    {
      "epoch": 9.72,
      "learning_rate": 5.981669078630006e-06,
      "loss": 0.004,
      "step": 6724
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.949509567454575e-06,
      "loss": 0.028,
      "step": 6725
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.917350056279144e-06,
      "loss": 0.0013,
      "step": 6726
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.885190545103714e-06,
      "loss": 0.0194,
      "step": 6727
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.8530310339282845e-06,
      "loss": 0.0001,
      "step": 6728
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.820871522752854e-06,
      "loss": 0.0241,
      "step": 6729
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.788712011577424e-06,
      "loss": 0.0012,
      "step": 6730
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.7565525004019935e-06,
      "loss": 0.0,
      "step": 6731
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.724392989226564e-06,
      "loss": 0.0007,
      "step": 6732
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.692233478051134e-06,
      "loss": 0.0001,
      "step": 6733
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.660073966875703e-06,
      "loss": 0.0106,
      "step": 6734
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.627914455700274e-06,
      "loss": 0.0002,
      "step": 6735
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.5957549445248435e-06,
      "loss": 0.0302,
      "step": 6736
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.563595433349413e-06,
      "loss": 0.0,
      "step": 6737
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.531435922173983e-06,
      "loss": 0.0136,
      "step": 6738
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.499276410998553e-06,
      "loss": 0.0087,
      "step": 6739
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.467116899823123e-06,
      "loss": 0.0008,
      "step": 6740
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.434957388647693e-06,
      "loss": 0.055,
      "step": 6741
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.402797877472262e-06,
      "loss": 0.0063,
      "step": 6742
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.370638366296833e-06,
      "loss": 0.0023,
      "step": 6743
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.3384788551214025e-06,
      "loss": 0.0401,
      "step": 6744
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.306319343945972e-06,
      "loss": 0.0358,
      "step": 6745
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.274159832770542e-06,
      "loss": 0.0413,
      "step": 6746
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.242000321595112e-06,
      "loss": 0.0003,
      "step": 6747
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.209840810419682e-06,
      "loss": 0.0029,
      "step": 6748
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.177681299244252e-06,
      "loss": 0.0313,
      "step": 6749
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.145521788068821e-06,
      "loss": 0.0449,
      "step": 6750
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.113362276893392e-06,
      "loss": 0.0057,
      "step": 6751
    },
    {
      "epoch": 9.76,
      "learning_rate": 5.0812027657179616e-06,
      "loss": 0.0603,
      "step": 6752
    },
    {
      "epoch": 9.77,
      "learning_rate": 5.049043254542531e-06,
      "loss": 0.0295,
      "step": 6753
    },
    {
      "epoch": 9.77,
      "learning_rate": 5.016883743367101e-06,
      "loss": 0.0779,
      "step": 6754
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.984724232191671e-06,
      "loss": 0.0092,
      "step": 6755
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.952564721016241e-06,
      "loss": 0.0002,
      "step": 6756
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.920405209840811e-06,
      "loss": 0.0007,
      "step": 6757
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.88824569866538e-06,
      "loss": 0.0139,
      "step": 6758
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.856086187489951e-06,
      "loss": 0.003,
      "step": 6759
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.8239266763145206e-06,
      "loss": 0.0001,
      "step": 6760
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.79176716513909e-06,
      "loss": 0.0083,
      "step": 6761
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.75960765396366e-06,
      "loss": 0.0262,
      "step": 6762
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.72744814278823e-06,
      "loss": 0.0002,
      "step": 6763
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.6952886316128e-06,
      "loss": 0.0086,
      "step": 6764
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.66312912043737e-06,
      "loss": 0.0005,
      "step": 6765
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.630969609261939e-06,
      "loss": 0.0038,
      "step": 6766
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.598810098086509e-06,
      "loss": 0.0361,
      "step": 6767
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.566650586911079e-06,
      "loss": 0.0007,
      "step": 6768
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.534491075735648e-06,
      "loss": 0.0235,
      "step": 6769
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.502331564560219e-06,
      "loss": 0.0007,
      "step": 6770
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.470172053384789e-06,
      "loss": 0.0001,
      "step": 6771
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.438012542209358e-06,
      "loss": 0.0269,
      "step": 6772
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.405853031033928e-06,
      "loss": 0.0012,
      "step": 6773
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.3736935198584984e-06,
      "loss": 0.0104,
      "step": 6774
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.341534008683068e-06,
      "loss": 0.0151,
      "step": 6775
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.309374497507638e-06,
      "loss": 0.0009,
      "step": 6776
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.2772149863322074e-06,
      "loss": 0.0003,
      "step": 6777
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.245055475156778e-06,
      "loss": 0.013,
      "step": 6778
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.212895963981348e-06,
      "loss": 0.0,
      "step": 6779
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.180736452805917e-06,
      "loss": 0.0012,
      "step": 6780
    },
    {
      "epoch": 9.81,
      "learning_rate": 4.148576941630487e-06,
      "loss": 0.0004,
      "step": 6781
    },
    {
      "epoch": 9.81,
      "learning_rate": 4.1164174304550574e-06,
      "loss": 0.0515,
      "step": 6782
    },
    {
      "epoch": 9.81,
      "learning_rate": 4.084257919279627e-06,
      "loss": 0.0315,
      "step": 6783
    },
    {
      "epoch": 9.81,
      "learning_rate": 4.052098408104197e-06,
      "loss": 0.0002,
      "step": 6784
    },
    {
      "epoch": 9.81,
      "learning_rate": 4.0199388969287664e-06,
      "loss": 0.0001,
      "step": 6785
    },
    {
      "epoch": 9.81,
      "learning_rate": 3.987779385753337e-06,
      "loss": 0.0046,
      "step": 6786
    },
    {
      "epoch": 9.81,
      "learning_rate": 3.955619874577907e-06,
      "loss": 0.0001,
      "step": 6787
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.923460363402476e-06,
      "loss": 0.0006,
      "step": 6788
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.891300852227047e-06,
      "loss": 0.0217,
      "step": 6789
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.8591413410516165e-06,
      "loss": 0.0063,
      "step": 6790
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.826981829876186e-06,
      "loss": 0.0003,
      "step": 6791
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.7948223187007562e-06,
      "loss": 0.0003,
      "step": 6792
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.762662807525326e-06,
      "loss": 0.0,
      "step": 6793
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.730503296349896e-06,
      "loss": 0.0037,
      "step": 6794
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.6983437851744656e-06,
      "loss": 0.0014,
      "step": 6795
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.6661842739990357e-06,
      "loss": 0.0003,
      "step": 6796
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.6340247628236054e-06,
      "loss": 0.0102,
      "step": 6797
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.6018652516481755e-06,
      "loss": 0.0042,
      "step": 6798
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.569705740472745e-06,
      "loss": 0.0103,
      "step": 6799
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.5375462292973152e-06,
      "loss": 0.0002,
      "step": 6800
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.505386718121885e-06,
      "loss": 0.0009,
      "step": 6801
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.473227206946455e-06,
      "loss": 0.0049,
      "step": 6802
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.4410676957710246e-06,
      "loss": 0.0819,
      "step": 6803
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.4089081845955947e-06,
      "loss": 0.0381,
      "step": 6804
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.3767486734201644e-06,
      "loss": 0.0475,
      "step": 6805
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.3445891622447345e-06,
      "loss": 0.0,
      "step": 6806
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.312429651069304e-06,
      "loss": 0.0127,
      "step": 6807
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.2802701398938734e-06,
      "loss": 0.0054,
      "step": 6808
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.2481106287184435e-06,
      "loss": 0.0087,
      "step": 6809
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.215951117543013e-06,
      "loss": 0.0058,
      "step": 6810
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.1837916063675832e-06,
      "loss": 0.0002,
      "step": 6811
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.151632095192153e-06,
      "loss": 0.0174,
      "step": 6812
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.119472584016723e-06,
      "loss": 0.0769,
      "step": 6813
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.087313072841293e-06,
      "loss": 0.0075,
      "step": 6814
    },
    {
      "epoch": 9.86,
      "learning_rate": 3.0551535616658627e-06,
      "loss": 0.0002,
      "step": 6815
    },
    {
      "epoch": 9.86,
      "learning_rate": 3.022994050490433e-06,
      "loss": 0.0974,
      "step": 6816
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.990834539315003e-06,
      "loss": 0.0698,
      "step": 6817
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.958675028139572e-06,
      "loss": 0.0002,
      "step": 6818
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.9265155169641423e-06,
      "loss": 0.0005,
      "step": 6819
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.894356005788712e-06,
      "loss": 0.0014,
      "step": 6820
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.862196494613282e-06,
      "loss": 0.0241,
      "step": 6821
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.8300369834378517e-06,
      "loss": 0.0162,
      "step": 6822
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.7978774722624218e-06,
      "loss": 0.0,
      "step": 6823
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.7657179610869914e-06,
      "loss": 0.0632,
      "step": 6824
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.7335584499115615e-06,
      "loss": 0.0283,
      "step": 6825
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.701398938736131e-06,
      "loss": 0.0188,
      "step": 6826
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.6692394275607013e-06,
      "loss": 0.0042,
      "step": 6827
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.637079916385271e-06,
      "loss": 0.0,
      "step": 6828
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.604920405209841e-06,
      "loss": 0.0234,
      "step": 6829
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.5727608940344107e-06,
      "loss": 0.0066,
      "step": 6830
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.5406013828589808e-06,
      "loss": 0.0002,
      "step": 6831
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.5084418716835504e-06,
      "loss": 0.0067,
      "step": 6832
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.4762823605081205e-06,
      "loss": 0.0472,
      "step": 6833
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.44412284933269e-06,
      "loss": 0.0212,
      "step": 6834
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.4119633381572603e-06,
      "loss": 0.0588,
      "step": 6835
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.37980382698183e-06,
      "loss": 0.0001,
      "step": 6836
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.3476443158064e-06,
      "loss": 0.0049,
      "step": 6837
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.3154848046309697e-06,
      "loss": 0.005,
      "step": 6838
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.2833252934555394e-06,
      "loss": 0.0035,
      "step": 6839
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.2511657822801095e-06,
      "loss": 0.0002,
      "step": 6840
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.219006271104679e-06,
      "loss": 0.0054,
      "step": 6841
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.1868467599292492e-06,
      "loss": 0.0663,
      "step": 6842
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.154687248753819e-06,
      "loss": 0.0003,
      "step": 6843
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.122527737578389e-06,
      "loss": 0.0001,
      "step": 6844
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.0903682264029586e-06,
      "loss": 0.0149,
      "step": 6845
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.0582087152275287e-06,
      "loss": 0.0197,
      "step": 6846
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.0260492040520984e-06,
      "loss": 0.0663,
      "step": 6847
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.9938896928766685e-06,
      "loss": 0.043,
      "step": 6848
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.961730181701238e-06,
      "loss": 0.0036,
      "step": 6849
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.9295706705258082e-06,
      "loss": 0.0006,
      "step": 6850
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8974111593503781e-06,
      "loss": 0.0061,
      "step": 6851
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.865251648174948e-06,
      "loss": 0.0383,
      "step": 6852
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8330921369995179e-06,
      "loss": 0.0002,
      "step": 6853
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8009326258240877e-06,
      "loss": 0.0,
      "step": 6854
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.7687731146486576e-06,
      "loss": 0.0184,
      "step": 6855
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.7366136034732275e-06,
      "loss": 0.0017,
      "step": 6856
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.7044540922977974e-06,
      "loss": 0.0263,
      "step": 6857
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.6722945811223672e-06,
      "loss": 0.0013,
      "step": 6858
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.6401350699469367e-06,
      "loss": 0.0201,
      "step": 6859
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.6079755587715066e-06,
      "loss": 0.0096,
      "step": 6860
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.5758160475960765e-06,
      "loss": 0.0138,
      "step": 6861
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.5436565364206465e-06,
      "loss": 0.01,
      "step": 6862
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.5114970252452164e-06,
      "loss": 0.0003,
      "step": 6863
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.479337514069786e-06,
      "loss": 0.0519,
      "step": 6864
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.447178002894356e-06,
      "loss": 0.0005,
      "step": 6865
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.4150184917189258e-06,
      "loss": 0.0432,
      "step": 6866
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.3828589805434957e-06,
      "loss": 0.0002,
      "step": 6867
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.3506994693680656e-06,
      "loss": 0.0001,
      "step": 6868
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.3185399581926355e-06,
      "loss": 0.0003,
      "step": 6869
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.2863804470172053e-06,
      "loss": 0.0001,
      "step": 6870
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.2542209358417752e-06,
      "loss": 0.0,
      "step": 6871
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.222061424666345e-06,
      "loss": 0.0411,
      "step": 6872
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.189901913490915e-06,
      "loss": 0.0221,
      "step": 6873
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.1577424023154849e-06,
      "loss": 0.0118,
      "step": 6874
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.1255828911400547e-06,
      "loss": 0.0001,
      "step": 6875
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.0934233799646246e-06,
      "loss": 0.0224,
      "step": 6876
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.0612638687891945e-06,
      "loss": 0.0076,
      "step": 6877
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.0291043576137644e-06,
      "loss": 0.0178,
      "step": 6878
    },
    {
      "epoch": 9.95,
      "learning_rate": 9.969448464383342e-07,
      "loss": 0.0214,
      "step": 6879
    },
    {
      "epoch": 9.95,
      "learning_rate": 9.647853352629041e-07,
      "loss": 0.012,
      "step": 6880
    },
    {
      "epoch": 9.95,
      "learning_rate": 9.32625824087474e-07,
      "loss": 0.0002,
      "step": 6881
    },
    {
      "epoch": 9.95,
      "learning_rate": 9.004663129120439e-07,
      "loss": 0.0303,
      "step": 6882
    },
    {
      "epoch": 9.95,
      "learning_rate": 8.683068017366137e-07,
      "loss": 0.0462,
      "step": 6883
    },
    {
      "epoch": 9.96,
      "learning_rate": 8.361472905611836e-07,
      "loss": 0.0025,
      "step": 6884
    },
    {
      "epoch": 9.96,
      "learning_rate": 8.039877793857533e-07,
      "loss": 0.0119,
      "step": 6885
    },
    {
      "epoch": 9.96,
      "learning_rate": 7.718282682103233e-07,
      "loss": 0.0005,
      "step": 6886
    },
    {
      "epoch": 9.96,
      "learning_rate": 7.39668757034893e-07,
      "loss": 0.0239,
      "step": 6887
    },
    {
      "epoch": 9.96,
      "learning_rate": 7.075092458594629e-07,
      "loss": 0.0432,
      "step": 6888
    },
    {
      "epoch": 9.96,
      "learning_rate": 6.753497346840328e-07,
      "loss": 0.0311,
      "step": 6889
    },
    {
      "epoch": 9.96,
      "learning_rate": 6.431902235086027e-07,
      "loss": 0.0004,
      "step": 6890
    },
    {
      "epoch": 9.97,
      "learning_rate": 6.110307123331725e-07,
      "loss": 0.0014,
      "step": 6891
    },
    {
      "epoch": 9.97,
      "learning_rate": 5.788712011577424e-07,
      "loss": 0.0,
      "step": 6892
    },
    {
      "epoch": 9.97,
      "learning_rate": 5.467116899823123e-07,
      "loss": 0.0296,
      "step": 6893
    },
    {
      "epoch": 9.97,
      "learning_rate": 5.145521788068822e-07,
      "loss": 0.0104,
      "step": 6894
    },
    {
      "epoch": 9.97,
      "learning_rate": 4.823926676314521e-07,
      "loss": 0.0,
      "step": 6895
    },
    {
      "epoch": 9.97,
      "learning_rate": 4.5023315645602193e-07,
      "loss": 0.0004,
      "step": 6896
    },
    {
      "epoch": 9.97,
      "learning_rate": 4.180736452805918e-07,
      "loss": 0.0317,
      "step": 6897
    },
    {
      "epoch": 9.98,
      "learning_rate": 3.8591413410516164e-07,
      "loss": 0.0009,
      "step": 6898
    },
    {
      "epoch": 9.98,
      "learning_rate": 3.5375462292973146e-07,
      "loss": 0.0588,
      "step": 6899
    },
    {
      "epoch": 9.98,
      "learning_rate": 3.2159511175430134e-07,
      "loss": 0.0007,
      "step": 6900
    },
    {
      "epoch": 9.98,
      "learning_rate": 2.894356005788712e-07,
      "loss": 0.0295,
      "step": 6901
    },
    {
      "epoch": 9.98,
      "learning_rate": 2.572760894034411e-07,
      "loss": 0.0574,
      "step": 6902
    },
    {
      "epoch": 9.98,
      "learning_rate": 2.2511657822801097e-07,
      "loss": 0.0482,
      "step": 6903
    },
    {
      "epoch": 9.98,
      "learning_rate": 1.9295706705258082e-07,
      "loss": 0.0005,
      "step": 6904
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.6079755587715067e-07,
      "loss": 0.035,
      "step": 6905
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.2863804470172055e-07,
      "loss": 0.0175,
      "step": 6906
    },
    {
      "epoch": 9.99,
      "learning_rate": 9.647853352629041e-08,
      "loss": 0.0177,
      "step": 6907
    },
    {
      "epoch": 9.99,
      "learning_rate": 6.431902235086027e-08,
      "loss": 0.0254,
      "step": 6908
    },
    {
      "epoch": 9.99,
      "learning_rate": 3.2159511175430136e-08,
      "loss": 0.0063,
      "step": 6909
    },
    {
      "epoch": 9.99,
      "learning_rate": 0.0,
      "loss": 0.0,
      "step": 6910
    },
    {
      "epoch": 9.99,
      "step": 6910,
      "total_flos": 695877609873408.0,
      "train_loss": 0.14138391603609243,
      "train_runtime": 47099.76,
      "train_samples_per_second": 2.348,
      "train_steps_per_second": 0.147
    }
  ],
  "max_steps": 6910,
  "num_train_epochs": 10,
  "total_flos": 695877609873408.0,
  "trial_name": null,
  "trial_params": null
}